Intrastromal Application of Riboflavin for Corneal Crosslinking

Abstract

Purpose.: To experimentally evaluate the efficacy of corneal crosslinking (CXL) by injecting the photomediator riboflavin into the corneal stroma via intrastromal channels.

Methods.: Five groups of pig corneas, nine each, were compared regarding stress–strain relationship and UV-absorption. Group 1 had intrastromal channels floated with riboflavin 0.5%-solution followed by UVA-irradiation (3 mW/cm² for 30 minutes); group 2 was handled like group 1, but were irradiated with 9 mW/cm² for 10 minutes; group 3 was treated according to the Dresden protocol (epi-off, 9 mW/cm² for 10 minutes); group 4 had the identical channel system, no riboflavin but identical irradiation; group 5 with native corneas served as a control group. The intrastromal channels were created with a femtosecond laser. The stress–strain relations were measured in corneal strips using a uniaxial material tester at strains up to 12%. The UV-transmission of the corneas was measured in groups 1, 3, and 5.

Results.: The stress needed for a 10% strain was significantly increased by 82% in the corneas treated with the Dresden protocol compared with native cornea (P = 0.0005). With intrastromal application of riboflavin the significant increase was 87% (P = 0.0005) in group 1 and 64% (P = 0.007) in group 2. The channel formation alone did not alter biomechanics (P = 0.923). The corneal UVA-transmission was 2.4% after intrastromal riboflavin application, 8.9% after the treatment according to the Dresden protocol, and 57.9% in native corneas.

Conclusions.: The experiments demonstrate the intrastromal application of riboflavin by means of intrastromal channels a feasible “epi-on” approach for CXL. More experimental data are needed before clinical testing.

Introduction

More than 15 years ago, corneal cross-linking (CXL) by means of riboflavin and UV light was proposed as a therapeutic approach to improve the biomechanical and biochemical properties of the cornea. ^1,2 Meanwhile, there is clinical evidence that CXL is a clinically useful procedure halting the progression of primary as well as secondary keratectasia with a failure rate of approximately 3% and a complication rate of 1% or less. ^3–7  

The majority of the serious complications like infectious keratitis and ulceration ^8–11 are closely related with the abrasion, which raises the demand for “epi-on” techniques. Current presented epi-on approaches have also resulted in significantly shallower CXL-domains ¹² and a failure rate of up to 50%. ¹³ Other techniques such as iontophoresis ¹⁴ and new pharmacologic formulations of the riboflavin drops ^15–19 are still investigational. 

In this paper, we present experimental results regarding efficacy and safety of an intrastromal application of riboflavin via a channel system that is created by a femtosecond laser with minimal dissection of tissue. 

Methods

Specimen Preparation

Forty-five fresh porcine cadaver eyes with intact epithelium and clear corneas were collected from the local slaughterhouse within 3 hours post mortem and treated within a total of 6 hours. Central pachymetry (SP-100; Tomey, Nagoya, Japan) of every eye was performed before any treatment. The eyes were randomly assigned to five groups, the nine eyes of group 1 and group 2 received an intrastromal channel system that was cut with a UV-femtosecond laser (Wavelight; Alcon, Erlangen, Germany) at a depth of 300 μm. The channels were 0.4 mm in width and the channel-free area had a diameter of 3.4 mm (Figs. 1A, 1B). The channel pattern was roughly centered on the geometrical center of the cornea. The area that was dissected by the laser was 9.4% of a normal cornea with a diameter of 11 mm. Through the two radial channels (horizontal) a solution of 0.5% riboflavin-5-monophosphate in 0.9% NaCl (TCI, Tokyo, Japan) was injected into the channel system, so that it was completely filled. The eye was kept in a wet chamber to allow the riboflavin to diffuse through the stroma for 30 minutes. Thereafter, the eyes of groups 1 and 2 were irradiated with UVA-light (λ = 365 nm) with 5.4 J/cm² using the UVX 2000 lamp (IROC Innocross, Zürich, Switzerland). The corneas of group 1 were irradiated for 30 minutes with an irradiance of 3 mW/cm², those of group 2 received an irradiance of 9 mW/cm² for 10 minutes. The reduction in irradiance was achieved by a standard neutral filter. The stress–strain measurements were performed 60 minutes after irradiation. The nine eyes of the fourth group were treated exactly the same way, except that the injection was performed with 0.9% NaCl solution, not including riboflavin. In the third group, the eyes were de-epithelialized mechanically with a hockey knife and received drops of 0.1% riboflavin-5-monophosphate in 16% dextran T-500 (TCI) every 2 minutes. After 30 minutes the eyes were exposed to 5.4 J/cm² (9mW/cm² for 10 minutes) UVA-light (Dresden protocol). The eyes of the fifth group received no treatment and served as a control. All experiments were performed at room temperature. 

Figure 1

 

(A) Schematic. (B) Channels immediately after riboflavin injection.

At the end of the treatment, a corneoscleral disc was excised and a 5-mm wide strip was cut through the central part of the cornea parallel to the vertical axis, including the inner ring of the channel pattern. 

Stress–Strain Measurement

As a first step, central pachymetry was performed in all strips. The strips, including the inner ring channel, were clamped horizontally at a distance of 6 mm between the two jaws of the UStrech (Cellscale, Waterloo, ON, Canada). To expose the tissue to the physiological stress range as well as to start the measurement from an equal tension, a prestress of 5 × 10³ Pa was applied for 120 seconds (the steady state was achieved after 80 seconds). The strain was then increased linearly with a velocity of 0.035 mm/second similar to the experiments of Spoerl et al. ²⁰ and the stress was measured every 10 msec until a strain of 12%. The whole stress–strain measurement was performed in a bath containing 16% dextran solution to stabilize hydration of the corneal strip. At the end of the stress–strain measurement the thickness of the cornea was measured a second time to guarantee stabile hydration of the cornea. 

UVA Transmission

The UVA transmission at a wavelength of 365 nm was measured in treated and untreated corneas. Three groups five eyes each were formed according to groups 1, 3, and 5 described above. A corneoscleral disk was excised and the UVA transmission was measured at 5, 10, 20, and 30 minutes after the complete treatment. The UVA light intensity of the beam with a diameter of 3 mm was measured by means of an UVA detector (YK-35UV; Lutron Electronic, Taipei, Taiwan). The detector was located 1 mm in behind the specimen in order to measure all light transmitted through the cornea. 

Numerical Evaluation

Stress and strain were derived from force and elongation measurements in the standard manor. ²⁰ All measurement data were directly exported to a Microsoft excel file (Microsoft, Redmond, WA, USA) and averages and SDs of stress and strain were determined for strains ranging from 0% to 12%. The stress of each group at 6%, 8%, 10%, and 12% strain was compared using the Mann-Whitney U test (WINSTAT 2012.1; R. Fitch Software, Bad Krozingen, Germany). Ultraviolet-A transmissions were also compared using the Mann-Whitney U test. Significance was accepted for P less than 0.05. 

Results

Stress–Strain Measurements

The stress–strain curves showed the typical nonlinear behavior of a viscoelastic tissue (Fig. 2) best fit with polynomials of third order (e.g., R ² = 0.9997 for the Dresden protocol group). The stress at 6%, 8%, 10%, and 12% of all five groups is listed in Table 1. The increase in stress at a strain of 10% was 82% in the Dresden protocol group compared with 87% (group 1) and 64% (group 2) in the intrastromal riboflavin groups. These increases correspond to 4% higher and a 22% lower increase compared with the Dresden protocol. All increases in stiffness are statistically significant different from 0 (P = 0.0005, 0.0005, and 0.007). Group 4 (channels but no riboflavin) showed an increase in stress compared with the native cornea group, which was not statistically significant (P = 0.923). 

Figure 2

 

Stress–strain relation of groups 1 through 5. The cross-linked corneas of groups 1, 2, and 3 demonstrate a statistically higher stress–strain relationship compared with noncrosslinked corneas.

Table 1

 

Stress (Pa) at Strain 6%, 8%, 10%, and 12%

Strain	6%	8%	10%	12%
Group 1: intrastromal riboflavin and CXL 3mW/cm2 for 30 min	95′015 ± 12′617	169′166 ± 26′184	250′305 ± 33′369	337′465 ± 47′482
Group 2: intrastromal riboflavin and CXL 9mW/cm2 for 10 min	80′445 ± 23′512	145′552 ± 37′229	218′478 ± 46′588	303′587 ± 55′550
Group 3: Dresden protocol 9mW/cm2 for 10 min	92′601 ± 13′467	165′897 ± 15′902	243′659 ± 12′112	331′529 ± 18′424
Group 4: intrastromal NaCl	48′141 ± 18′529	85′805 ± 29′803	138′135 ± 40′908	204′737 ± 49′597
Group 5: native cornea	42′108 ± 14′109	79′824 ± 25′930	133′556 ± 37′987	200′002 ± 46′346

UVA Transmission

The UVA transmission was lowest in group 1 with 2.4% at 30 minutes after intrastromal riboflavin application (Table 2), significantly lower compared with native cornea 57.9% (P = 0.009) and corneas treated with the Dresden protocol with 8.9% (P = 0.009). 

Table 2

 

UVA Transmission (%), Average ± SD

	5 min	10 min	20 min	30 min
Group 1: intrastromal application	5.92 ± 0.81	4.16 ± 0.82	2.94 ± 0.36	2.35 ± 0.21
Group 3: Dresden protocol	24.99 ± 5.86	16.76 ± 4.29	11.86 ± 2.18	8.85 ± 2.05
Group 5: native cornea	59.81 ± 1.16	57.43 ± 3.03	58.46 ± 2.50	57.89 ± 1.89

Discussion

The major findings of this study were: (1) using intrastromal application of riboflavin for CXL an increase of stiffness of 87% respectively 64% was measured comparable to the stiffening achieved by the Dresden protocol, (2) the UVA transmission through cornea is significantly smaller compared with the Dresden protocol, (3) the creation of channels presented here does not result in a measurable weakening of the cornea. 

The first numerical information about the increase in biomechanical properties resulting from crosslinking using the Dresden protocol was published by Spoerl et al. ²¹ confirmed by Wollensak. ²² The increase in stress at 10% reported was approximately 77%, very close to the 82% presented in this paper. The intrastromal application of riboflavin has a comparable effect (87% and 64% vs. 82%). The stronger effect in the 3 mW/cm² group compared with 9 mW/cm² group, at equal total energy, reflects the previously documented stronger crosslinking in low irradiance treatments compared with high irradiance treatments: Hammer and coworkers ²³ found a more than double stiffening effect in corneas treated with 3 mW/cm² than in corneas treated with 9 mW/cm². Most probably, this stronger action corresponds with the longer irradiation time, because during that time more oxygen can diffuse into the cornea and replace the oxygen consumed during the crosslinking process. ²⁴ The smaller effect in the intrastromal 9 mW/cm²-group is not explained by the absorption of UVA-light of the epithelium because the pig epithelium alone absorbs at 365 nm only 5 ± 2% (own results, not published) and with 3 mW/cm² the difference between intrastromal application and “epi-off” was not significant. Such reduction in crosslinking may easily be compensated by an increase of energy density from 5.4 J/cm² to approximately 7.0 J/cm². 

The smaller transmission of UVA light in corneas with intrastromal application of riboflavin compared with the imbibition from the surface has the advantage of a better protection of the endothelium and more posterior structures of the eye. It is not clear whether this is a result of the higher concentration of riboflavin solution used (0.5% vs. 0.1%) or the different riboflavin concentration gradient inside the stroma. We chose 0.5% riboflavin-5-monophosphate concentration in order to optimize the stiffening effect because the stiffening effect is strongly dependent on the number of riboflavin molecules in the stroma. 

There is no significant difference in biomechanical properties due to the channel system. As the dissection is parallel to the surface of the cornea, there is only minimal cutting of collagen lamellae and, therefore, obviously only minimal impact on the biomechanical integrity of the cornea. Whether this is correct for keratoconus corneas is unclear and needs further investigation. 

After 30 minutes of riboflavin application at the surface of the cornea (epi-off) Geerling and coworkers ²⁵ found an almost linear decrease of riboflavin concentration from anterior to posterior. A similar decay was measured by Brillouin spectroscopy where the Brillouin shift regressed linearly from anterior to posterior. ²⁶ The Brillouin shift is related with biomechanical “bulk modulus,” ²⁷ which leads directly to a strong relation between the riboflavin concentration and the resulting increase in biomechanical stiffening. Spoerl and coworkers ²⁸ proved experimentally that during crosslinking by means of the Dresden protocol the anterior cornea is much more stiffened compared with the posterior part, which can be explained by Beer's law. By means of intrastromal application of riboflavin the usual riboflavin gradient may be modified and the highest concentration of riboflavin (and, therefore, the biomechanical stiffening) may not be necessarily located at the anterior surface of the cornea but can be moved to deeper layers. From a biomechanical point of view this is an interesting approach because it allows improving stiffness of the naturally weaker layers of the cornea, which are located in the posterior stroma. ²⁶ In an ongoing study, we are currently investigating the influence of the depth of the channel system on the crosslinking effect. 

The Dresden protocol including abrasion of the epithelium is considered the gold standard of CXL, however, this epi-off technique is plagued with negative side effects such as pain and perioperative infections. ^7–11,29 To avoid these side effects various approaches of transepithelial riboflavin imbibition have been investigated. A presumed loosening of epithelial tight junctions resulted in a reduced crosslinking effect with a depth as small as 100 μm. ¹² Consequently, this led to a reoperation rate of up to 50% after 24 months. ¹³ Another approach for transepithelial CXL is iontophoresis, which has not yet gained clinical approval. ¹⁴ The intrastromal channel method presented here may prove to be a more advantageous epi-on technique. 

Intrastromal application has been proposed before ^30,31 where riboflavin was injected in femtosecond laser–prepared pockets of 7-mm diameter ³⁰ or even not reported geometrical dimensions. ³¹ Although surface-parallel cut such large dissections may jeopardize the biomechanical stability of an anyway weak cornea, however, neither Kanellopoulos ³⁰ nor Alió and coworkers ³¹ did present any biomechanical basic research. Experimental data were presented by Wollensak and coworkers ³² indicating a substantial weakening of more than 10% due to the dissection of more than 50% of corneal area. The channel system proposed here undermines only 9% of the corneal area and in a healthy cornea, this has no measurable effect on corneal biomechanics. 

The experiments presented in this article demonstrate the feasibility of intrastromal application of riboflavin for CXL in two ways: first, the stiffening effect is comparable with the standard epi-off technique, and second, the riboflavin-shielding effect is as least as efficient compared with the accepted Dresden protocol. Clearly, we need more experimental information about the optimal depth of the channels, potential higher order aberrations induced by the channel system and potential endothelial damage before this approach can be tested clinically. 

Acknowledgments

The authors thank Jörg Klenke, Johannes Lörner, Yao Zhang, and Christian Wüllner (Wavelight; Alcon, Erlangen, Germany) for their support during the preparation of the intrastromal channel system using the femtosecond laser. 

Disclosure: T.G. Seiler, P; I. Fischinger, None; T. Senfft, None; G. Schmidinger, None; T. Seiler, Alcon (F, C), Wavelight (C), P