Equivalence of Biomechanical Changes Induced by Rapid and Standard Corneal Cross-linking, Using Riboflavin and Ultraviolet Radiation

Abstract

Purpose.: Ultraviolet (UV) corneal cross-linking is an accepted method for treating corneal ecstatic disorders. The authors evaluated whether a rapid treatment protocol (higher intensity and shorter irradiation time) could achieve the same increase in corneal stiffness as the currently used standard protocol.

Methods.: Stress–strain measurements were performed on porcine corneal strips. The corneas (n = 72) were cut into three strips, each randomly receiving a different treatment: rapid (10 mW/cm², 9 minutes), standard (3 mW/cm², 30 minutes), or no (control, 0 mW/cm²) irradiation. After irradiation, the Young's modulus of each strip was determined. The results of the stress–strain measurements were analyzed statistically.

Results.: Statistical analysis showed that, after irradiation, the median value of Young's modulus from both active treatment groups (rapid, 3.83 N/mm²; standard, 3.88 N/mm²) was significantly higher (P < 0.05) than that of the control group (2.91 N/mm²). Treatment increased Young's modulus by a factor of 1.3. However, there was no significant difference (P = 0.43) between the rapid and standard groups in the median of Young's modulus.

Conclusions.: Rapid UV cross-linking treatment can be regarded as equivalent to the standard procedure in terms of increase in corneal stiffness. The new rapid protocol shortens the treatment duration by more than two thirds, from 30 to 9 minutes. The safety of the higher intensities must be addressed in further clinical studies.

Ultraviolet (UV) cross-linking (CXL) is an established method for the treatment of corneal ectatic disorders, such us keratoconus, pellucid marginal degeneration, and iatrogenic keratectasia after laser in situ keratomileusis (LASIK). ^{1 –7} Since its first introduction by Seiler and Spoerl ⁸ in 1997, a standardized protocol has been developed ⁹ : An abrasion of the corneal epithelium is performed, followed by riboflavin application. During the treatment phase, a recommended illumination intensity of 3 mW/cm² is applied to a 9-mm zone in the cornea for 30 minutes. This intensity corresponds to a total energy dose of 3.4 J or a radiant exposure of 5.4 J/cm². The photochemical process that induces the additional cross-links between the corneal fibers is thereby dependent on the applied radiant exposure of UV light. ^{8 –10}  

A specific radiant exposure can be achieved by either applying low intensity for a longer illumination time or higher intensity for a shorter time; theoretically, the same results can be achieved with different sets of illumination parameters. To address the patient's wish for a shorter treatment duration and the surgeon's need for a higher throughput of patients, a shorter CXL procedure would be desirable. However, no experimental data have yet been published that demonstrate the equivalence of the biomechanical stability of the corneal stroma between a standard (3 mW/cm² and 30 minutes) and a rapid cross-linking procedure that uses higher intensities and shorter treatment times. The aim of this work was therefore to experimentally investigate the equivalence of the biomechanical stability of porcine corneas by increasing the intensity to 10 mW/cm² and simultaneously decreasing the treatment time to 9 minutes (i.e., to less than one third), achieving the same radiant exposure of 5.4 J/cm². 

Methods

Experimental Study Design

To prove that the rapid method was equivalent to the standard, we investigated the change in corneal stiffness that was evoked by the cross-linking treatment of ex vivo porcine corneal tissue. Each cornea was cut into three strips, which were later randomly assigned to one of the following three groups, to minimize systematic errors due to sample preparation. The standard group received standard parameters (i.e., 3 mW/cm², 30 minutes), the rapid group was treated with the new parameters using higher irradiation intensity and shorter times (10 mW/cm², 9 minutes), and the control group was not irradiated with UV light at all, but still underwent the same procedure. The experiment involved three steps: sample preparation, sample treatment with UV irradiation, and stress–strain measurements. 

Sample Preparation

Some 72 porcine eyes were enucleated postmortem at a local abattoir and prepared for the experiments within 8 hours after the pigs were killed. All corneas were undamaged, were clear, and possessed no visual autolytic changes, haze, or scratches. In a first step, saline solution was injected into the eye bulb. This process not only compensated for the loss of intraocular pressure due to dehydration, but also facilitated subsequent epithelial abrasion and good riboflavin diffusion into the sample. After successful abrasion, the corneal tissue was extracted from the eye bulb, and its thickness was determined by five pachymeter (SP-2000; Tomey, Nagoya, Japan) measurements. With a four-blade knife, each sample was cut into three strips measuring 1 mm in width and 14 mm in length. All three strips were next placed into small, numbered containers filled with 0.1% riboflavin solution mixed with one part vitamin B₂-riboflavin-5-phosphate 0.5% (G. Streuli & Co AG, Uznach, Switzerland) and four parts 20% dextran T-500 (Carl Roth GmbH & Co. KG, Karlsruhe, Germany). The samples were kept cool overnight in 0.1% riboflavin solution, to ensure a homogenous distribution of riboflavin. 

The average thickness of the porcine corneas used in this investigation was 993 ± 75 μm and ranged from 750 to 1161 μm. The standard deviation of the five thickness measurements of each single cornea was 21 μm, corresponding to an error of approximately 2% on the cross-sectional area. This variability in the average corneal thickness thus adds an additional 2% error to the measured stress–strain results. 

Cross-linking Procedure

The cross-linking treatments were performed with two modified UV lamps (UV-X; IROC AG, Zurich, Switzerland) that permitted the irradiation intensity to be set between 1 and 10 mW/cm². Table 1 shows the treatment parameters of the standard and rapid cross-linking treatment. 

Table 1.

 

Treatment Parameters of the Rapid and Standard Cross-linking Procedure

Treatment Method	Intensity (mW/cm2)	Treatment Time (min)	Dose (J/cm2)
Standard	3	30	5.4
Rapid	10	9	5.4

The UV lamps were calibrated with a power meter (LaserMate-Q; Coherent, Santa Clara) to 3 and 10 mW/cm², respectively. For the treatment, the UV lamp was placed over the corneal strips, according to the manufacturer's instructions, and the maximum internal aperture of 9 mm ensured full irradiation of the sample area. 

The three samples from each cornea were treated in the same order, to ensure separation from other corneas. All three strips were taken from the container and randomly placed in one of the three different groups. Thereby, it is hoped that the free-floating strips within the containers were randomly mixed as they were transported from one laboratory to the other. To avoid dehydration during the irradiation phase, samples were kept moist by the repeated application of 0.1% riboflavin solution (every 5 minutes). The control group sample was placed under a black box to avoid any interaction with light. The other two samples were placed underneath the UV lamps. The rapid group was irradiated with an intensity of 10 mW/cm² for 9 minutes and then clamped for biomechanical evaluation. The stress–strain measurement was performed on this sample and next on the control group (∼20 minutes). Finally, after 30 minutes of irradiation time, the standard group sample, which had been irradiated with 3 mW/cm², was measured. 

Biomechanical Measurements

Corneal stiffness was determined by stress–strain measurements with a commonly used material-testing machine (Materialprüfmaschine MPM-145670; Zwick GmbH & Co., Ulm-Einsingen, Germany) adapted for small biological samples. The corneal samples were fixed in a small clamp designed in-house (Fig. 1a), which used a specially designed spacer for optimal tissue alignment (Fig. 1b). After the clamped sample had been tightened by two screws, it was then placed in the material-testing machine (Fig. 1c). In this procedure, a uniform sample area of 7 mm (length) × 1 mm (width) of the 14 × 1-mm sample was subjected to stress–strain testing. The thickness was individually checked during sample preparation. The proportion of 7:1 was kept to ensure that the force stayed parallel to the stretching direction to minimize deformation on the sample rims. 

Figure 1.

 

(a) A 7-mm long corneal sample (3) held by the clamp (1); (b) a sample inserted into a clamp (1) with the help of a spacer (2); (c) material-testing machine.

For each measurement, the following procedure was used: The lower crosshead of the material-testing machine is moved until the load sensor (force sensor) detects a force (load). As soon as resistance is detected, its position is recorded, and the reference force is calibrated to 0. Subsequently, the sample is preconditioned by applying a force of 0.1 N five times. This corresponds to a strain of between 1% and 5%, depending on the sample. Afterward, the actual measurement is performed. The sample is stretched with a velocity of 10 mm/min, up to a maximum strain of 12%. During the measurement, the load from the strain is automatically recorded and converted to stress by dividing it by the cross-sectional area (i.e., the sample's width times its thickness). Figure 2a shows a typical stress–strain curve. The stress–strain relationship of biological (corneal) tissue can be generally divided into the following regions (Fig. 2b): Region 1 is an area of nonlinear elasticity; region 2 shows linear elastic behavior; and region 3 is where irreversible plastic deformation occurs. 

Figure 2.

 

(a) Typical stress–strain curve; (b) regions of different elasticity.

The region of low strain, from 2% to 6%, was the area of interest in this investigation. The stress–strain curve was fitted to an exponential function in this region  

 

where σ is the stress, ε is the strain, and a and b are the fitting parameters. The stiffness (Young's modulus), E, is the first derivative of this function:  

 

For the following statistical analysis, Young's modulus was consistently evaluated at 4% strain. 

Statistical Analysis

Using a standard power calculation for normally distributed data, the following minimal number of samples per group was necessary to determine the statistically significant differences between two treatment methods (using a two-sided t-test):  

 

where z is the quantile of the normal distribution, σ_SD is the standard deviation, and L is the maximum allowed difference between the means of the two groups. Our calculation was based on an α = 0.05 level of significance and a statistical power of 1 − β = 0.9. The standard deviation, σ_SD = 31%, was taken from an older study of corneal cross-linking on porcine eyes for Young's modulus at 4% strain. ¹¹ Because of the large variability found in biological tissues, a difference of L = 20% was chosen, thereby resulting in a minimum of n = 52 samples per group. 

All groups were tested for normality using the Kolmogorov-Smirnov test and Lilliefors' correction. If not all groups were normally distributed, the nonparametric Friedman test was used to test the null hypothesis that the medians in all three groups and again the medians of the two treatment groups did not differ significantly. Equivalence was established by showing that the two-sided 95% confidence interval of the differences between rapid and standard treatment of the individual corneae was entirely within an interval (−Δ, +Δ). ^12,13 The margin of clinical equivalence, Δ, was set at 1.5 times the standard deviation obtained from the published literature, ¹¹ as no other equivalence value has yet been established. The equivalence parameter for Young's modulus was thus defined as Δ = 0.66 N/mm². 

Results

Excluded from the data analysis were 12 samples where the untreated samples measured Young's modulus greater than 5 N/mm² or where a procedural error systematically occurred (e.g., slip of the probe in the probe holder). Thus, only 60 of the 72 sample triplets were used for data analysis. 

Figure 3 shows the average Young's modulus of each group as a box plot. The median (M) of the untreated group M_untreated = 2.91 N/mm² is lower than the medians of the treated groups, which are M_rapid = 3.83 N/mm² for the rapid and M_standard = 3.88 N/mm² for the standard group. Consequently, the rapid group's median increased by a factor of 1.32, and the standard group's median rose by a factor of 1.33, compared with that of the untreated group. Figure 4 shows the average increase in Young's modulus of both treated groups compared with that of the untreated control. The medians of the average increase and the interquartile ranges are of comparable size. 

Figure 3.

 

Box-and-whisker plot of Young's modulus of the three sample groups. Light gray, untreated group; dark gray, rapid group; medium gray, standard group. −, min/max value; ×, 1%/99% value; ■, mean value.

Figure 4.

 

Box-and-whisker plot of the difference in Young's modulus between treated and untreated samples. Dark gray, rapid minus untreated; medium gray, standard minus untreated; −, min/max value; ×, 1%/99% value; ■, mean value.

The difference between the median values of all three groups was found to be statistically significant (P < 0.05, Friedman test). In contrast, no statistical significance was found when the two treatment groups were compared (P = 0.43, Friedmann test). The test for equivalence of the difference between the two treatment methods showed also that the methods are equivalent in terms of an increase in stiffness. The average difference in Young's modulus of the two methods is 0.16 ± 1.70 N/mm², with a 95% confidence interval of (−0.28 to 0.60 N/mm²). Thus, the lower bound of the confidence interval is larger than the negative of the equivalence parameter (−0.28 N/mm² > −Δ = −0.66 N/mm²) and the upper boundary of the confidence interval is smaller than the equivalence parameter (0.60 N/mm² < Δ = 0.66 N/mm²). Therefore, the performed study shows, in a statistically significant manner, the equivalence of both the rapid and standard UV corneal cross-linking procedure. 

Discussion

The presented data show the equivalence of the standard and rapid CXL procedures—that is, the increase in biomechanical stability. The rapid group, with its increased illumination intensity of 10 mW/cm² and three times shorter illumination time of 9 minutes, demonstrated an increase in Young's modulus by a factor of 1.3. This increase was statistically equivalent to that in the standard treatment group, having an intensity of 3 mW/cm² and a required illumination time of 30 minutes. 

The reported increase in stability was somewhat lower than the data published by Wollensak et al., ¹¹ who measured a 1.8× increase in stability. ¹¹ This difference can be attributed to the experimental procedure. The measured stress in the material-testing machine is highly dependent on experimental conditions, such as the load on the fixation clamp, the prestress force of the machine, and the condition of the tissue samples. It should be noted that, in our method, we used tissue samples that had much narrower widths than those used in the work published by Wollensak et al., and this may directly affect the total energy absorbed by the tissue samples. Other possible explanations for the differences between the literature data and the increase in stability reported here may be due to the nonlinear stress–strain behavior of the cornea. 

Other studies investigating the change in stiffness of porcine corneae during cross-linking have been performed. Tanter et al. ¹⁴ used a supersonic shear-imaging technique to measure corneal stiffness and showed an increase by a factor of 4.6. He et al. ¹⁵ obtained a factor of 1.04 by using a quantitative ultrasound method. Kling et al. ¹⁶ found an average 1.6× increase in Young's modulus in porcine eyes by comparing the corneal geometry changes when the intraocular pressure was changed. 

To show the equivalence of the standard and rapid CXL procedure, the equivalence parameter is critical with respect to the decision of whether a paired group should be considered equal or different. Choosing a more rigorous value for the equivalence parameter, such as Δ = 0.44 N/mm² instead of Δ = 0.66 N/mm² for Young's modulus, might lead to the conclusion that the rapid and the standard group differ from each other. However, to derive these possible differences statistically, a substantially higher number of eyes is needed. In addition, it is questionable whether the experimental setup used in this study would be capable of detecting these further differences because the handling of the tissue sample also causes large standard deviations. 

The equivalence of the rapid and standard CXL procedures, as shown in this study in porcine cornea, should be extrapolatable to human cornea. It could be shown by Wollensak et al. ¹⁷ that human and porcine corneas show similar behavior in terms of change in biomechanical stability due to cross-linking treatment. To show a statistically significant equivalence in human cornea, more than 50 human corneas would be needed, which seems unjustified in ethical terms. 

According to the Bunsen-Roscoe law, ¹⁰ the effect of a photochemical or photobiological reaction is directly proportional to the total irradiation dose, irrespective of the time span over which the dose is administered. This law may be valid within a certain dose range for pure photochemical reactions; however, the response of cells and tissue to electromagnetic radiation is more complex. Thus, a linear dose–time relationship is less likely. Currently, only very few studies analyze different tissues within a range in which this law can be applied. Moreover, the few published studies show diverging results. On the one hand, they support the direct application of the Bunsen-Roscoe law but, on the other hand, they also show that biological tissues possess a protective mechanism that can be damaged by electromagnetic radiation when the threshold intensity is exceeded for a prolonged period. ¹⁸  

On the supposition that the Bunsen-Roscoe law is valid for cross-link formation during the corneal cross-linking procedure, radicals generating these cross-links are formed in a shorter period when the radiation intensity is increased from 3 to 10 mW/cm². We assume that the possible protective mechanism of the tissue depends not only on irradiation time, but also on the rate of radical formation, and thus that the period during which the tissue is protected is dependent on the radiation intensity. At lower irradiation intensities, the tissue is protected for a longer time than at higher intensities. The results of this study show that, at the intensity and time regimen used, the possible protective mechanism had an equal influence on both treatment methods as both study groups showed similar increases in stiffness. It may even be possible to further increase the intensity to reach treatment times below 9 minutes. 

The Bunsen-Roscoe law can also be applied to estimate the endothelial safety during the illumination of the cornea with higher intensity. Endothelial cell damage or death is caused by the total energy deposited in the tissue, due to a cascade of photoinduced chemical effects. In our study, we increased the intensity from 3 to 10 mW/cm² and reduced the illumination time from 30 to 9 minutes, thus ensuring an equal energy dose of 5.4 J/cm² for both cases. Therefore, the expected photoinduced chemical effects on the endothelial cells are assumed to be the same in the rapid and the standard procedures. Since the endothelial cells are not damaged during the standard procedure, no damage is expected to arise as a result of the rapid procedure. This is currently under investigation by our group in human corneas. 

To make a clear statement about the transferability of the results obtained in this study from ex vivo tissue and about the validity of the Bunsen-Roscoe reciprocity law on living human corneas requires further clinical study. The safety of the deeper layers of the cornea, especially the endothelium, has to be investigated in vivo at higher irradiation intensities. If endothelial safety is ensured, this new rapid cross-linking method could replace the standard cross-linking protocol as developed by Spoerl et al. ⁸  

Footnotes

 Disclosure: S. Schumacher, IROC AG (E); L. Oeftiger, None; M. Mrochen, IROC AG (E), P