Abstract
Introduction
This phase 3, multicenter, sham procedure/placebo-controlled, double-masked, randomized clinical trial evaluated the safety and efficacy of oxygen-enriched epithelium-on corneal collagen cross-linking (CXL) for the treatment of keratoconus.
Methods
The trial enrolled patients aged 13 to 51 years diagnosed with keratoconus. Eyes were randomized in a 2:1 ratio to CXL treatment or sham/placebo. The CXL treatment group (200 eyes) received cross-linking using riboflavin 5′-phosphate ophthalmic solutions 0.239% and 0.177% with ultraviolet A (UV-A) irradiation and supplemental oxygen while the sham/placebo group (112 eyes) received placebo solutions and a sham irradiation procedure. The primary efficacy endpoint was the between-group difference in least squares (LS) mean change from baseline in maximum corneal curvature (Kmax) at month 12. Safety outcomes included adverse events and ophthalmic assessments.
Results
At month 12, the LS mean Kmax in the CXL treatment group improved as a change from baseline by 0.5 D (95% confidence interval [CI] 0.7, 0.3; p < 0.0001) and the LS mean Kmax in the untreated sham/placebo group deteriorated by 0.4 D (95% CI 0.1, 0.8: p = 0.0045). The difference between groups was − 1.0 D (95% CI − 1.3, − 0.6; p < 0.0001). Therefore, the study met the primary efficacy endpoint criterion with both a statistically significant and clinically meaningful change in the mean Kmax. There were no serious ocular adverse events nor severe treatment-related adverse events in the study eye. The most common adverse event was punctate keratitis (6.5% vs. 1.8% in the CXL and sham/placebo groups, respectively).
Conclusion
Epithelium-on CXL using riboflavin ophthalmic solutions (riboflavin 5′-phosphate ophthalmic solutions 0.239% and 0.177%) with UV-A irradiation and supplemental oxygen is safe and effective for the treatment of keratoconus in pediatric patients and adults.
Trial Registration
ClinicalTrials.gov identifier NCT 05759559.
Keywords: Epithelium-on, Oxygen-enriched, Corneal collagen cross-linking, Keratoconus
Key Summary Points
| Why carry out this study? | |
| Corneal collagen cross-linking is the only known treatment that has been shown to slow or halt progression of keratoconus. However, until recently, the only United States (US) Food and Drug Administration (FDA)-approved corneal cross-linking procedure involved debridement of the corneal epithelium (epithelium-off). | |
| We conducted a 12-month sham procedure/placebo-controlled study to evaluate the safety and efficacy of an oxygen-enriched corneal cross-linking procedure that does not involve debridement of the corneal epithelium (epithelium-on). | |
| What was learned from the study? | |
| The study met the pre-specified efficacy criterion of a statistically significant and clinically meaningful between-group difference at month 12, and demonstrated that the oxygen-enriched epithelium-on corneal cross-linking procedure was safe and well tolerated. The cross-linking treatment group demonstrated a significant improvement in maximum corneal curvature over the 12-month period while the sham procedure/placebo group exhibited a worsening of keratoconus. | |
| Oxygen-enriched epithelium-on corneal cross-linking is a promising new approach for the treatment of keratoconus in adult and pediatric patients. |
Introduction
Keratoconus is a type of corneal ectasia which causes thinning and weakening of the cornea, resulting in a subsequent cone-like topography. This type of irregular astigmatism leads to loss of uncorrected visual acuity (UCVA) and best spectacle-corrected visual acuity (BSCVA), producing symptoms such as blurred and distorted vision, poor night vision, and photophobia which significantly decrease quality of life in patients and cause detrimental psychological, social, and professional consequences [1]. Although visual acuity can be improved with treatments such as soft, hybrid, scleral, or rigid corneal contact lenses in addition to refractive surgical procedures such as photorefractive keratectomy and intracorneal ring segment implantation for mild to moderate stages of keratoconus, these interventions are merely short-term solutions which do not hinder disease progression or eradicate the condition, especially in more severe cases [2, 3]. More recent investigational techniques include corneal allogenic intrastromal ring segments (CAIRS) [4] and corneal tissue addition keratoplasty (CTAK) [5].
Corneal collagen cross-linking (CXL) is the only known treatment that has been shown to slow or halt progression of keratoconus [6, 7]. CXL involves soaking the cornea with riboflavin, followed by exposing the cornea to ultraviolet A (UV-A) irradiation to allow the cross-linking reaction to occur, thereby strengthening the corneal stroma. Until recently, the only United States (US) Food and Drug Administration (FDA)-approved CXL procedure involved debridement of the corneal epithelium (epithelium-off) to allow for better absorption of the riboflavin reagent and oxygen into the corneal stroma. However, epithelial debridement is accompanied by potential side effects such as infectious and non-infectious keratitis, delayed re-epithelization, increased pain and discomfort, corneal haze, corneal edema, and even thinning of the cornea [8]. Therefore, procedures in which the corneal epithelium is left intact (epithelium-on) have been developed to improve patient comfort [9].
Various methods have been studied to optimize epithelium-on CXL efficacy, including modification of the riboflavin formulations with a permeability enhancer (benzalkonium chloride) and viscosity enhancing agent (hydroxypropyl methylcellulose) to ensure ample penetration of riboflavin through the intact corneal epithelium into the stroma, and the use of supplemental oxygen during UV-A irradiation which is essential to create optimal conditions for efficient and effective corneal collagen cross-linking.
Because the intact corneal epithelium prevents sufficient diffusion of oxygen into the stroma under normoxic conditions (21% oxygen concentration), supplemental oxygen is essential to create hyperoxic conditions (90% or greater oxygen concentration) under which rapid diffusion of oxygen through the intact corneal epithelium occurs [10, 11]. Mechanistically, cross-linking can occur through either a type I slow reaction or a type II fast reaction; oxygen is absolutely necessary for the type II fast cross-linking reaction [12, 13]. Specifically, the presence of oxygen allows the excited photoenhancer (riboflavin) to react with non-reactive triplet oxygen and create the reactive singlet oxygen, which through other highly reactive and short-lived intermediates induces swift and more efficient cross-link formation between collagen fibers [12, 13].
Epithelium-on cross-linking procedures with high-fluence UV-A irradiation but shorter exposure time have also been developed to increase efficiency, decrease treatment time, and enhance patient comfort. However, continuous high-fluence UV-A exposure rapidly depletes oxygen, and thus both supplemental oxygen combined with pulsed irradiation (1 s on/1 s off) is necessary to maintain effective and efficient type II cross-linking reactions with the corneal epithelial intact [10, 14, 15].
Therefore, the purpose of this phase 3 randomized clinical trial in pediatric and adult patients was to evaluate the safety and efficacy of an epithelium-on CXL procedure using modified riboflavin formulations and oxygen-enriched pulsed high-fluence UV-A irradiation for the treatment of keratoconus.
Methods
Design
This was a phase 3, multicenter, randomized, sham procedure/placebo-controlled study which evaluated the safety and efficacy of epithelium-on corneal collagen cross-linking in impeding the progression of, and/or reducing maximum corneal curvature (Kmax), in eyes with keratoconus. This study was conducted in 28 clinical sites in the USA from January 2023 to June 2024.
The study was conducted in accordance with US regulations and the International Council for Harmonisation Guidelines on Good Clinical Practice which embody the tenets of the Declaration of Helsinki of 1964 and its later amendments. Institutional review board (IRB) approval (Alpha IRB, San Clemente, CA; GLK-202-02 approval date 12 December 2022) was obtained for the study. All patients provided written informed consent prior to their participation. Furthermore, those who were under the age of 18 years (or had not yet reached the age of majority per local regulations) signed the most recent IRB-approved assent form, and their parent or legal guardian signed the most recent IRB-approved informed consent form.
The clinical trial was prospectively registered and is available at ClinicalTrials.gov (identifier NCT05759559).
Participants
Patients, 12 to 55 years of age, were required to meet all inclusion criteria and none of the exclusion criteria prior to their treatment randomization.
Ocular inclusion criteria included topographic and clinical evidence of keratoconus defined as the following: axial topography consistent with keratoconus, maximum corneal curvature as measured by Kmax of ≥ 47.00 D, presence of central or inferior steepening on the Pentacam topography map, and presence of one or more findings associated with keratoconus, such as Fleischer ring, Vogt striae, corneal thinning, corneal scarring, and/or scissoring of the retinoscopic reflex. In addition, patients had to have a BSCVA of ≥ 1 letter and ≤ 85 letters on the Early Treatment of Diabetic Retinopathy Study (ETDRS) chart. For patients who wore contact lenses, removal of lenses was required for a 1-week period prior to the screening visit and patients must have agreed to remain out of their contact lenses from the time of treatment until the 1-month visit was completed. A contact lens wearer was defined as someone who had worn contact lenses in the eye to be treated in the prior 30 days.
Study exclusion criteria comprised contraindications, sensitivity or known allergy to the use of the riboflavin or placebo solutions, or their components. In addition, if a participant was female, she must not have been pregnant, nursing, planning a pregnancy, or have a positive urine pregnancy test prior to treatment of either eye during the study. In addition, women of childbearing potential must not have been lactating, and must have agreed to use a medically acceptable form of birth control for at least 1 week prior to treatment and for 1 month following the last treatment.
Ocular exclusion criteria included eyes classified as either normal, atypical normal, or keratoconus suspect on the severity grading scheme, as well as a history of previous corneal surgery, limbal relaxing incision procedure, or insertion of Intacs® in the eye(s) to be treated. Patients also were excluded if their corneal pachymetry value was < 325 μm at the thinnest point measured by Pentacam tomography in the eye to be treated; if the eye to be treated was aphakic; or the eye to be treated was pseudophakic and did not have a UV-blocking lens implanted. Furthermore, patients could not have a previous ocular condition (other than refractive error) in the eye to be treated that may have predisposed the eye for future complications. For instance, these consisted of a history of corneal disease (e.g., herpes simplex, herpes zoster keratitis, corneal melt, corneal dystrophy), clinically significant corneal scarring in the cross-linking treatment zone that was not related to keratoconus or, in the study investigator’s opinion, would have interfered with the cross-linking procedure, a history of delayed epithelial healing in the eye to be treated or a condition that may have interfered with or prolonged epithelial healing, nystagmus or any other condition that would have prevented a steady gaze during the treatment or other diagnostic tests, or history of previous corneal cross-linking treatment in the eye to be treated. Patients also should not have used an investigational drug or device within 30 days of screening or have been concurrently enrolled in another investigational drug or device trial.
Randomization and Masking
Patients were screened 45 to 1 day(s) before treatment and eyes were randomized using interactive response technology in a 2:1 ratio on the day of treatment (day 1) to receive either the CXL treatment or sham/placebo, respectively. Randomization was stratified by baseline keratoconus severity (mild vs. moderate/severe) in the study eye, based on a grading scheme adapted from McMahon et al. [16] which uses the flat keratometry readings obtained via Pentacam tomography, as well as by age (< 30 vs. ≥ 30 years) within each clinical site. In addition, randomization of eyes in patients ≥ 30 years of age was limited to no more than approximately 25% of the overall sample size, which equates to 73 eyes. To prevent the interjection of bias and to maintain data integrity, the study was double masked at the patient and sponsor level. Furthermore, beginning with the month 1 visit, the clinician who assessed BSCVA, UCVA, and manifest refraction was masked to treatment assignment. The investigator was not masked to the treatment administered to the patient.
The study employed a crossover design in which patients who initially randomized to sham/placebo were crossed over and received CXL treatment after 12 months, and were subsequently followed for an additional 6 months. However, the data presented herein are up until the 12-month visit before crossover occurred.
Procedures and Visits
CXL Treatment Group
In the CXL treatment group, investigators did not remove the corneal epithelium. In the first step of the procedure, the investigator applied topical anesthetic, inserted a lid speculum, and prepped the corneal surface using a wet Weck-Cel sponge soaked with riboflavin 5′-phosphate ophthalmic solution, 0.239% to remove the mucin layer with 4 to 10 gentle swipes in the horizontal and vertical meridian. The investigator then instilled two drops of riboflavin 5′-phosphate ophthalmic solution, 0.239% onto the cornea of the study eye every 60 s for 4 min, followed by two drops of riboflavin 5′-phosphate ophthalmic solution, 0.177% every 30 s for 6 min. Investigators gently rinsed the corneal epithelium with approximately 5 mL of balanced salt solution and performed ultrasound pachymetry to ensure the cornea remained at ≥ 325 µm. In order to induce cross-linking through supplemental oxygen, they immediately applied the oxygen goggles and turned on the humidified oxygen flow. The oxygen saturation level remained at approximately ≥ 90% on the corneal surface. The study eye was re-anesthetized and irradiated in a uniform circular area with a diameter of 9 mm at 30 mW/cm2 for 11 min and 6 s with an on/off cycle of 1 s UV-A on/1 s UV-A off (total UV-A dose of 10 J/cm2) at a wavelength of 365 nm using the UV-A irradiation system. During corneal irradiation, balanced salt solution was applied every 2 min or more for hydration as needed. After the UV-A treatment was completed, the oxygen goggles and speculum were removed, a slit-lamp examination was conducted, and a bandage contact lens was applied. Post-treatment medications consisted of a topical non-steroidal anti-inflammatory drug (NSAID), an oral NSAID, a broad-spectrum topical antibiotic, and artificial tears. Follow-up visits to assess medical and ocular status occurred 1 day, 3 days, 1 week, 1 month, and 3, 6, and 12 months after CXL treatment.
Sham Procedure/Placebo Group
Investigators performed a similar procedure in the sham/placebo group with some minor differences. In parallel to the CXL procedure, they did not remove the corneal epithelium, applied topical anesthetic, and inserted a lid speculum. Investigators prepped the corneal surface in a similar fashion with 4 to 10 horizontal and vertical meridian swipes using a wet Weck-Cel sponge, although they soaked it with placebo solution instead. They placed two drops of placebo solution (riboflavin 5′-phosphate ophthalmic solution, 0%) into the study eye every 60 s for 4 min followed by two drops of placebo solution every 30 s for 6 min. Investigators gently flushed the corneal surface with balanced salt solution, performed ultrasound pachymetry, and applied the oxygen goggles with the oxygen flow turned on. They re-anesthetized and irradiated the eye at 6 mW/cm2 for 11 min and 6 s with an on/off cycle of 1 s UV-A on/1 s UV-A off (total UV-A dose of 2 J/cm2) at 365 nm with application of balanced salt solution every 2 min or more for hydration as needed. After UV-A treatment, the oxygen goggles and speculum were removed, and a slit-lamp examination was conducted before applying a bandage contact lens. The investigators prescribed the same post-treatment medications and assessed patients in the sham procedure/placebo group at the same time intervals as patients in the CXL group.
A UV-A irradiance of 6 mW/cm2 was chosen to effectively mask the subject by providing perceptible illumination. The low level fluence (total UV-A dose of 2 J/cm2) was substantially lower than that used in the CXL treatment group and, in the absence of riboflavin, did not result in corneal collagen cross-linking as evidenced by the continued worsening of Kmax in the sham procedure/placebo group.
Efficacy and Safety Outcome Measures
The most significant indicator of keratoconus progression that was evaluated in this study was Kmax, which was measured on the Pentacam corneal topography image. At each required visit at least three acceptable images were to be captured with the median Kmax value determined from the best three images. Kmax was assessed at baseline and compared to its value at month 12 to evaluate the primary efficacy endpoint, and at month 6 to evaluate the secondary efficacy endpoint. The baseline value was defined as the most recent non-missing value obtained immediately prior to administration of treatment. Change from baseline was calculated by subtracting the baseline value from the post-procedure assessment for each study eye. For the primary efficacy endpoint, the testing criterion was a statistically significant (p < 0.05) mean difference of at least 1.0 D between the CXL treatment group and sham procedure/placebo group. The between-group difference of 1.0 D or greater at month 12 served as one of the two criteria to achieve the primary efficacy endpoint based on the requirement from the US FDA and prior precedence in pivotal trials of corneal cross-linking treatments [6].
Summary statistics for Kmax also were stratified by age (< 30, ≥ 30 years), sex (male, female), race (white, non-white), and baseline keratoconus severity (mild, moderate/severe) as additional efficacy endpoints.
Safety outcomes were evaluated by the incidence of treatment-emergent ocular adverse events and non-ocular adverse events. Treatment-emergent adverse events were summarized by treatment group and categorized by system organ class and preferred term using the Medical Dictionary for Regulatory Activities (MedDRA version 25.1). Treatment-emergent adverse events are adverse events that occurred after study treatment administration or pre-existing conditions that worsened after study treatment administration. All adverse events were categorized by severity grade (mild, moderate, severe), as well as by relationship to treatment (related vs. not related).
Other additional safety outcomes assessed included corneal thickness, central subfield thickness of the retina, intraocular pressure (IOP), and endothelial cell counts. The change from baseline in corneal thickness was measured by Pentacam central pachymetry at the thinnest locale at baseline and at each follow-up exam beginning with month 1. Macular central subfield thickness was evaluated at baseline and follow-up visits using optical coherence tomography. IOP was measured by Goldmann applanation tonometry at baseline and each follow-up visit. Similarly, endothelial cell counts were measured by specular microscopy at baseline and follow-up visits.
Statistical Analysis
The sample size for this study was based on unequal group sizes with the number of eyes in the CXL group being two times greater than in the control group, providing a 2:1 randomization ratio. A total of 261 eyes (174 in the CXL treatment group and 87 in the sham/placebo group) yielded 85% probability (two-sided alpha = 0.05) of observing a ≤ − 1.0 D difference in the mean change from baseline in Kmax at month 12 between the CXL and control groups. This also assumed a common standard deviation (SD) of 2.2 D. The target sample size for this study was 290 eyes to account for approximately a 10% dropout rate over the 12-month follow-up; the actual sample size was 312 eyes.
Statistical analyses were primarily performed on the intent-to-treat (ITT) analysis set for efficacy analyses which included all randomized study eyes, as well as on the per-protocol (PP) analysis set for supporting analyses. This was a subset of the ITT and comprised study eyes that completed the study without any major protocol deviations. Similarly, the safety analyses were conducted on the safety analysis set which consisted of all treated study eyes. For safety analyses in which the patient was the unit of analysis, a patient was placed in the CXL treatment group if either study eye was in the CXL treatment group. However, if a patient had both eyes in the control group, or only one eye in the study which was in the control group, then that patient was in the control group.
As stated previously, the primary efficacy endpoint used to determine whether the study was a success was a statistically significant (p < 0.05) month 12 change from baseline in the least squares (LS) mean Kmax between the CXL treatment and sham/placebo groups, and the criterion for clinical significance was a LS mean difference of 1.0 D. The primary efficacy analysis used a mixed model repeated measures (MMRM) analysis of covariance (ANCOVA) model with change from baseline Kmax value at month 12 as the response, treatment as a factor, and the baseline Kmax value, baseline keratoconus severity, and age stratum as covariates. This model controlled for the within-patient correlation resulting from a patient potentially having both eyes in the study using an unstructured covariance matrix.
Similarly, the secondary efficacy endpoint was the month 6 change from baseline in Kmax. The secondary efficacy analysis used the same model as the primary efficacy analysis. As a hierarchical testing strategy to control the type I error rate, the secondary efficacy endpoint analysis was only tested if the primary efficacy endpoint analysis was statistically significant.
In addition, multiple sensitivity analyses were used to support the primary analysis. These analyses were performed on the ITT and PP analysis sets using a combination of methods such as multiple imputation (MI), last observation carried forward (LOCF), and observed case to account for dropouts and missing data. Discontinuations due to lack of efficacy or adverse events were assumed to be missing not at random (MNAR), and missing data without discontinuation or discontinuation due to reasons other than lack of efficacy or adverse events were assumed to be missing at random (MAR).
If the MAR data did not follow a monotonic pattern, the Markov chain Monte Carlo (MCMC) imputation method was applied [17]. In order to produce a monotonic missing data pattern, monotonic missing data were imputed assuming a MAR pattern based on a regression model that incorporated treatment assignment, age (years), as well as the Kmax values at the baseline, and months 1, 3, 6, and 12. The MAR assumption implies that the covariates in the model account for differences in the distribution of missing variables for observed and missing cases. Therefore, the choice of treatment, age, and baseline Kmax were chosen as it was considered reasonable that outcomes for participants who dropped out were expected to be similar to outcomes for participants who did not drop out with these similar baseline characteristics.
Monotonic missing data were imputed assuming they were MNAR based on a regression model which incorporated the same variables as the MAR monotonic multiple imputation except for treatment assignment, while using the distribution from the worse half of control eyes which were defined as eyes whose change from baseline in Kmax was greater than the median for the control group at month 12. The initial imputation to produce a monotonic missing data pattern for the MNAR imputation was performed by using the same method as described for the MAR imputation.
All statistical analyses were performed using SAS Version 9.4.
Results
Participant Disposition and Demographics
A total of 312 eyes from 208 patients were randomized into the study, of which 200 eyes were randomized to CXL treatment and 112 eyes to sham/placebo. All randomized eyes were treated and comprised the ITT and safety analysis sets. A total of 12 eyes, 9 in the CXL treatment group and 3 in the sham/placebo group, were removed from the PP analysis set due to a major protocol deviation. In 11 patients, this was due to an inordinate delay between the onset of supplemental oxygen and initiation of UV-A irradiation, while one patient in the CXL group was removed because they did not meet BSCVA inclusion criteria. Patient demographics and baseline characteristics followed a similar distribution in both the CXL and sham/placebo groups in terms of age, sex, race, ethnicity, mean Kmax, keratoconus severity grade, and corneal thickness (Table 1). The mean ± SD age of the patient population was 26.7 ± 8.14 years. In addition, the majority of patients were male (67.6%), and identified as white (71.8%), and not of Hispanic or Latino ethnicity (63.8%).
Table 1.
Demographics and study eye baseline characteristics
| CXL treatment (N = 200) | Sham procedure/placebo (N = 112) | Total (N = 312) | |
|---|---|---|---|
| Age (years) | |||
| n | 200 | 112 | 312 |
| Mean (SD) | 26.9 (8.30) | 26.4 (7.89) | 26.7 (8.14) |
| Minimum, maximum | 13, 50 | 14, 51 | 13, 51 |
| Age, n (%) | |||
| < 18 years | 22 (11.0) | 14 (12.5) | 36 (11.5) |
| 18 to < 65 years | 178 (89.0) | 98 (87.5) | 276 (88.5) |
| < 30 years | 155 (77.5) | 87 (77.7) | 242 (77.6) |
| ≥ 30 years | 45 (22.5) | 25 (22.3) | 70 (22.4) |
| Sex, n (%) | |||
| Male | 136 (68.0) | 75 (67.0) | 211 (67.6) |
| Female | 64 (32.0) | 37 (33.0) | 101 (32.4) |
| Race, n (%) | |||
| White | 145 (72.5) | 79 (70.5) | 224 (71.8) |
| Black or African American | 38 (19.0) | 22 (19.6) | 60 (19.2) |
| Asian | 3 (1.5) | 1 (0.9) | 4 (1.3) |
| American Indian or Alaska Native | 3 (1.5) | 1 (0.9) | 4 (1.3) |
| Native Hawaiian or Other Pacific Islander | 1 (0.5) | 2 (1.8) | 3 (1.0) |
| Multiple | 1 (0.5) | 1 (0.9) | 2 (0.6) |
| Other | 2 (1.0) | 0 | 2 (0.6) |
| Unknown | 7 (3.5) | 6 (5.4) | 13 (4.2) |
| Ethnicity, n (%) | |||
| Hispanic or Latino | 71 (35.5) | 35 (31.3) | 106 (34.0) |
| Not Hispanic or Latino | 124 (62.0) | 75 (67.0) | 199 (63.8) |
| Unknown | 5 (2.5) | 2 (1.8) | 7 (2.2) |
| Baseline Kmax (D) | |||
| n | 200 | 112 | 312 |
| Mean (SD) | 58.0 (8.0) | 58.1 (8.6) | 58.0 (8.2) |
| Minimum, maximum | 47.0, 87.9 | 47.0, 90.7 | 47.0, 90.7 |
| Keratoconus severity grade, n (%) | |||
| Mild | 172 (86.0) | 95 (84.8) | 267 (85.6) |
| Moderate/severe | 28 (14.0) | 17 (15.2) | 45 (14.4) |
| Corneal thickness (µm) | |||
| n | 200 | 112 | 312 |
| Mean (SD) | 458.5 (46.10) | 467.5 (42.79) | 461.8 (45.08) |
| Minimum, maximum | 339, 568 | 341, 545 | 339, 568 |
Counts and percentages are based on the number of study eyes in the intent-to-treat analysis set. Patients with 2 eyes in the study are represented twice
CXL corneal collagen cross-linking, D diopters, Kmax maximum corneal curvature, SD standard deviation
Efficacy
Kmax Analyses
Two efficacy analyses were conducted: change from baseline within the two groups (i.e., CXL and sham/placebo groups) and change from baseline between the two groups. While the difference between groups defines the overall treatment effect, the change from baseline analysis within a group has important clinical implications. In the change from baseline analysis within a group, an increase from baseline in Kmax over time signifies worsening of the disease, no change from baseline in Kmax signifies stabilization of the disease, and a decrease from baseline in Kmax signifies improvement of the disease.
Kmax at Month 12
At baseline, Kmax values were comparable in the CXL and sham/placebo groups with a mean ± SD of 58.0 ± 8.0 D and 58.1 ± 8.6 D, respectively. At month 12, LS mean Kmax in the CXL-treated group improved as a change from baseline by 0.5 D (95% confidence interval [CI] 0.7, 0.3; p < 0.0001) which signified not only stopping the worsening of the keratoconus disease but also restoration of the healthy corneal geometry lost to the disease worsening prior to the CXL treatment. At the same time, LS mean Kmax in the untreated sham/placebo group deteriorated by 0.4 D (95% CI 0.1, 0.8: p = 0.0045), signifying worsening of the disease over the course of 12 months. As a combined outcome at month 12, the difference in LS mean Kmax change from baseline between the groups was a statistically superior − 1.0 D (95% CI − 1.3, − 0.6; p < 0.0001) (Fig. 1). Therefore, the study met the primary efficacy endpoint criterion with both a statistically significant and clinically meaningful difference between groups in LS mean change from baseline for Kmax (Table 2).
Fig. 1.
Graph showing the change from baseline in least squares mean values for maximum corneal curvature (Kmax) (± standard error of the mean) (diopters, D) for the corneal collagen cross-linking (CXL) treatment group and sham procedure/placebo group at the month 1, 3, 6, and 12 visits. The CXL treatment group exhibits a reduction/improvement in Kmax and the sham procedure/placebo group exhibits an increase/worsening in Kmax. Difference represents the least squares mean change from baseline in Kmax (± standard error of the difference) for the CXL treatment group minus the least squares mean change from baseline in Kmax in the sham procedure/placebo group
Table 2.
Change from baseline in Kmax at month 12; primary analysis on the primary endpoint (ITT analysis set)
| CXL treatment (N = 200) | Sham procedure/placebo (N = 112) | |
|---|---|---|
| LS mean change from baseline (SEM) | − 0.5 D (0.1) | 0.4 D (0.2) |
| 95% CI for LS mean change from baseline | (− 0.7, − 0.3) | (0.1, 0.8) |
| p value | < 0.0001 | 0.0045 |
| Differences of LS mean from sham/placebo (SED) | − 1.0 D (0.2) | |
| 95% CI for LS mean difference from sham/placebo | (− 1.3, − 0.6) | |
| p value | < 0.0001 |
CI confidence interval, CXL corneal collagen cross-linking, D diopter, ITT intent-to-treat, Kmax maximum corneal curvature, LS least squares, SED standard error of the difference, SEM standard error of the mean
The robustness of the results of the primary endpoint was supported by multiple sensitivity analyses on the ITT and PP analysis sets using various methods to impute missing post-baseline Kmax data. In all nine analyses, the trend was the same as that observed in the primary analysis; the CXL treatment group had a reduction/improvement from baseline in Kmax and the sham/placebo group had an increase/worsening from baseline in Kmax at month 12 with a difference between groups in the LS mean change from baseline of − 0.8 D to − 1.0 D (Table 3). In addition, five of the nine sensitivity analyses reached the efficacy criterion of a clinically meaningful difference in Kmax of ≤ − 1.0 D between the CXL treatment group and the sham/placebo group.
Table 3.
Change from baseline in Kmax at month 12; sensitivity analyses
| CXL treatment (N = 200) | Sham procedure/placebo (N = 112) | |
|---|---|---|
| Sensitivity analysis 1 (ITT analysis set) | ||
| LS mean change from baseline (SEM) | − 0.5 D (0.1) | 0.4 D (0.2) |
| 95% CI for LS mean change from baseline | (− 0.7, − 0.3) | (0.1, 0.8) |
| Differences of LS mean from sham/placebo (SED) | − 1.0 D (0.2) | |
| 95% CI for LS mean difference from sham/placebo | (− 1.3, − 0.6) | |
| p value | < 0.0001 | |
| Sensitivity analysis 2 (ITT analysis set) | ||
| LS mean change from baseline (SEM) | − 0.3 D (0.1) | 0.5 D (0.2) |
| 95% CI for LS mean change from baseline | (− 0.6, − 0.1) | (0.2, 0.8) |
| Differences of LS mean from sham/placebo (SED) | − 0.8 D (0.2) | |
| 95% CI for LS mean difference from sham/placebo | (− 1.2, − 0.5) | |
| p value | < 0.0001 | |
| Sensitivity analysis 3 (ITT analysis set) | ||
| LS mean change from baseline (SEM) | − 0.5 D (0.1) | 0.5 D (0.2) |
| 95% CI for LS mean change from baseline | (− 0.7, − 0.3) | (0.2, 0.8) |
| Differences of LS mean from sham/placebo (SED) | − 1.0 D (0.2) | |
| 95% CI for LS mean difference from sham/placebo | (− 1.3, − 0.6) | |
| p value | < 0.0001 | |
| Sensitivity analysis 4 (ITT analysis set) | ||
| LS mean change from baseline (SEM) | − 0.5 D (0.1) | 0.5 D (0.2) |
| 95% CI for LS mean change from baseline | (− 0.7, − 0.3) | (0.1, 0.8) |
| Differences of LS mean from sham/placebo (SED) | − 1.0 D (0.2) | |
| 95% CI for LS mean difference from sham/placebo | (− 1.3, − 0.6) | |
| p value | < 0.0001 | |
| Sensitivity analysis 5 (ITT analysis set) | ||
| LS mean change from baseline (SEM) | − 0.5 D (0.1) | 0.4 D (0.2) |
| 95% CI for LS mean change from baseline | (− 0.7, − 0.3) | (0.1, 0.8) |
| Differences of LS mean from sham/placebo (SED) | − 1.0 D (0.2) | |
| 95% CI for LS mean difference from sham/placebo | (− 1.3, − 0.6) | |
| p value | < 0.0001 | |
| Sensitivity analysis 6 (ITT analysis set) | ||
| LS mean change from baseline (SEM) | − 0.5 D (0.1) | 0.4 D (0.2) |
| 95% CI for LS mean change from baseline | (− 0.7, − 0.2) | (0.1, 0.7) |
| Differences of LS mean from sham/placebo (SED) | − 0.9 D (0.2) | |
| 95% CI for LS mean difference from sham/placebo | (− 1.3, − 0.5) | |
| p value | < 0.0001 | |
| Sensitivity analysis 7 (ITT analysis set) | ||
| n | 182 | 102 |
| LS mean change from baseline (SEM) | − 0.5 D (0.1) | 0.3 D (0.2) |
| 95% CI for LS mean change from baseline | (− 0.7, − 0.3) | (0.0, 0.6) |
| Differences of LS mean from sham/placebo (SED) | − 0.8 D (0.2) | |
| 95% CI for LS mean difference from sham/placebo | (− 1.2, − 0.5) | |
| p value | < 0.0001 | |
| Sensitivity analysis 8 (PP analysis set) | ||
| n | 191 | 109 |
| LS mean change from baseline (SEM) | − 0.5 D (0.1) | 0.5 D (0.2) |
| 95% CI for LS mean change from baseline | (− 0.8, − 0.3) | (0.2, 0.8) |
| Differences of LS mean from sham/placebo (SED) | − 1.0 D (0.2) | |
| 95% CI for LS mean difference from sham/placebo | (− 1.4, − 0.6) | |
| p value | < 0.0001 | |
| Sensitivity analysis 9 (ITT analysis set) | ||
| LS mean Change from baseline (SEM) | − 0.5 D (0.1) | 0.4 D (0.2) |
| 95% CI for LS mean change from baseline | (− 0.7, − 0.3) | (0.1, 0.7) |
| Differences of LS mean from sham/placebo (SED) | − 0.9 D (0.2) | |
| 95% CI for LS mean difference from sham/placebo | (− 1.3, − 0.6) | |
| p value | < 0.0001 | |
CI confidence interval, CXL corneal collagen cross-linking, D diopter, ITT intent-to-treat, Kmax maximum corneal curvature, LS least squares mean, PP per protocol, SED standard error of the difference, SEM standard error of the mean
Kmax at Month 6
At month 6, the CXL-treated group improved as a change from baseline by 0.4 D (95% CI 0.6, 0.2; p < 0.0001) which signified not only stopping the worsening of the keratoconus disease but also restoration of the healthy corneal geometry lost to the disease worsening prior to the CXL treatment. At the same time, Kmax in the untreated sham/placebo group increased by 0.1 D (95% CI − 0.1, 0.4; p = 0.3858), potentially signifying a gradual worsening of the disease. As a combined outcome at month 6, the difference in LS mean Kmax change from baseline between the groups was a statistically significant − 0.6 D (95% CI − 0.9, − 0.2; p = 0.0008) (Table 4).
Table 4.
Change from baseline in Kmax at month 6 (ITT analysis set)
| CXL treatment (N = 200) | Sham procedure/placebo (N = 112) | |
|---|---|---|
| LS mean change from baseline (SEM) | − 0.4 D (0.1) | 0.1 D (0.1) |
| 95% CI for LS mean change from baseline | (− 0.6, − 0.2) | (− 0.1, 0.4) |
| p value | < 0.0001 | 0.3858 |
| Differences of LS mean from sham/placebo (SED) | − 0.6 D (0.2) | |
| 95% CI for LS mean difference from sham/placebo | (− 0.9, − 0.2) | |
| p value | 0.0008 |
CI confidence interval, CXL corneal collagen cross-linking, D diopter, ITT intent-to-treat, Kmax maximum corneal curvature, LS least squares mean, SED standard error of the difference, SEM standard error of the mean
The same results were obtained through the two separate sensitivity analyses using the previously described methods which were performed on the PP and ITT analysis sets, respectively (Table 5). Both supporting analyses also showed a 0.4-D decrease/improvement in LS mean Kmax from baseline to month 6 in the CXL treatment group and a 0.1-D increase/worsening in LS mean Kmax in the sham/placebo group. As a combined outcome, the difference in LS mean Kmax change from baseline between the two groups for both sensitivity analyses was a statistically significant − 0.6 D (95% CI − 0.9, − 0.2; p = 0.0011).
Table 5.
Change from baseline in Kmax at month 6: sensitivity analyses
| CXL treatment (N = 200) | Sham treatment/control (N = 112) | |
|---|---|---|
| PP analysis set | ||
| n | 191 | 109 |
| LS mean (SEM) | − 0.4 (0.1) | 0.1 (0.1) |
| 95% CI for LS mean | (− 0.6, − 0.2) | (− 0.1, 0.4) |
| Differences of LS mean from sham/control (SED) | − 0.6 (0.2) | |
| 95% CI for LS mean difference from sham/control | (− 0.9, − 0.2) | |
| p value | 0.0011 | |
| ITT analysis set | ||
| n | 200 | 112 |
| LS mean (SEM) | − 0.4 (0.1) | 0.1 (0.1) |
| 95% CI for LS mean | (− 0.6, − 0.2) | (− 0.1, 0.4) |
| Differences of LS mean from sham/control (SED) | − 0.6 (0.2) | |
| 95% CI for LS mean difference from sham/control | (− 0.9, − 0.2) | |
| p value | 0.0011 | |
CI confidence interval, CXL corneal collagen cross-linking, D diopter, ITT intent-to-treat, Kmax maximum corneal curvature, LS mean least squares mean, PP per protocol, SED standard error of the difference, SEM standard error of the mean
Kmax in Subgroups at Month 12
The primary analysis was used for the age strata subgroup analyses and observed case (i.e., ignoring missing data) was used for all other subgroup analyses (comparing treatment groups by sex, race, and baseline keratoconus severity).
For the age strata subgroup analyses (Table 6), baseline Kmax was comparable between the CXL and sham/placebo groups of patients < 30 years of age (58.4 ± 7.7 D and 58.7 ± 8.7 D, respectively), as well as between the two groups of patients ≥ 30 years of age (56.5 ± 8.7 D and 55.9 ± 8.0 D, respectively).
Table 6.
Change from baseline in Kmax (D) at month 12: by subgroup (ITT analysis set)
| CXL treatment (N = 200) | Sham procedure/placebo (N = 112) | |
|---|---|---|
| Age strata: < 30 years | ||
| n | 155 | 87 |
| LS mean change from baseline (SEM) | − 0.5 D (0.1) | 0.5 D (0.2) |
| 95% CI for LS mean change from baseline | (− 0.8, − 0.2) | (0.2, 0.9) |
| p value | 0.0002 | 0.0029 |
| Differences of LS mean from sham/placebo (SED) | − 1.1 D (0.2) | |
| 95% CI for LS mean difference from sham/placebo | (− 1.5, − 0.6) | |
| p value | < 0.0001 | |
| Age strata: ≥ 30 years | ||
| n | 45 | 25 |
| LS mean change from baseline (SEM) | − 0.6 D (0.2) | − 0.0 D (0.3) |
| 95% CI for LS mean change from baseline | (− 1.0, − 0.1) | (− 0.6, 0.5) |
| p value | 0.0114 | 0.8650 |
| Differences of LS mean from sham/placebo (SED) | − 0.5 D (0.3) | |
| 95% CI for LS mean difference from sham/placebo | (− 1.2, 0.2) | |
| p value | 0.1448 | |
| Sex: male | ||
| n | 122 | 68 |
| LS mean change from baseline (SEM) | − 0.4 D (0.1) | 0.3 D (0.2) |
| 95% CI for LS mean change from baseline | (− 0.7, − 0.1) | (− 0.0, 0.7) |
| p value | ||
| Differences of LS mean from sham/placebo (SED) | − 0.8 D (0.2) | |
| 95% CI for LS mean difference from sham/placebo | (− 1.2, − 0.3) | |
| p value | 0.0015 | |
| Sex: female | ||
| n | 60 | 34 |
| LS mean change from baseline (SEM) | − 0.6 D (0.2) | 0.3 D (0.2) |
| 95% CI for LS mean change from baseline | (− 1.0, − 0.3) | (− 0.2, 0.8) |
| p value | ||
| Differences of LS mean from sham/placebo (SED) | − 0.9 D (0.3) | |
| 95% CI for LS mean difference from sham/placebo | (− 1.5, − 0.3) | |
| p value | 0.0047 | |
| Race: white | ||
| n | 137 | 78 |
| LS mean change from baseline (SEM) | − 0.5 (0.1) | 0.4 (0.2) |
| 95% CI for LS mean change from baseline | (− 0.7, − 0.2) | (0.1, 0.7) |
| p value | ||
| Differences of LS mean from sham/placebo (SED) | − 0.8 (0.2) | |
| 95% CI for LS mean difference from sham/placebo | (− 1.2, − 0.5) | |
| p value | < 0.0001 | |
| Race: non-white | ||
| n | 45 | 24 |
| LS mean change from baseline (SEM) | − 0.6 (0.2) | 0.1 (0.3) |
| 95% CI for LS mean change from baseline | (− 1.1, − 0.1) | (− 0.5, 0.8) |
| p value | ||
| Differences of LS mean from sham/placebo (SED) | − 0.7 (0.4) | |
| 95% CI for LS mean difference from sham/placebo | (− 1.6, 0.1) | |
| p value | 0.0987 | |
| Baseline keratoconus severity: mild | ||
| n | 158 | 87 |
| LS mean change from baseline (SEM) | − 0.5 D (0.1) | 0.4 D (0.1) |
| 95% CI for LS mean change from baseline | (− 0.7, − 0.3) | (0.1, 0.7) |
| p value | ||
| Differences of LS mean from sham/control (SED) | − 0.9 D (0.2) | |
| 95% CI for LS mean difference from sham/control | (− 1.2, − 0.5) | |
| p value | < 0.0001 | |
| Baseline keratoconus severity: moderate/severe | ||
| n | 24 | 15 |
| LS mean change from baseline (SEM) | − 0.5 D (0.5) | 0.0 D (0.7) |
| 95% CI for LS mean change from baseline | (− 1.6, 0.6) | (− 1.4, 1.5) |
| p value | ||
| Differences of LS mean from sham/control (SED) | − 0.5 D (0.9) | |
| 95% CI for LS mean difference from sham/control | (− 2.4, 1.3) | |
| p value | 0.5461 | |
CI confidence interval, CXL corneal collagen cross-linking, D diopter, ITT intent-to-treat, Kmax maximum corneal curvature, LS least squares mean, SED standard error of the difference, SEM standard error of the mean
In the < 30-year age strata, Kmax in the CXL-treated group improved as a change from baseline by 0.5 D (95% CI 0.8, 0.2; p = 0.0002) and in the sham/placebo group deteriorated by 0.5 D (95% CI 0.2, 0.9; p = 0.0029) at month 12. As a combined outcome at month 12, the difference in LS mean Kmax change from baseline between the groups was a statistically significant − 1.1 D (95% CI − 1.5, − 0.6; p < 0.0001).
In the ≥ 30-year age strata, Kmax in the CXL-treated group improved from baseline by 0.6 D (95% CI 1.0, 0.1; p = 0.0114) and in the sham/placebo group remained unchanged from baseline ( − 0.0 D; 95% CI − 0.6, 0.5; p = 0.8650). As a combined outcome, the difference in LS mean Kmax change from baseline between groups was − 0.5 D (95% CI − 1.2, 0.2; p = 0.1448). While the change from baseline in the CXL group was statistically significant, the combined outcome (i.e., the difference between the CXL and sham/placebo groups) trended toward statistical significance but did not reach it likely due to the small sample sizes (45 in the CXL group and 25 in the sham/placebo group).
In the analysis by sex, the mean baseline Kmax was similar in the CXL treatment and sham/placebo groups of male (58.3 ± 8.3 D and 58.9 ± 8.6 D, respectively) and female patients (57.4 ± 7.2 D and 56.4 ± 8.4 D, respectively). At month 12, the difference between the two groups in the LS mean Kmax change from baseline was similar in male patients ( − 0.8 D; 95% CI − 1.2, − 0.3; p = 0.0015) and female patients ( − 0.9 D; 95% CI − 1.5, − 0.3; p = 0.0047) (Table 6).
In the analysis by race, the mean baseline Kmax was similar in the CXL and sham/placebo groups of white patients (56.7 ± 7.1 D and 57.0 ± 8.0 D, respectively), as well as between treatment groups of non-white patients (61.3 ± 9.1 D and 60.6 ± 9.6 D, respectively). At month 12, the difference between the CXL and placebo/sham groups in the LS mean Kmax change from baseline was − 0.8 D (95% CI − 1.2, − 0.5; p < 0.0001) in white patients and − 0.7 D (95% CI − 1.6, 0.1; p = 0.0987) in non-white patients (Table 6). The difference trended toward statistical significance in eyes of non-white patients but did not reach it, which may be a reflection of the small sample size (45 in the CXL group and 24 in the sham/placebo group).
In the analysis according to baseline keratoconus severity, the mean baseline Kmax was similar in the CXL and sham/placebo groups with mild keratoconus severity (55.8 ± 5.7 D and 55.5 ± 5.8 D, respectively), and was higher but similar in the CXL and placebo/sham groups with moderate/severe severity (71.4 ± 6.8 D and 72.5 ± 7.4 D, respectively). At month 12, in eyes with mild keratoconus severity, the difference in LS mean Kmax change from baseline between the CXL and sham/placebo groups was − 0.9 D (95% CI − 1.2, − 0.5; p < 0.0001), whereas in eyes of moderate/severe keratoconus severity, the difference in LS mean Kmax change from baseline between the two groups was − 0.5 D (95% CI − 2.4, 1.3; p = 0.5461) (Table 6). This lack of a statistically significant treatment effect in eyes with moderate/severe keratoconus severity is likely due to the small sample size (24 in the CXL group and 15 in the sham/placebo group).
Safety
Adverse Events
No serious ocular adverse events occurred in the study. Through month 12, a total of 106 ocular adverse events occurred in 61 of 200 (30.5%) eyes in the CXL group, most of which were reported at or prior to month 3 (94 of 106 events). In contrast, a total of 13 ocular adverse events occurred in 12 of 112 (10.7%) eyes in the sham/placebo group with the majority (8 of 13 events) after month 3. Most of these ocular adverse events were mild in severity. For instance, among the 61 eyes that experienced adverse events in the CXL group, 40 were mild. Among the 12 eyes that experienced adverse events in the sham/placebo group, 8 were mild. Moderate adverse events occurred in 20 CXL-treated eyes and 3 sham/placebo eyes. A severe ocular adverse event occurred in a single eye within each of the treatment groups.
Punctate keratitis was the most common ocular adverse event that occurred in 13 of 200 (6.5%) eyes in the CXL group and in 2 of 112 (1.8%) eyes in the sham/placebo group from baseline through month 12. Other ocular adverse events that occurred at an incidence of ≥ 3.0% in the CXL group included corneal opacity (described as “haze”), photophobia, reduced visual acuity, conjunctival hyperemia, and conjunctivitis allergic (Table 7). No ocular adverse event occurred at an incidence of ≥ 3.0% in the sham/placebo group.
Table 7.
Most common ocular adverse events (≥ 3.0%) (safety analysis set)
| MedDRA Preferred Term | CXL treatment (N = 200) n (%) [events] | Sham procedure/placebo (N = 112) n (%) [events] |
|---|---|---|
| Punctate keratitis | 13 (6.5) [14] | 2 (1.8) [2] |
| Corneal opacity | 8 (4.0) [10] | 1 (0.9) [1] |
| Visual acuity reduced | 7 (3.5) [8] | 1 (0.9) [1] |
| Conjunctival hyperaemia | 7 (3.5) [7] | 1 (0.9) [1] |
| Photophobia | 8 (4.0) [8] | 0 |
| Conjunctivitis allergic | 6 (3.0) [6] | 0 |
CXL corneal collagen cross-linking, MedDRA Medical Dictionary for Regulatory Activities
Non-ocular adverse events occurred in similar proportions of patients across treatment groups, with adverse events in 11 of 164 (6.7%) patients in the CXL group and 2 of 44 (4.5%) patients in the sham/placebo group. There were no severe non-ocular adverse events or treatment-related non-ocular adverse events that occurred from baseline through month 12.
Visual Acuity
At baseline, the mean ± SD BSCVA was 69.0 ± 13.21 and 69.3 ± 13.99 ETDRS letters in the CXL and sham/placebo groups, respectively. At month 12, the change from baseline in BSCVA was similar in the two groups (3.5 ± 0.63 and 2.7 ± 0.79 ETDRS letters in the CXL and sham/placebo groups, respectively).
Similarly, at baseline, mean ± SD UCVA was 47.7 ± 18.97 and 48.0 ± 20.88 ETDRS letters in the CXL and sham/placebo groups, respectively. At month 12, the change from baseline in UCVA was similar in the two groups (3.7 ± 0.83 and 4.1 ± 1.07 ETDRS letters in the CXL and sham/placebo groups, respectively).
Corneal Thickness
At baseline, the mean ± SD corneal thickness was 458.5 ± 46.10 μm and 467.5 ± 42.79 μm in the CXL and sham/placebo groups, respectively. At month 12, the change from baseline was not clinically meaningful in either treatment group ( − 1.1 ± 12.33 μm and − 1.9 ± 9.54 μm in the CXL and sham/placebo groups, respectively).
Endothelial Cell Count
At baseline, the mean ± SD endothelial cell count was 2812.0 ± 257.56 cells/ mm2 in the CXL group and 2790.2 ± 282.83 cells/ mm2 in the sham/placebo group. At month 12, the mean ± SD change from baseline in endothelial cell count was − 1.6 ± 157.45 cells/mm2 in the CXL treatment group and − 36.7 ± 188.00 cells/ mm2 in the sham/placebo group.
Intraocular Pressure (IOP)
At baseline, the mean ± SD IOP was 13.1 ± 2.75 mmHg in the CXL treatment group and 13.2 ± 2.34 mmHg in the sham/placebo group. At month 12, the mean ± SD change from baseline was 0.3 ± 3.01 mmHg in the CXL treatment group and − 0.2 ± 3.00 mmHg in the sham/placebo group.
Central Subfield Thickness
At baseline, mean ± SD retinal central subfield thickness was 257.0 ± 31.94 μm in the CXL treatment group and 252.3 ± 40.81 μm in the sham/placebo group. At month 12, the mean ± SD change from baseline was − 1.6 ± 17.92 μm in the CXL treatment group and − 0.5 ± 37.09 μm in the sham/placebo group.
Discussion
This was the second of two pivotal studies conducted to assess the efficacy and safety of oxygen-enriched, epithelium-on corneal cross-linking as a treatment for keratoconus in adults and pediatric patients. The study met its primary efficacy endpoint of both a statistically significant and clinically meaningful difference between the CXL and sham/placebo groups in the change from baseline to month 12 for mean Kmax, thereby demonstrating that oxygen-supplemented epithelium-on CXL treatment is an effective treatment option with a safer risk profile than epithelium-off CXL. These efficacy results were robustly supported by multiple sensitivity analysis models which showed the same trend of a decrease in Kmax after CXL treatment and an increase in Kmax after sham procedure/placebo, with results from the majority of these models being clinically meaningful and statistically significant. Therefore, the results strongly indicate that epithelium-on CXL treatment in the presence of supplemental oxygen reduces corneal curvature and subsequently has the potential to prevent long-term vision loss in patients with keratoconus.
In addition to a statistically significant treatment effect in Kmax at month 12, the treatment effect based on the primary analysis was statistically significant at month 6, and was supported by sensitivity analyses.
While the primary efficacy analysis was based on the difference between the treated and untreated groups, the change from baseline analysis within a group also has important clinical implications. In the change from baseline analysis within a group, an increase from baseline in Kmax over time signifies worsening of keratoconus, no change from baseline in Kmax signifies stabilization of the disease, and a decrease from baseline in Kmax signifies improvement in the disease. A statistically significant decrease/improvement from baseline in Kmax was observed in the entire CXL-treated group at month 6 and month 12, as well as in the CXL-treated subgroups in the < 30-year age strata, the ≥ 30-year age strata, in male patients, in female patients, in white patients, and in eyes with mild baseline keratoconus severity, thereby providing compelling evidence that epithelium-on CXL with oxygen supplementation is an effective treatment for keratoconus. Other subgroups also showed a treatment effect but did not reach statistical significance due to small sample size.
Therefore, the study offers valuable insight into keratoconus disease management strategies which can be used to prevent the risk of irreversible vision loss and improve the quality of life for patients before it is too late and their disease progresses too far.
Importantly, an equal improvement in Kmax change from baseline was observed in the CXL-treated eyes with mild keratoconus severity and those with moderate/severe keratoconus severity, despite the small sample size of eyes with moderate/severe keratoconus. However, a difference between mild vs. moderate/severe keratoconus subgroups was observed in the increase/worsening of Kmax change from baseline wherein the sham/placebo mild group worsened more than the sham/placebo moderate/severe subgroup over the 12-month duration of the trial.
Therefore, this may indicate that early detection and treatment of keratoconus is vital to reduce the economic burden of the disease by diminishing the need for corneal transplantation. According to Rapuano et al. [18], patients who received CXL treatment were less likely to receive penetrating keratoplasty surgery, spent fewer years in advanced keratoconus disease stages, had fewer medical costs, and higher quality-adjusted life years compared to patients who did not receive CXL. This further underscores the need for early keratoconus detection and treatment.
In addition, the results demonstrate that CXL has a similar treatment effect in both sexes. Similarly, this trend was observed in the analysis comparing treatment differences in patients who identified as white and non-white. An equal improvement in Kmax change from baseline was observed in the CXL-treated eyes of white and non-white patients, despite the small sample size of non-white patients in the study. However, a difference between white and non-white subgroups was observed in the increase/worsening of Kmax change from baseline wherein the sham/placebo subgroup of white patients worsened more than the sham/placebo subgroup of non-white patients over the 12-month duration of the trial. As a result of the small sample size, additional research is needed to determine the effect of CXL treatment on patients of different races in order to identify which populations are at greatest risk for vision loss and in need of swift intervention.
Furthermore, the results indicate that epithelium-on CXL with oxygen supplementation is not only an effective treatment but is also safe and well tolerated. Few ocular adverse events occurred in patients who underwent epithelium-on CXL treatment, with the majority of these adverse events occurring in the early period following treatment. As aforementioned, the most common ocular adverse event was punctate keratitis which resolved quickly. Some ocular adverse events are to be expected due to the nature of the treatment; however, the low incidence rates and mild severity of most events suggest that epithelium-on CXL in the presence of supplemental oxygen is a safe and well-tolerated procedure. This was further illustrated by the complete absence of serious ocular adverse events in both groups. In addition, there were no treatment-related non-ocular adverse events.
Furthermore, the lack of meaningful changes in corneal thickness, corneal endothelial cell counts, IOP, and central subfield thickness of the retina adds additional validation of the safety and tolerability of epithelium-on CXL in the presence of supplemental oxygen. These results are also supported in the literature through a multitude of other studies which evaluate the safety of epithelium-on CXL with supplemental oxygen to treat keratoconus [19–22]. There have been trials that compared epithelium-off CXL to epithelium-on CXL and have reported an increased recovery time as a result of epithelium debridement in the epithelium-off group. A study conducted by Spadea et al. [23] reported a statistically significant reduction of corneal sensitivity for up to 3 months for individuals who underwent epithelium-off CXL, which was not observed in the epithelium-on CXL group for as long of a duration (7 days) nor was it statistically significant. In addition, several studies have reported an increased risk of infection for epithelium-off CXL which commonly occurs during the re-epithelialization phase. The infection rate after epithelium-off CXL has been reported to range from 0.12% [24] to 0.6% [25]. Further analysis of the specific types of infection has been described by Murchison et al. [26] wherein 31 of the 49 reported eyes that developed an infection following epithelium-off CXL had bacterial keratitis, while the other 18 eyes had viral, fungal, or parasitic keratitis. Twenty-eight eyes were infected with Staphylococcus aureus, 14 of which exhibited antibiotic resistance. Therefore, there is less risk of infection with epithelium-on CXL with supplemental oxygen than with the epithelium-off procedure. Multiple studies including the current study indicate that epithelium-on CXL with supplemental oxygen is an effective method of treating keratoconus, while also associated with a lower incidence of side effects that resolve more quickly compared to the epithelium-off procedure [20]. The appealing safety profile of epithelium-on CXL with supplemental oxygen is highly meaningful in terms of a patient’s comfort and well-being, which also has the potential to increase quality of life in multiple areas. Meanwhile debridement of the cornea in the epithelium-off procedure, which causes a plethora of uncomfortable and longer-lasting side effects such as increased pain, corneal haze, photophobia, and eye irritation, may result in a delay in returning to work or resuming normal activities and, more critically, could be enough to deter a patient from seeking care altogether. Conversely, the lack of debridement of the cornea in the epithelium-on CXL procedure with supplemental oxygen leads to quicker recovery, lack of epithelial defect, and potentially a faster return to work or school. For a disease such as keratoconus which has extremely detrimental and sight-threatening consequences if not treated as early as possible, epithelium-on CXL with supplemental oxygen is a safe and effective procedure and represents a promising new approach for the treatment of keratoconus.
This study has many strengths including the gold-standard randomized, sham procedure/placebo-controlled research design, and the similar demographic profile of participants in the CXL and sham/placebo groups, which limits bias. In addition, the sample size of 312 eyes provides sufficient power underpinning valid statistical analyses. Furthermore, the sensitivity analyses, using different analysis populations and algorithms to account for missing data, add credibility to the results and limit bias by addressing different data assumptions.
Limitations of the study include the inability to mask the investigator providing the CXL treatment or sham procedure/placebo. However, this potential source of bias was mitigated by the objective nature of the Kmax measurement and by having manifest refraction and visual acuity assessments performed by an individual who was masked to treatment assignment. In addition, although the participants were followed for up to a year, it is possible for keratoconus progression to recur after this time. As a result, an extension clinical trial has already been initiated which enrolled patients who completed the current trial to monitor them for longer duration of time. Furthermore, although the study population contained a similar demographic distribution between the two groups and all races observed a similar trend in the reduction of Kmax, some races were represented with relatively small sample size.
Conclusion
The results of this study provide evidence that epithelium-on CXL in the presence of supplemental oxygen is efficacious in reducing corneal curvature, is safe and well tolerated in the treatment of keratoconus in adult and pediatric patients, and thus represents a promising new approach for treating keratoconus.
Acknowledgements
We thank all the investigators, their staff, and patients who participated in the clinical trial. Investigators for the trial included Zaina Al-Mohtaseb (Spring, TX), Brandon Ayres (Bala Cynwyd, PA), Brandon Baartman (Omaha, NE), Michael Robert Banitt (Seattle, WA), Kenneth A. Beckman (Westerville, OH), John Berdahl (Sioux Falls, SD), Michelle Boyce (Kansas City, MO), James Fox (Grand Junction, CO), Joseph Gira (St. Louis, MO), Steven A. Greenstein (Teaneck, NJ), Michael Greenwood (West Fargo, ND), Robert F. Haverly (Erie, PA), Sebastian Heersink (Dothan, AL), Mark Kelland Herschel (Winter Park, FL), Stephen Khachikian (Rapid City, SD), Brent Kramer (Alexandria, MN), James Lee (Colorado Springs, CO), Samuel Hyung Lee (Sacramento, CA), James Lehmann (San Antonio, TX), Eva Liang (Las Vegas, NV), Jason Mayer (Fort Collins, CO), Adam M. Moss (Edina, MN), John Steven Parker Jr (Vestavia Hills, AL), Gregory Parkhurst (San Antonio, TX), Michael Saidel (Petaluma, CA), Russell Swan (Bozeman, MT), Dan B. Tran (Orange, CA), and David T. Vroman (Ladson, SC).
Medical Writing/Editorial Assistance
Support for medical writing was provided by Sarah Cherukury, MPH, CHES of Glaukos Corporation.
Author Contributions
All authors met the International Committee of Medical Journal Editors (ICMJE) criteria for authorship. The study was designed by Valerie Smith, Angela C. Kothe and Tomas Navratil; data collection was performed by Kenneth A. Beckman, Gregory D. Parkhurst, James H. Lee, Zaina N. Al-Mohtaseb, Michael D. Greenwood, Valerie Smith and Marco Armijo; statistical analyses were performed by Erin H. Leone; data interpretation was performed by Erin H. Leone, Angela C. Kothe and Tomas Navratil. Authoring and/or critical revisions, and final approval of the manuscript were performed by Kenneth A. Beckman, Gregory D. Parkhurst, James H. Lee, Zaina N. Al-Mohtaseb, Michael D. Greenwood, Marco Armijo, Valerie Smith, Erin H. Leone, Angela C. Kothe and Tomas Navratil.
Funding
The study was sponsored by Glaukos Corporation (Aliso Viejo, CA, USA). Glaukos participated in the design of the study, data management, analysis and interpretations, and the preparation, review, and approval of the manuscript. The journal’s Rapid Service Free was funded by Glaukos.
Data Availability
The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.
Declarations
Conflict of Interest
The authors have made the following disclosures: Kenneth A. Beckman received financial and nonfinancial support from Glaukos, is a lecturer for Glaukos and is on the advisory board for Glaukos. Gregory D. Parkhurst and James H. Lee received financial support from Glaukos. Zaina N. Al-Mohtaseb is a consultant and lecturer for Glaukos. Michael D. Greenwood received financial and nonfinancial support from Glaukos, and is a consultant and lecturer for Glaukos. Marco Armijo, Valerie Smith, Erin H. Leone, Angela C. Kothe and Tomas Navratil are employees of Glaukos Corporation and may own stock and/or stock options.
Ethical Approval
The study was conducted in accordance with US regulations and the International Council for Harmonisation Guidelines on Good Clinical Practice which embody the tenets of the Declaration of Helsinki of 1964 and its later amendments. Institutional review board (IRB) approval (Alpha IRB, San Clemente, CA; GLK-202-02; approval date 12 December 2022) was obtained for the study. All patients provided written informed consent prior to their participation. Furthermore, those who were under the age of 18 years (or had not yet reached the age of majority per local regulations) signed the most recent IRB-approved assent form, and their parent or legal guardian signed the most recent IRB-approved informed consent form.
Footnotes
Prior Presentation: This manuscript is based on work that was presented in part at the 2025 annual meeting of the Association for Research in Vision and Ophthalmology (poster #B0083, May 4–8, 2025, Salt Lake City, UT) and the 2025 annual meeting of the American Academy of Ophthalmology (poster #PO212, October 17–20, 2025, Orlando, FL).
References
- 1.Fournié P, Acquadro M, Touboul D, et al. Keratoconus and the impact of treatment on patients’ quality of life: a qualitative study. Ophthalmol Ther. 2023;12(4):1939–56. 10.1007/s40123-023-00717-w. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Cortina MS, Greiner MA, Kuo AN, et al. Safety and efficacy of epithelium-off corneal collagen cross-linking for the treatment of corneal ectasia: a report by the American Academy of Ophthalmology. Ophthalmology. 2024;131(10):1234–42. 10.1016/j.ophtha.2024.05.006. [DOI] [PubMed] [Google Scholar]
- 3.Santodomingo-Rubido J, Carracedo G, Suzaki A, Villa-Collar C, Vincent SJ, Wolffsohn JS. Keratoconus: an updated review. Cont Lens Anterior Eye. 2022;45(3):101559. 10.1016/j.clae.2021.101559. [DOI] [PubMed] [Google Scholar]
- 4.Levy I, Mukhija R, Nanavaty MA. Corneal allogenic intrastromal ring segments. J Clin Med. 2025;14(4):1340. 10.3390/jcm14041340. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Greenstein SA, Yu AS, Gelles JD, Eshraghi H, Hersh PS. Corneal tissue addition keratoplasty: new intrastromal inlay procedure for keratoconus using femtosecond laser-shaped preserved corneal tissue. J Cataract Refract Surg. 2023;49(7):740–6. 10.1097/j.jcrs.0000000000001187. [DOI] [PubMed] [Google Scholar]
- 6.Hersh PS, Stulting RD, Muller D, et al. United States multicenter clinical trial of corneal collagen crosslinking for keratoconus treatment. Ophthalmology. 2017;124:1259–70. [DOI] [PubMed] [Google Scholar]
- 7.Larkin DFP, Chowdhury K, Burr JM, et al. Effect of corneal cross-linking versus standard care on keratoconus progression in young patients: the Keralink randomized controlled trial. Ophthalmology. 2021;128:1516–26. [DOI] [PubMed] [Google Scholar]
- 8.Borchert GA, Kandel H, Watson SL. Epithelium-on versus epithelium-off corneal collagen crosslinking for keratoconus: a systematic review and meta-analysis. Graefes Arch Clin Exp Ophthalmol. 2024;262(6):1683–92. 10.1007/s00417-023-06287. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Hafezi F, Kling S, Hafezi NL, et al. Corneal cross-linking. Prog Retin Eye Res. 2025;104:101322. 10.1016/j.preteyeres.2024.101322. [DOI] [PubMed] [Google Scholar]
- 10.Kamaev P, Friedman MD, Sherr E, Muller D. Photochemical kinetics of corneal cross-linking with riboflavin. Invest Ophthalmol Vis Sci. 2012;53(4):2360–7. 10.1167/iovs.11-9385. [DOI] [PubMed] [Google Scholar]
- 11.McCall AS, Kraft S, Edelhauser HF, et al. Mechanisms of corneal tissue cross-linking in response to treatment with topical riboflavin and long-wavelength ultraviolet radiation (UVA). Invest Ophthalmol Vis Sci. 2010;51(1):129–38. 10.1167/iovs.09-3738. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Raiskup F, Spoerl E. Corneal crosslinking with riboflavin and ultraviolet A. I. principles. Ocul Surf. 2013;11(2):65–74. 10.1016/j.jtos.2013.01.002. [DOI] [PubMed] [Google Scholar]
- 13.Raiskup F, Herber R, Lenk J, Pillunat LE, Spoerl E. Crosslinking with UV-A and riboflavin in progressive keratoconus: from laboratory to clinical practice - developments over 25 years. Prog Retin Eye Res. 2024;102:101276. 10.1016/j.preteyeres.2024.101276. [DOI] [PubMed] [Google Scholar]
- 14.Seiler TG, Komninou MA, Nambiar MH, et al. Oxygen kinetics during corneal cross-linking with and without supplementary oxygen. Am J Ophthalmol. 2021;223:368–76. 10.1016/j.ajo.2020.11.001. [DOI] [PubMed] [Google Scholar]
- 15.Hill J, Liu C, Deardorff P, et al. Optimization of oxygen dynamics, UV-A delivery, and drug formulation for accelerated epi-on corneal crosslinking. Curr Eye Res. 2020;45(4):450–8. 10.1080/02713683.2019.1669663. [DOI] [PubMed] [Google Scholar]
- 16.McMahon TT, Szczotka-Flynn L, Barr JT, et al. A new method for grading the severity of keratoconus: the Keratoconus Severity Score. Cornea. 2006;25:794–800. [DOI] [PubMed] [Google Scholar]
- 17.Berglund P, Heeringa S. Multiple imputation of missing data using SAS®, SAS® Institute, 2014.
- 18.Rapuano CJ, Lindstrom RL, Donnenfeld E, et al. Economics of corneal cross-linking for keratoconus treatment. J Med Econ. 2025. 10.1080/13696998.2025.2564576. [DOI] [PubMed] [Google Scholar]
- 19.D’Oria F, Palazón A, Alio JL. Corneal collagen cross-linking epithelium-on vs. epithelium-off: a systematic review and meta-analysis. Eye Vis (Lond). 2021;8(1):34. 10.1186/s40662-021-00256-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Borchert GA, Watson SL, Kandel H. Oxygen in corneal collagen crosslinking to treat keratoconus: a systematic review and meta-analysis. Asia Pac J Ophthalmol (Phila). 2022;11:453–9. [DOI] [PubMed] [Google Scholar]
- 21.Mazzotta C, Sgheri A, Bagaglia SA, et al. Customized corneal cross-linking for treatment of progressive keratoconus: clinical and OCT outcomes using a transepithelial approach with supplemental oxygen. J Cataract Refract Surg. 2020;46:1582–7. [DOI] [PubMed] [Google Scholar]
- 22.Ng SM, Ren M, Lindsley KB, Hawkins BS, Kuo IC. Transepithelial versus epithelium-off corneal crosslinking for progressive keratoconu. Cochrane Database System Rev. 2021. 10.1002/14651858.CD013512.pub2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Spadea L, Salvatore S, Paroli MP, Vingolo EM. Recovery of corneal sensitivity after collagen crosslinking with and without epithelial debridement in eyes with keratoconus. J Cataract Refract Surg. 2015;41(3):527–32. 10.1016/j.jcrs.2014.06.030. [DOI] [PubMed] [Google Scholar]
- 24.Farrokhpour H, Soleimani M, Cheraqpour K, et al. A case series of infectious keratitis after corneal cross-linking. J Refract Surg. 2023;39(8):564–72. [DOI] [PubMed] [Google Scholar]
- 25.Serrao S, Lombardo G, Lombardo M. Adverse events after riboflavin/UV-A corneal cross-linking: a literature review. Int Ophthalmol. 2022;42(1):337–48. 10.1007/s10792-021-02019-1. [DOI] [PubMed] [Google Scholar]
- 26.Murchison CE, Petroll WM, Robertson DM. Infectious keratitis after corneal crosslinking: systematic review. J Cataract Refract Surg. 2021;47(8):1075–80. 10.1097/j.jcrs.0000000000000620. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Data Availability Statement
The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.