Keratoconus is a bilateral progressive corneal ectatic disease inducing visual impairment due to irregular astigmatism or corneal opacities,1 which may require corneal transplantation in severe forms. Corneal collagen cross-linking (CXL) is a surgical technique first described by Wollensak et al.2,3 aimed at stabilizing the cornea and halting progression of the keratoconus.
In conventional CXL (C-CXL), the corneal stroma is soaked with a riboflavin solution before exposure to a uniform beam of ultraviolet-A (UVA) radiation. However, biomechanical analyses of keratoconic corneas have shown that weakening is concentrated within the area of the keratoconic cone.4,5 A CXL procedure targeted to stiffen this area specifically presents an interesting strategy to redistribute biomechanical stresses in the cornea. This topography-guided CXL (TG-CXL) demonstrates potential to optimize the standard procedure and improve the quality of vision by normalizing the corneal curvature.6 Thus, we conducted a prospective clinical study to compare TG-CXL to C-CXL.
Patients and Methods
This prospective, comparative, nonrandomized clinical study was conducted in the National Reference Keratoconus Center (Department of Ophthalmology, Purpan Hospital, Toulouse, France). Approval was obtained from the Ethics Committee of the French Society of Ophthalmology (IRB 00008855 Société Française d'Ophtalmologie IRB#1) and conducted in accordance with the tenets of the Declaration of Helsinki.
We included 60 eyes of 60 patients between November 2014 and July 2015. Thirty eyes were enrolled in the TG-CXL group. After completion of the study group, standard C-CXL was applied to a control group of 30 eyes matched to the study group for age, sex, keratoconus stage, maximum and minimum keratometry values (Kmax and Kmin), and uncorrected distance and corrected distance visual acuity (UDVA and CDVA).
Inclusion criteria were age 16 years or older; progressive keratoconus, defined as an increase of Kmax greater than 1.00 diopters (D) in the previous 12 months or less; and central corneal thickness of 400 μm or greater. Patients signed an informed consent form prior to enrollment in the study. Patients younger than 18 years signed an assent form in addition to having a parent or legal guardian sign an informed consent form. Patients with other ocular diseases, aphakic or pseudophakic eyes, history of chemical injury, or previous corneal surgery or insertion of corneal rings in the eye to be treated were excluded.
Patients were examined preoperatively and at 2 days (for therapeutic lens removal and to check for early complications), 1, 3, and 6 months, and 1 year postoperatively. Follow-up measurements included corneal topography (WaveLight Oculyzer II; Alcon Laboratories, Inc., Fort Worth, TX), optical coherence tomography (OCT) (Spectralis; Heidelberg Engineering, Heidelberg, Germany), confocal microscopy (HRT Rostock Cornea Module; Heidelberg Engineering), specular microscopy (SP 2000P; Topcon, Tokyo, Japan), UDVA, CDVA, and slit-lamp examination.
C-CXL Procedure. The C-CXL was performed in accordance with the Dresden protocol.2 After administration of a topical anesthetic (Tetracaïne; Novartis, Basel, Switzerland), the central 9-mm epithelium was removed mechanically with an Amoils epithelial scrubber (Innovative Excimer Solutions, Toronto, Canada). One drop of riboflavin 0.1% (Ricrolin; Sooft, Rome, Italy) was instilled every minute for 20 minutes. The stroma was irradiated with a UVA light source (VEGA CBM-X-Linker; C.S.O., Florence, Italy) at 3 mW/cm2 for 30 minutes.
TG-CXL Procedure. After administration of topical anesthetic (Tetracaïne), the epithelium was removed mechanically over the region designated for UVA treatment. Riboflavin 0.1% (VibeX Rapid; Avedro, Waltham, MA) was instilled on the debrided stroma every 2 minutes for 10 minutes. A UVA irradiation pattern was programmed for each patient using the CE marked (EU1507401) KXL II device (Avedro), designed as three superimposed concentric circular zones centered on the maximum posterior elevation determined by Oculyzer posterior float. The diameter of the inner region encompassed the area of abnormal posterior elevation and the location of Kmax, whereas the outer circles were placed to encompass the remaining region of abnormal anterior axial curvature. Transitions between zones occurred at the regions of greatest change in anterior axial corneal curvature. Total energy delivered across the treatment zone ranged from 5.4 J/cm2 in the outermost zone to 15 J/cm2 in the innermost zone (Figure 1), delivered using 30 mW/cm2 UVA, pulsed at 1-second intervals.
Example of customized treatment on topography map. The maximum ultraviolet-A energy is focused on top of the cone (15 J) and decrease gradually (from 10 to 5.4 J) to the periphery.
At the end of the C-CXL or TG-CXL procedures, a topical antibiotic (Quinofree; Laboratoires Théa, Clermont-Ferrand, France) and a therapeutic contact lens were applied. The postoperative treatment included an antibiotic (Quinofree) for 7 days and an anti-inflammatory drop (Ocufen; Horus Pharma, Saint Laurent du Var, France) for 15 days, which was started at 1 week postoperatively.
The main outcome measure was comparison of Kmax between 1 year and preoperative values on anterior axial curvature map.
Secondary outcome measures included the following: qualitative analysis of modifications from difference map (between 1 year and preoperative anterior axial curvature map) and quantitative evaluation using a corneal normalization analysis: an S index (superior index) corresponds to the mean keratometry values from 5 points on the top of the 3-mm diameter circle (crossing the 30°, 60°, 90°, 120° and 150° axes) and an I index (inferior index) corresponds to the mean keratometry values of 5 points at the bottom of the same circle (crossing 210°, 240°, 270°, 300°, and 330° axes) (Figure A, available in the online version of this article). A normalization analysis was performed corresponding to the change in I index and S index at 1 year compared to preoperative values. UDVA and CDVA assessment using a standardized scotopic decimal projection chart at a viewing distance of 5 m converted into logMAR notation. Demarcation line depth was observed by OCT. We compared the region of the keratoconus cone (cone area) to the “normal” cornea 180° away (surrounding area). Nerves and cell densities were analyzed by confocal microscopy at the cone area and surrounding area. Two independent blinded examiners (MC and KP) studied the images using Neuron J software (National Institutes of Health, Bethesda, MD) to evaluate the nerve density and used Image J software (National Institutes of Health) to measure keratocyte density.
Superior (S) and inferior (I) index calculated on a topography map.
The presence of the demarcation line observed with OCT was compared using the chi-square test. Comparisons of other parameters between the two groups were performed using the Student's t test. Data were expressed as average ± standard deviation and a P value of less than .05 was considered statistically significant.
Baseline characteristics were comparable between the two groups (Table 1).
Patient Demographics and Ocular Characteristics
The qualitative analysis of differential maps showed apical flattening of the cone. A significant decrease of Kmax was observed in the TG-CXL group at 6 months (−1.29 ± 2.44 D [range: −7.90 to 4.70 D], P = .0069) and 1 year (−1.07 ± 1.70 D [range: −4.40 to 1.60 D], P < .001), with no significant changes in C-CXL eyes (0.44 ± 1.61 D [range: −3.00 to 3.40 D], P = .2282 and 0.40 ± 1.75 D [−3.30 to 4.10 D, P = .2598], respectively), representing a statistically significant difference between the two groups (P < .01). The percentage of eyes decreasing by 1.00 D or more was significantly higher in the TG-CXL than in the C-CXL group (P < .02) (Figure 2).
Histogram of the postoperative maximum keratometry (Kmax) values at 1 year in terms of diopters changes. TG-CXL = topography-guided corneal collagen cross-linking; C-CXL = conventional corneal collagen cross-linking
We observed a significant decrease of I index at 6 months (−1.017 ± 1.367 [range: −5.720 to 0.6, P < .001]) and 1 year (−0.966 ± 1.204 [range: −3.83 to 1.02], P < .001) in the TG-CXL group, whereas there were no significant changes in S index (0.063 ± 1.943 [range: −8.02 to 2.740], P = 0.859 and 0.645 ± 1.832 [range: −5.22 to 7], P = .059, respectively). In the C-CXL group, there was no significant change in either I index (0.87 ± 2.369 [range: −2.260 to 8.360], P = .085 at 6 months and 0.50 ± 2.423 [range: −2.450 to 8.32], P = .281 at 1 year) or S index (0.580 ± 2.03 [range: −2.120 to 8.94], P = .1748 at 6 months and 0.636 ± 2.242 [range: −3.15 to 9.88], P = .145 at 1 year).
Visual Acuity Outcomes
At 1 year, UDVA and CDVA improved to 0.6540 ± 0.4036 (P = .557) and 0.2162 ± 0.2495 (P < .05) logMAR, respectively, in the TG-CXL group and 0.6149 ± 0.3515 (P = .197) and 0.2648 ± 0.2574 (P = .104) logMAR, respectively, in the C-CXL group. An improvement of at least one line on the Monoyer scale occurred in 45.6% of the C-CXL group and 64.5% of the TG-CXL group. No changes in visual acuity were observed in 24.1% of eyes in the C-CXL group and 12.9% of eyes in the TG-CXL group (Figure 3). A loss of at least one line on the Monoyer scale occurred in 30.3% in the C-CXL group and 22.6% in the TG-CXL group. There were no statistically significant differences between the two groups in postoperative CDVA.
The bar graph depicts the changes in corrected distance visual acuity from the preoperative examination to the 1-year postoperative examination in terms of the number of Monoyer lines changes. TG-CXL = topography-guided corneal collagen cross-linking; C-CXL = conventional corneal collagen cross-linking
OCT Demarcation Line
In both groups, the demarcation line was observed at 1 month postoperatively at a similar depth in the cone area (Figure 4). In the C-CXL group, the mean depth was equivalent between the cone area and the surrounding area (232 ± 43 and 245 ± 73 μm, respectively, P = .391). However, in the TG-CXL group, we found a deeper demarcation line in the cone area than in the surrounding area (242 ± 46 and 182 ± 59 μm, respectively, P < .0001). After 1 month, no visible demarcation line was observed in either group.
Examples of corneal stromal demarcation line in both groups at 1 month. In the topography-guided corneal collagen cross-linking (TG-CXL) group, the mean demarcation line depth is more important in the cone area than in the surrounding area. In the conventional corneal collagen cross-linking (C-CXL) group, the demarcation line is as deep in the cone area as in the surrounding area.
Confocal microscopy results are listed in Table A (available in the online version of this article).
Confocal Microscopy Follow-up Regarding Nerve and Cell Densities
Nerve density decreased in both groups within the cone area. Nerve healing occurred progressively over several months without a significant difference between the TG-CXL and C-CXL groups at 1 month (1.94 ± 3.45 vs 1.81 ± 1.10 nerves/mm2, P = .900), 3 months (2.02 ± 2.29 vs 2.58 ± 1.73 nerves/mm2, P = .574), 6 months (6.93 ± 3.59 vs 5.64 ± 3.44 nerves/mm2, P = .423), and 1 year (12.35 ± 3.02 vs 10.17 ± 1.02 nerves/mm2, P = .075).
In the TG-CXL group, nerve density within the untreated surrounding area was comparable to the preoperative nerve density (5.90 ± 5.04 vs 13 ± 4.69 nerves/mm2, P < .01) at 1 month. This denervation was less significant than was observed in the cone area of the TG-CXL group (1.94 ± 3.45 nerves/mm2, P < .001) or the surrounding area of the C-CXL group (2.52 ± 1.61 nerves/mm2, P < .001) (Figures 5A–5B).
(A–B) Examples of corneal confocal microscopy images in both groups, showing nerve density in the surrounding area at 1 month. In the topography-guided corneal collagen cross-linking (TG-CXL) group, mean nerve density decrease was less than in conventional corneal collagen cross-linking (C-CXL). (C–D) Examples of corneal confocal microscopy images of anterior keratocytes in both groups, showing the surrounding area at 1 month. Mean cells density decrease was less in the TG-CXL group than in the C-CXL group.
The nerve density in the C-CXL group was similar in the surrounding area and the cone area at 1 month (2.52 ± 1.61 vs 1.81 ± 1.10 nerves/mm2, P = .144), 3 months (2.36 ± 1.65 vs 2.58 ± 1.73 nerves/mm2, P = .563), 6 months (5.77 ± 3.05 vs 5.64 ± 3.44 nerves/mm2, P = .862), and 1 year (10.1 ± 1.09 vs 10.17 ± 1.02 nerves/mm2, P = .713).
In the TG-CXL group, the difference between the surrounding area and the cone area was statistically significant at 1 month (5.90 ± 5.04 vs 1.94 ± 3.45 nerves/mm2, P < .001) and 3 months (5.98 ± 4.65 vs 2.02 ± 2.29 nerves/mm2, P < .001). This difference was no longer significant at 6 months (7.29 ± 3.72 vs 6.93 ± 3.59 nerves/mm2, P = .617) and 1 year (12.01 ± 2.94 vs 12.35 ± 3.02 nerves/mm2, P = .285).
Within the cone area, cellular apoptosis was comparable between the TG-CXL and C-CXL groups at 1 month (205.5 ± 70.1 vs 179.4 ± 77.4 cells/mm2, P = .335), 3 months (215.1 ± 71.1 vs 238.1 ± 40.2 cells/mm2, P = .489), 6 months (266.4 ± 63.4 vs 264.1 ± 111.4 cells/m2, P = .95), and 1 year (309.7 ± 57.5 vs 286.6 ± 43.7 cells/mm2, P = .344).
In the TG-CXL group, we observed a loss of cells at 1 month compared to the preoperative keratocyte density within the surrounding area (264.3 ± 84.5 vs 328.4 ± 116.6 cells/mm2, P < .02). However, cellular apoptosis in the surrounding area was less than in the cone area of the TG-CXL group (205.5 ± 70.1 cells/mm2, P < .001) and the surrounding area of the C-CXL group (176.3 ± 65 cells/mm2, P < .01) (Figures 5C–5D).
The cell density in the C-CXL group was similar in the surrounding area and the cone area at 1 month (176.3 ± 65 vs 179.4 ± 77.4 cells/mm2, P = .787), 3 months (230.6 ± 36.4 vs 238.1 ± 40.2 cells/mm2, P = .528), 6 months (271.4 ± 72.8 vs 264.1 ± 111.4 cells/mm2, P = .683), and 1 year (295.5 ± 54.2 vs 286.6 ± 43.7 cells/mm2, P = .504).
In the TG-CXL group, the difference between the surrounding and the cone areas was statistically significant at 1 month (264.3 ± 84.5 vs 205.5 ± 70.1 cells/mm2, P < .004) and 3 months (271.3 ± 101.8 vs 215.1 ± 71.1 cells/mm2, P < .002). This difference was no longer significant at 6 months (284.6 ± 46.9 vs 266.3 ± 63.4 cells/mm2, P = .443) and 1 year (310.6 ± 67.8 vs 309.7 ± 57.5 cells/mm2, P = .872).
In the TG-CXL group, we noted an asymmetric haze, more pronounced on the cone area than on the surrounding side (Figure B, available in the online version of this article), whereas it was symmetric in the C-CXL group.
Photography of an asymmetrical haze 1 month after topography-guided corneal collagen cross-linking treatment.
Similar ocular side effects were observed in both groups and included itching, pain, and blurry vision that quickly improved within the first week on reepithelialization. There were no persistent ocular side effects.
Endothelial cell count was stable among all follow-up visits in the TG-CXL group (2,684 ± 276 cells/mm2 preoperatively and 2,855 ± 497 cells/mm2 at 1 year, P = .146) and C-CXL group (2,669 ± 233 cells/mm2 preoperatively and 2,688 ± 281 cells/mm2 at 1 year, P = .917).
Although histological studies have shown that keratoconic histopathological changes involve the whole cornea,1 clinical manifestations of keratoconus are typically limited to the inferior temporal cornea,7,8 with normal curvature of the superior cornea observed with topography. Additionally, studies with Brillouin biomicroscopy,4,5 biomechanical mapping of cornea, suggest that keratoconus is a focal disease.
C-CXL has been demonstrated to slow or halt the progression of the ectatic disease.9,10 However, in 2015 the Cochrane analysis concluded there was limited evidence to support the use of CXL in the management of keratoconus due to the lack of properly conducted randomized controlled trials.11 Nevertheless, many recent studies have contributed additional direct12–15 and indirect16 data in favor of CXL efficacy.
Hence, it seemed logical to specifically target the weakened area of the cornea. Thus, we performed a prospective study comparing TG-CXL to C-CXL. To our knowledge, the current study is the first report of this customized therapeutic CXL.
TG-CXL UVA treatment profiles included three concentric circular zones centered on maximum posterior elevation to mimic a graded treatment pattern based on the theoretical model proposed by Sinha Roy and Dupps, which suggests that spatial variation of UVA intensity results in the greatest flattening effect.6
Our results demonstrate that TG-CXL flattens the cone at 6 months and 1 year, with significant decrease in Kmax (−1.29 ± 2.44 and −1.07 ± 1.70 D, respectively), whereas eyes treated by C-CXL did not achieve the same effect. However, this study only includes 1-year data and we will continue to observe these patients because, as in many studies,13–15 a further decrease in Kmax may be seen later. TG-CXL could improve CXL directly by specifically treating the diseased area or indirectly by limiting healing responses, which delay flattening.
The corneal normalization analyses, determined by the I and S index,17,18 lead us to consider corneal reshaping. In the TG-CXL group, a statistically significant decrease in I index was observed, with no increase in S index, demonstrating that flattening of the cone area does not induce a bulge of the surrounding area.
Flattening of the cone in eyes treated by TG-CXL is associated with a significant improvement in visual acuity. Optimizing the UVA treatment profile may induce greater flattening and consequently greater improvement in visual acuity.
An indirect clinical outcome of CXL efficacy is the corneal demarcation line observed by OCT.19 It is best viewed at 1 month postoperatively and its depth was not statistically different between TG-CXL and C-CXL within the cone area, indicating comparable efficacy of the new TG-CXL technique. Epithelium-off CXL induces significant alterations in stromal corneal components.19–23 Nerve and cell density decreased immediately after treatment and recovered progressively for both the TG-CXL and C-CXL groups in the cone area. The effect on cell and nerve density was similar in both the surrounding area and the cone area in the C-CXL group, whereas in the TG-CXL group we observed biological differences between the cone area and the surrounding area, which did not receive UVA irradiation. Within this surrounding area, a shallower and less pronounced demarcation line was observed than in the treated area, with less significant denervation and cell apoptosis. The cells and nerves recovered more quickly in the surrounding area than in the cone area in the TG-CXL group. Given that ocular surface parameters (eg, dry eye and tear break-up time) depend in part24 on corneal nerve status and keratocyte activation, we anticipate quicker recovery.20,25
Finally, it is surprising that we saw changes superiorly in the TG-CXL group because this part of the cornea had not been deepithelialized and did not receive UVA. Three complementary considerations could explain this biological gradient effect.
It is well known that CXL leads to chemokine release.26 Corneal stromal architecture is a collagen network27–29 that can enable chemokines to diffuse into the untreated corneal area and induce local damages. This paracrine effect could be described as a “sponge” effect, and is probably insufficient alone to explain the biological alterations.
Riboflavin may diffuse upward in the corneal stroma and environment exposure to UV irradiation may have occurred because patients were not advised to wear sunglasses beyond the first few hours after treatment.
A third explanation could be mechanical bias due to the size of the confocal microscopy optical head. Although the optical window enables a 400 × 400 μm2 image, the optical head is too big to focus accurately on a precise area.
We did not conduct a randomized study due to limited access to the KXL II device. Although the patients undergoing TG-CXL were paired with patients matched for age, sex, keratoconus stage, Kmax, Kmin, UDVA, and CDVA undergoing C-CXL, this study design could constitute a limitation compared with a randomized approach. Other randomized studies including a largest cohort and a longer follow-up should confirm these encouraging results.
Our study demonstrates that TG-CXL induces a gradient in the biological response to treatment from the cone area to the surrounding area and a flattening effect that is more significant than with C-CXL. Although longer follow-up and a larger cohort are necessary, this technique, which can be optimized in the future, seems to be a promising procedure, which results in greater flattening of the cone than C-CXL and consequently improves visual acuity.