The Science and Clinical Evolution of Corneal Cross-Linking: Mechanism of Action, Ultra-Structural Changes, Clinical Indications, and Emerging Treatment Strategies

Abstract

Keratoconus is the most common corneal ectatic disorder worldwide. Typically manifesting in adolescence, it imposes a substantial burden on quality of life and, if untreated, may progress to advanced stages requiring corneal transplantation. Riboflavin/UV-A corneal cross-linking (CXL) has become the cornerstone therapy for stabilizing corneal biomechanics and halting disease progression. CXL acts through UV-A photo-activated riboflavin-driven photochemistry that induces covalent cross-links within stromal proteins. Extensive theoretical, experimental, and clinical evidence has clarified its mechanism of action, demonstrating ultrastructural stromal reorganization, resulting in biomechanical stiffening and a sustained long-term effect. Over time, riboflavin ophthalmic formulations for CXL, delivery strategies, and treatment protocols have progressed and expanded beyond the epithelium-off approach to include safer and comparably effective transepithelial techniques. In addition, novel chemical cross-linkers are being investigated as non-light-based alternatives for corneal stabilization. Beyond corneal ectatic disorders, CXL is increasingly being explored for additional clinical indications, including infectious keratitis, bullous keratopathy, and inflammatory dry eye. This systematic review describes the most recent advances in understanding the mechanism of action of CXL and associated ultrastructural changes and synthesizes current evidence on the evolution of treatment protocols and their expanding clinical indications. By critically appraising available data, we aim to guide clinical decision-making and identify priorities toward safer, more effective, and individualized CXL therapies.

Key Summary Points

Corneal cross-linking (CXL) has evolved from the original Dresden protocol into a broad therapeutic platform for keratoconus and other corneal disorders through advances in riboflavin delivery, UV-A photoactivation, and personalized treatment strategies.

Emerging transepithelial approaches have improved the safety profile of CXL while progressively narrowing the efficacy gap with conventional epithelium-off protocols.

Theranostic-guided corneal UV therapy enables real-time monitoring of stromal riboflavin concentration and photoactivation, supporting personalized and precise treatment strategies and enabling accurate prediction of the CXL outcome.

Emerging CXL strategies are exploring spatially controlled riboflavin delivery and adaptive UV-A-mediated photoactivation of riboflavin to enable patient-specific biomechanical corneal remodeling based on individual clinical and tomographic characteristics.

CXL is performed in indications beyond corneal ectasia, including as an adjunctive treatment for infectious keratitis and in selected cases of bullous keratopathy and dry eye syndrome.

Introduction

Keratoconus is a progressive, usually bilateral, chronic disease of the cornea characterized by stromal thinning, biochemical and biomechanical tissue weakening that leads to corneal bulging and visual impairment. The etiopathogenesis of keratoconus is multifactorial, including genetic and environmental risk factors. A familial history is present in approximately 18% of cases, as reported in various epidemiological studies, underscoring the importance of hereditary factors in the disease’s development [21, 112, 126]. In addition to its genetic heterogeneity, several environmental factors, notably dysregulated immunity responses associated with allergy and the habit of eye rubbing, play a significant role in triggering or accelerating disease progression [31, 113, 162]. Keratoconus often has an insidious clinical onset in adolescence and manifests with progressive visual decline caused by irregular astigmatism, myopia, high-order aberrations (HOAs), and in advanced cases, corneal scarring. Due to its onset at a young age, the disease can have a profound negative impact on patients’ quality of life [60, 85].

Treatment of keratoconus begins with glasses or contact lenses to improve visual acuity. While contact lenses can often provide improved visual performance, up to 40% of patients may not wear them due to poor tolerance, ocular surface discomfort, difficulty with fitting and handling, or advancing corneal irregularity. Moreover, contact lenses do not affect the natural course of the disease or halt its progression, which remains closely associated with worsening visual function over time [12]. Para-surgical intervention, such as corneal cross-linking with riboflavin and ultraviolet-A (UV-A) light (CXL), aims to stabilize the corneal structure and biomechanics, thereby slowing or arresting disease progression. In more advanced cases, corneal transplantation, either deep anterior lamellar keratoplasty (DALK) or, less frequently, penetrating keratoplasty (PK), may be required to replace severely damaged corneal tissue and restore vision. However, corneal transplantation remains associated with significant long-term risks and graft failure, reinforcing the importance of early disease stabilization. Notably, the risk for re-grafting in keratoconus is higher than 30% over a long-term period (≥ 10 years), which makes corneal graft not a definite solution for the disease since the risk of graft rejection persists throughout life [50, 206]. For these reasons, an early, safe, and effective intervention, such as CXL, represents a fundamental strategy for reducing disease burden. By slowing ectatic progression, CXL may reduce both visual disability and healthcare burden [30, 57].

The CXL procedure was originally designed for the treatment of corneal ectatic disorders, such as keratoconus and post-laser refractive surgery ectasia. More recently, it has shown promise in additional indications, including the treatment of bullous keratopathy and the decontamination of corneal tissue in cases of infectious keratitis caused by pathogenic microorganisms [194]. New indications of crosslinking are emerging in the field of inflammatory dry eye management, due to its capacity in reducing metalloproteinases activity and ocular surface inflammation [129].

Since its introduction in 2003 in Europe, and its later adoption in the United States in 2016, CXL has become a widely studied and accepted treatment. To date, over 3000 scientific publications have documented its experimental and clinical outcomes. The original Dresden protocol involves epithelial removal (epi-off), followed by 30 min of 20% dextran-enriched 0.1% riboflavin and UV-A irradiation at 3 mW/cm² for 30 min (total dose 5.4 J/cm²). Despite its efficacy, epi-off CXL carries inherent risks. The most frequent adverse events are mild and transient, particularly corneal haze and corneal edema, which generally resolve completely within 1 year. Severe adverse events leading to permanent sequelae are uncommon, with a cumulative incidence ≤ 1.5%, including infectious keratitis (0.6%) and corneal scarring (0.5%) [174]. Furthermore, the conventional Dresden protocol requires a minimum stromal thickness of 400 µm after de-epithelialization to minimize the risk of endothelial damage, a concern that may be exacerbated by dextran-induced intraoperative corneal thinning and inadequate riboflavin saturation [54]. To overcome these limitations, transepithelial (epi-on) protocols have been developed to preserve the epithelium, reduce complications, and expand eligibility to thinner corneas, although their clinical efficacy has only recently improved. Advances in epi-on CXL have addressed earlier limitations through improved riboflavin formulations, optimized delivery methods, and refined UV irradiation, enabling treatments increasingly tailored to the individual cornea [24, 36, 156]. Beyond UV-A light-induced cross-linking, emerging chemical cross-linkers offer a promising avenue for corneal stabilization, potentially providing non-light-based strategies to strengthen the cornea and complement existing CXL therapies.

In this systematic review, we aim to provide a comprehensive overview of the evolving role of CXL in the treatment of keratoconus and corneal ectatic disorders, as well as in other emerging indications such as bullous keratopathy, infectious keratitis, and inflammatory dry eye. We update current knowledge on the mechanism of action of CXL and associated ultrastructural changes and summarize evidence from experimental and clinical studies on treatment protocols. By critically evaluating the available evidence, this review aims to inform clinical decision-making and identify future directions for more personalized treatment strategies of pathological corneal conditions.

Basic Science of Corneal Cross-Linking

Photochemistry of CXL

Riboflavin is a molecule classified as one of the B vitamins (B₂). It is a heterocyclic compound consisting of a flavin molecule to which a chain formed by ribitol is attached [89]. It is a unique photochemical agent because it can act as a photosensitizing agent and the light-excited molecule itself can also act as an electron donor [76]. Riboflavin absorbs maximally in the blue (λ_max = 445 nm) and in the near ultraviolet (λ_max = 373 nm); it also has two absorbance peaks in UV-C at about 260- and 225-nm [115, 182]. When excited by near ultraviolet or by blue light, riboflavin exhibits a strong green fluorescence; the UV-induced fluorescence emission has a maximum at 534 nm (range, 480–600 nm) [79]. Upon UV-A irradiation, riboflavin in its excited triplet state can trigger two complementary photochemical pathways, type I and type II mechanisms, and can photosensitize the oxidization of a great variety of substrates, such as amino acids (Fig. 1) [32, 115, 140, 159].

Fig. 1.

Schematic representation of the riboflavin/UV-A light-induced photochemical reaction underlying corneal cross-linking. The diagram illustrates the type I (³RF) and type II (¹O₂) photo-chemical pathways, the generation of reactive oxygen species generation (ROS; ³O₂ and ¹O₂), and the formation of covalent chemical bonds between stromal proteins (CH = N), which collectively lead to biomechanical stiffening of the corneal tissue. UV-A light absorbed by riboflavin-enriched corneal tissue, either with or without intact epithelium, initiates the photo-polymerization process

In the type I mechanism, triplet riboflavin abstracts an electron or hydrogen atom from the substrate, forming riboflavin radicals and substrate radicals. These reactive intermediates, particularly involving amino acids such as lysine, hydroxylysine, arginine, cysteine and tyrosine, can recombine directly or react with oxidized groups to form new covalent bonds or undergo further reactions. A key outcome is the formation of Schiff bases between lysine or hydroxylysine residues, which contributes to the generation of cross-linking bonds among collagen fibrils and proteoglycan core proteins. This radical-mediated process is favored when riboflavin concentration is high, can proceed under hypoxic conditions and is considered the main pathway responsible for the biomechanical stiffening effect of CXL [107, 110, 171]. Oxygen may, however, contribute to riboflavin regeneration by oxidizing both riboflavin radicals and reduced riboflavin, thereby slowing the decrease in riboflavin concentration. This allows more riboflavin molecules to be excited to high-energy states, promoting the crosslinking reaction [185].

In the type II mechanism, the excited triplet riboflavin transfers energy to molecular oxygen, generating singlet oxygen (¹O₂) as the main reactive oxygen species (ROS), and to a lesser extent superoxide anion (O^2–) [84, 173]. Singlet oxygen is highly electrophilic and reacts with susceptible amino acid side chains, including methionine, cysteine, and aromatic amino acids, in particular tyrosine. These oxidative modifications promote the formation of covalent cross-links, such as dityrosine bonds and methionine sulfoxide bridges. Because oxygen availability can influence this pathway by affecting the steady-state concentration of tyrosyl radicals and their dimerization efficiency, its contribution is greater where oxygen tension is relatively high [84, 172, 185]. However, oxygen has poor solubility in the aqueous stromal environment, and under the highly photo-oxidative conditions generated by UV-A light exposure in the presence of riboflavin, it is rapidly and irreversibly consumed. Consequently, the pathway’s activity is transient, producing an immediate, localized depletion zone that restricts both the spatial extent and the uniformity of oxygen-mediated cross-linking to the most anterior stromal layers. Clinical strategies such as increasing ambient oxygen concentration and using pulsed UV irradiation have been explored to modulate intrastromal oxygen availability, although reported results remain contradictory. Notably, excess oxygen may even reduce cross-linking efficiency by promoting radical quenching rather than bond formation [84].

Overall, the relative contribution of type I and type II mechanisms to the formation of corneal cross-linking bonds depends on riboflavin concentration, UV-A irradiance and effective riboflavin photoactivation, oxygen availability, and the abundance of oxidizable substrates, particularly lysine, hydroxylysine, methionine, and tyrosine. Type II can contribute at the beginning of UV irradiation and in the most anterior stroma, while type I is more relevant as oxygen is consumed and in anterior as well as deeper stromal regions. Enhancing riboflavin penetration may therefore strengthen type I-mediated cross-linking, overall improving the homogeneity of biomechanical stiffening across the corneal thickness. It is now well established that in the presence of an adequate stromal concentration of riboflavin (≥ 0.012%), the type I direct photochemical pathway plays a central role in generating additional cross-linking bonds among stromal proteins [110]. Experimental studies on human donor eyes have shown that corneal riboflavin concentration decreases exponentially during UV-A irradiation at both 3 mW/cm² and 10 mW/cm² for total 5.4 J/cm² energy dose [118]. This behavior is consistent with the knowledge that the UV-A light-mediated photo-degradation of riboflavin in the stroma is spatially homogeneous and temporally monotonic decreasing [114, 118, 184]. Since the intrastromal riboflavin concentration is relatively low (≤ 0.04%), the overall photochemical process can be considered pseudo-first order under anaerobic conditions [114, 118]; consequently, an exponential kinetic consumption of riboflavin is expected. This implies that, in an ambient environment (i.e., 21% partial pressure of oxygen), the role of type I mechanism is predominant in driving the formation of new chemical bonds between stromal proteins during CXL; in addition, the concentration of riboflavin directly influences the efficiency of oxygen utilization [84, 118, 185]. However, there remains a lack of studies systematically assessing the effects of oxygen supplementation in relation to quantified corneal riboflavin concentrations. Further studies are needed to elucidate the effect of high-irradiance and/or high-energy dose UV-A protocols on the kinetics of CXL and the relative contributions of type I and type II pathways on clinical outcomes.

Ultra-Structural and Biomechanical Changes Induced by CXL

Keratoconic corneas themselves represent a challenge. The biomechanical, structural, and morphological properties of the keratoconic cornea vary widely between patients, and even across regions of the same cornea. Beyond inter- and intra-corneal variability, the photochemistry of CXL is strongly influenced by the penetration of riboflavin, the absorption of UV-A light radiation, and local oxygen availability, all of which ultimately determine the depth and magnitude of stromal changes.

Experimental studies have demonstrated that CXL significantly enhances the biomechanical strength of the corneal tissue by increasing its rigidity, inducing ultrastructural changes such as increased interfibrillar spacing and greater collagen fiber diameter [15]. Translating these findings into clinical practice, the improved mechanical resistance enables the cornea to better withstand normal physiological stress, such as intraocular pressure, thereby stabilizing the progression of keratoconus and other corneal ectatic disorders.

On the ultra-structural level, one of the earliest and most consistent findings after CXL is keratocyte apoptosis within the anterior stroma [142, 168]; this process occurs almost immediately following treatment. Repopulation by activated keratocytes begins at deeper stromal layers after one to three months and is associated with increased stromal reflectivity and transient corneal haze. Most corneas show substantial recovery by six to twelve months after CXL. Sub-basal nerve fibers are also affected, with early degeneration followed by slow regeneration over months to years, explaining the transient reduction in corneal sensitivity and changes in nerve tortuosity observed with confocal microscopy that may persist long term.

Collagen fibril architecture undergoes measurable remodeling after CXL. Studies using sophisticated imaging techniques (electron microscopy, two-photon microscopy, X-ray scattering) have documented nearly constant riboflavin diffusion between 100 µm and 250 µm in epi-on delivery protocols and to 300 µm in epi-off delivery protocols before declining toward the endothelium. Following UV-A irradiation, these studies also reported modest increases in fibril diameter, reduced interfibrillar spacing, and enhanced packing regularity within the treated stroma [3, 20, 72]. Anterior collagen lamellae demonstrate compaction and improved interlamellar cohesion, changes that are believed to reduce the tendency for lamellar slippage under stress. Biochemically, there is evidence for new covalent bonds among collagen fibrils resembling those formed by natural enzymatic processes, such as lysyl oxidase activity [141]. These changes improve biomechanical tissue strength and increase resistance to enzymatic digestion, overall conferring greater structural integrity [183]. Proteoglycans also participate in the remodeling process; CXL has been shown to upregulate small leucine-rich proteoglycans, such as lumican, decorin, and keratocan, while downregulating matrix metalloproteinases, which further stabilizes the extracellular matrix and contributes to improved fibril spacing and transparency [208]. Based on these insights, in 2015 the authors proposed revising the terminology of the procedure, recommending that the term “collagen” be omitted from “corneal collagen cross-linking” in order to more accurately reflect the procedure’s broader impact on the corneal proteins and extracellular matrix [115, 116].

The mentioned ultra-structural modifications translate into measurable biomechanical improvements. Ex vivo experiments using several testing methodologies (atomic force microscopy, strip extensiometry, inflation testing, dynamic tonometry, Brillouin microscopy, optical coherence elastography) demonstrated a significant increase in tissue elasticity and decrease in hysteresis after CXL, with effects concentrated in the anterior 200 μm of the stroma [15, 96]. Atomic force microscopy (AFM) has been pivotal in quantifying nano-scale biomechanical effects of CXL in human corneas, determining the supra-molecular changes in collagen fibril/bundle mechanics and the surrounding proteoglycan matrix [38, 101, 102]. Localized stiffening of the anterior stroma has also been confirmed by ex vivo depth-resolved Brillouin microscopy and by in vivo optical coherence elastography imaging, with both techniques independently validating this finding [51, 170]. Several testing techniques further confirmed increases in tangent, tensile and shear moduli, highlighting the anisotropic and depth-dependent nature of CXL-induced stiffening [15, 69, 93].

Taken together, ultrastructural and biomechanical studies have shown that CXL induces measurable, depth-dependent increases in stromal stiffness, with the greatest effect localized to the anterior 200 μm [205]. These changes arise from a cascade of molecular and supramolecular modifications, including changes in collagen fibrils’ interactions, proteoglycan stabilization, and reduced tissue viscoelasticity, which are directly linked to the spatial dynamics of riboflavin distribution and UV-A light absorption [79, 114, 118, 171, 185]. Nonetheless, the influence of heterogeneity in baseline ultrastructure, biomechanics, and treatment response remains incompletely understood. The traditional empirical ‘one-protocol-fits-all’ CXL treatment approach cannot adequately account for variability in corneal architecture and biomechanics, nor does it consider riboflavin stromal distribution and UV-A-mediated photoactivation, thereby contributing to the unpredictability of treatment outcomes and reinforcing the need for tailored strategies to maximize the therapeutic benefit of CXL for each patient.

Clinical Science and Evidence of Corneal Cross-Linking

Eligibility Criteria for CXL

Current recommendations for eligibility to perform CXL prioritize treatment of progressive keratoconus. Nevertheless, definition of disease progression remains a subject of ongoing debate [59]. The most widely accepted approach involves documenting progression through multiple single indicators assessed over at least two consecutive visits within the past 12 months or a longer observation period.

The K_max index remains the most endorsed single metric for detecting clinically meaningful steepening of the cornea and is therefore retained as the primary endpoint for defining progression, supported by additional confirmatory structural and functional parameters to ensure diagnostic robustness. Currently widely accepted eligibility criteria are summarized in Table 1. Candidates must have evidence of manifest keratoconus (i.e., K_max ≥ 47 D) and sufficient corneal thickness, which is typically required to be ≥ 400 µm at the thinnest point, although the most recent 0.22% riboflavin formulations are approved for use in thinner corneas, down to 370 µm [117].

Table 1.

Parameters used to define progressive keratoconus and determine eligibility to corneal cross-linking (CXL) treatment

Parameter	Parameter threshold value
Maximum keratometry index (Kmax)	≥ 47.0 D
Corneal thickness (thinnest point)	≥ 400 µm
Kmax change*	Steepening of ≥ 1.0 D within the past 12 months
Central corneal thickness (CCT) change	Decrease of ≥ 10 µm within the past 12 months
Manifest refraction spherical equivalent (MRSE)	Increase in myopic MRSE of ≥ 0.5 D within the past 12 months
Manifest astigmatism (cylinder)	Increase in manifest astigmatism of ≥ 0.5 D within the past 12 months
Uncorrected distance visual acuity (UDVA)	Reduction of ≥ 1 ETDRS line within the past 12 months
Corrected distance visual acuity (CDVA)	Reduction of ≥ 1 ETDRS line within the past 12 months
Steepest 3-mm central simulated keratometry (steep Ksim) index	Steepening of ≥ 1.0 D within the past 12 months
Mean 3-mm central Ksim index	Steepening of ≥ 1.0 D within the past 12 months

*The K_max change is the primary determinant for defining disease progression and CXL treatment eligibility, accompanied by at least one additional supporting structural or functional parameter recorded over the same period

Several authors have emphasized the need to improve the prediction of disease progression and optimize the timing of CXL treatment by incorporating multiple tomographic biomarkers and novel analytical approaches based on zonal indices of anterior curvature and corneal thickness [82]. Recent studies have shown that zonal average analysis of corneal tomography data surrounding the point of maximum curvature is more sensitive and accurate than single-point indices for detecting progressive disease over long-term observation [82, 123]. Likewise, the ABCD grading system, which evaluates a 3-mm optical zone centered on the thinnest corneal point, has been shown to be more sensitive to identify disease progression than K_max alone [196].

There is broad expert consensus that CXL should not be restricted by patient age in keratoconic eyes showing evidence of progression, although most patients are treated between 12 and 40 years [59]. Given the higher likelihood of rapid progression in younger individuals, experts recommend performing CXL even at first diagnosis in adolescents with high-risk factors, such as a positive family history, atopy, allergy, chronic conjunctivitis, blepharitis, or associated skin diseases. Disease progression within 1 year has been reported in up to 88% of pediatric patients, and keratoconus remains one of the most frequent indications for corneal transplantation in childhood, accounting for approximately 15–20% of all corneal transplants performed in individuals under 18 years of age [27, 95].

The Original Dresden Protocol

Riboflavin/UV-A corneal cross-linking according to the original Dresden protocol represents the historical cornerstone of keratoconus management. The procedure was first introduced by Wollensak, Spoerl, and Seiler in 2003 at the University of Dresden, marking the first clinical translation of photochemical tissue stabilization into ophthalmology. The original technique involves the removal of the central 9 mm of corneal epithelium (“epithelium-off” or “epi-off”) followed by application of a 0.1% riboflavin solution in 20% dextran T-500 onto the stroma for 30 min. Subsequently, the cornea is irradiated with UV-A light (370 ± 5 nm wavelength) at an irradiance of 3 mW/cm² for 30 min, delivering a total energy dose of 5.4 J/cm² [203].

The clinical efficacy of the original Dresden protocol has been established across numerous randomized controlled trials and large case series. In progressive keratoconus, stabilization or improvement in maximum keratometry (K_max) is observed in a median of 70% of eyes at 1 year and longer period [73, 156, 202]. Numerous long-term studies demonstrate sustained disease control, minimal endothelial cell loss, and low rates of retreatment up to 15 years after treatment [147, 156, 174, 175]. Failure rates, defined as an increase in K_max ≥ 1.0 D or continued visual deterioration, ranged from 14 to 33%, depending on disease severity, corneal thickness, and preoperative K_max [49, 97, 156].

The most frequent complications included delayed epithelial healing (~ 5%), sterile corneal infiltrates (~ 2%), infectious keratitis (~ 1%), transient corneal haze (≥ 30%), and corneal scarring (~ 1%) [27, 37, 97, 174]. Rare but severe events, such as corneal melt and endothelial decompensation have been described mostly in females with thin corneas or following infection, or abuse of nonsteroidal anti-inflammatory drug (NSAID) eye drops or improper riboflavin application and/or UV irradiation [97, 174, 188].

Despite its proven efficacy and status as the benchmark for corneal cross-linking, the standard Dresden protocol has several limitations. These include the long duration of the procedure and the requirement for epithelial removal, which can cause postoperative pain, increase the risk of infection, lead to corneal scarring, and delay visual recovery [97]. In addition, dextran-induced corneal dehydration during riboflavin soaking may result in significant stromal thinning, increasing the risk of endothelial damage if corneal thickness falls below 400 µm [204]. While the Dresden protocol’s long-term safety and efficacy are well established, these concerns have prompted multiple protocol modifications aimed at reducing treatment time, improving patient comfort, and extending eligibility to thinner corneas. Such advancements have led to accelerated, transepithelial, and personalized approaches, which are described in the following sections.

“Accelerated” CXL Protocols

To reduce the treatment time of the standard Dresden protocol, accelerated CXL (A-CXL) protocols were developed based on the Bunsen–Roscoe law of reciprocity, which assumes that equivalent total fluence yields comparable biological effects when irradiance and exposure time are inversely adjusted. Common regimens include 9 mW/cm² for 10 min, 18 mW/cm² for 5 min, and 30 mW/cm² for 3 min (5.4 J/cm² total dose).

However, experimental and clinical data indicate that reciprocity is only partially valid, as oxygen availability and riboflavin photochemistry significantly influence efficacy. The law of reciprocity between irradiance and exposure time applies only within a limited range, as shown by Lombardo et al. [118] using theranostic UV device. These findings explain why the 9 mW/cm² for 10 min regimen is now regarded as the optimal accelerated protocol, providing an effective balance between treatment efficiency and safety [70, 130, 179, 192].

Higher irradiances (≥ 18 mW/cm²) have been associated with reduced biomechanical stiffening and shallower stromal demarcation lines, and irradiances ≥ 45 mW/cm² have been reported not to further increase corneal stiffness [65, 98, 201].

Despite these limitations, A-CXL is widely adopted due to shorter procedure time, improved comfort, and generally comparable efficacy and safety to the standard protocol. Meta-analyses show that 9 mW/cm² for 10 min achieves outcomes similar to the standard Dresden protocol CXL, while higher irradiances yield more variable results. The 9- or 10- mW/cm² protocols are therefore considered the optimal balance between efficiency and predictability [65, 70, 98, 118, 128, 130, 134, 177, 179, 192].

Thin corneas require individualized protocols to avoid excessive photochemical effects and endothelial risk. Pachymetry-guided approaches, such as the M nomogram, adjust irradiance and exposure time while maintaining a fixed fluence of 5.4 J/cm² (Table 2) [134]. These strategies improve control of cross-linking depth and have shown favorable long-term outcomes in keratoconus and post-refractive ectasia [134]

Table 2.

Pachymetry-based epithelium-off accelerated corneal cross-linking: M Nomogram*

Parameter	Variables (Thin cornea)	Variables (Intermediate cornea)	Variables (Standard cornea)
Corneal thinnest point (µm)	≤ 350 µm	> 350µm < 400µm	≥ 400 µm
Fluence (total) (J/cm2)	5.4J/cm2	5.4 J/cm2	5.4 J/cm2
Irradiance (mW/cm2)	30 mW/cm2	15 mW/cm2	9 mW/cm2
Irradiation time	6 min	12 min	10 min
Irradiation mode (interval)	Pulsed (1 s on, -1 s off)	Pulsed (1 s on, 1 s off)	Continuous

*The M Nomogram has been validated using isotonic, 1% HPMC enriched, 0.1% riboflavin-5-phosphate applied for 10 min onto the de-epithelialized cornea

Selective Epithelial Removal and Excimer Laser-Assisted Techniques in Corneal Cross-Linking

Several CXL approaches have incorporated selective epithelial removal strategies to balance treatment efficacy with faster visual recovery and reduced postoperative discomfort. Different patterns of partial de-epithelialization have been explored, including vertical stripe debridement techniques designed to facilitate riboflavin penetration through exposed stromal areas while preserving epithelial bridges that may promote faster epithelial healing. In a matched comparative study, Hashemi et al. [67] reported that both total and partial epithelial removal were effective in stabilizing keratoconus progression, with mean K_max changes of 0.4 D and 0.0 D, respectively. Partial debridement was associated with slightly greater improvement in corrected distance visual acuity and better preservation of central corneal thickness. Similarly, Galvis et al. [55] reported long-term ectasia stabilization over a 5-year follow-up using a partial de-epithelialization approach.

Other geometry-based approaches have included preservation of a central epithelial island to protect the visual axis and improve postoperative recovery. This strategy has been associated with reduced postoperative discomfort, lower rate of corneal haze, and faster visual rehabilitation. In a recent comparative clinical trial, Khalil et al. [83] reported effective stabilization of paracentral progressive keratoconus with significantly lower postoperative pain and faster visual recovery compared with conventional complete epithelial removal.

Beyond manual techniques, excimer laser-assisted epithelial removal has further refined debridement strategies through transepithelial phototherapeutic keratectomy (t-PTK). Unlike conventional mechanical debridement, t-PTK enables highly uniform epithelial removal at a predetermined depth. Comparative studies between t-PTK CXL (targeted ablation depth of 50 μm over 7 mm optical zone) and conventional CXL have demonstrated favorable visual and keratometric outcomes compared with the t-PTK approach [77]. Although termed “epi-on”, PTK-assisted customized epi-on (PACE) CXL has been proposed to selectively remove the epithelium over the cone apex based on OCT epithelial mapping. However, no clinical data have been published to date. Therefore, prospective controlled studies are needed to establish the long-term efficacy and potential clinical advantages of partial epithelial debridement approaches compared with conventional CXL protocols [77].

Transepithelial CXL Protocols

Transepithelial (“epithelium-on” or “epi-on”) CXL was developed as a less invasive alternative to the epi-off procedure, with the aim of improving patient comfort, accelerating visual recovery, and reducing the risk of complications associated with epithelial removal.

The original transepithelial CXL (2010) used 15% dextran-enriched 0.1% riboflavin applied to intact epithelium but showed poor stromal penetration due to the epithelial barrier and hypertonic dextran-based formulation. Stabilization rates were low (~ 10%), highlighting the need for improved formulations and delivery strategies [23, 29, 176, 181]. The intact epithelium acts as a substantial barrier to dextran-enriched riboflavin formulation, resulting in very low or virtually no stromal concentration of the photosensitizer.

Over the past two decades, substantial efforts have focused on optimizing riboflavin formulations to enhance corneal penetration through the intact epithelium and to achieve reproducible and effective outcomes with transepithelial CXL techniques. New formulations (dextran-free, hypotonic, higher riboflavin concentration, and permeation enhancers such as ethylenediaminetetraacetic acid (EDTA), trometamol (Tris), sodium chloride (NaCl), benzalkonium chloride (BAK), sodium iodide (NaI), and vitamin E TPGS) improved epithelial penetration, with hypotonicity and increased riboflavin concentration being key determinants, as hypotonicity creates an osmotic gradient that facilitates the penetration of riboflavin through epithelial tight junctions [24, 25, 71, 79, 90, 117, 157]. Viscoelastic-free, high-concentration formulations (0.22% riboflavin) achieved the best stromal uptake, explaining variability in clinical outcomes [71, 117]. These differences underscored the critical importance of carefully considering the composition and physicochemical properties of riboflavin formulations among different products, which may partly explain the high variability in clinical outcomes.

To further enhance riboflavin penetration through the epithelium, corneal iontophoresis-assisted delivery has been developed [46, 61]. This method applies a mild electrical current to drive riboflavin into the stroma, achieving higher intrastromal concentrations than manual passive diffusion while maintaining epithelial integrity [8, 119]. The ionized substances are driven into the tissue primarily by two mechanisms, such as electromigration (i.e., the motion of molecules based on flow of current) and electro-osmosis (i.e., the motion of the solvent). Experimental studies have demonstrated that although the stromal concentration of riboflavin after corneal iontophoresis was lower than conventional epi-off delivery, the stiffening effect on the anterior cornea (after 10 mW/cm² for 9 min UV-A irradiation) was almost comparable to that of conventional epi-off CXL [124]. In addition, clinical studies have assessed the long-term safety and efficacy of this approach in the treatment of progressive keratoconus with transepithelial CXL approach [16, 22, 119, 120, 135, 163, 197]. Long-term studies, including the RCT NCT02117999 have shown considerable variability, with K_max flattening ranging from – 0.3 D to – 1.1 D and progression rates exceeding 25%, particularly in younger patients who exhibit higher failure rates [22, 36, 119, 120, 163].

Intraoperative oxygen supplementation has also been proposed to improve clinical outcomes in transepithelial CXL approaches [159]. Laboratory studies demonstrated that delivering oxygen directly to the corneal surface during irradiation increases the availability of reactive oxygen species (ROS) and enhances the biomechanical stiffening effect of CXL [75]. Clinical translation of these findings has confirmed that oxygen supplementation can improve outcomes, especially in accelerated and transepithelial protocols [132]. Mazzotta et al. [132] reported significant K_max flattening when oxygen was continuously delivered during pulsed UV-A exposure in a high-irradiance transepithelial protocol (30 mW/cm² power density for total 7.2 J/cm² energy dose). These results suggested that maintaining adequate oxygen concentration at the corneal surface could restore the efficiency of the type II photochemical pathway and enhance the transepithelial cross-linking effect, particularly in conditions where oxygen diffusion is limited, such as during high-irradiance UV-A exposure. In this view, recent experimental studies and mathematical models have shown that oxygen may contribute to riboflavin regeneration by oxidizing both riboflavin radicals and reduced riboflavin, thereby sustaining a type I photopolymerization mechanism and prolonging the availability of reactive species during UV-A irradiation [185]. Taken together, these findings support the concept that controlled oxygen availability may have implications for optimizing treatment protocols. However, further evidence is needed, and real-time monitoring of riboflavin during CXL (as enabled by emerging theranostic technology) may help clarify the relative contributions of type I and type II mechanisms to corneal cross-link formation, particularly when intraoperative oxygen supplementation is used to improve the efficacy and reproducibility of epi-on CXL protocols.

Pulsed UV-A irradiation, typically using a 1-s on/1-s off pattern, has been shown in a few clinical studies to produce deeper stromal demarcation lines compared with continuous exposure at the same total fluence [18, 131, 143, 154]. The rationale for high-fluence and pulsed irradiation protocols derives from the dynamic interplay between UV-A exposure, oxygen consumption, and replenishment. Continuous high-intensity irradiation rapidly depletes stromal oxygen, limiting ROS-mediated cross-linking photo-chemistry; in contrast, introducing intermittent “off” phases between light pulses may allow partial oxygen re-diffusion into the stroma, sustaining the type II photo-polymerization pathway [84, 185, 201]. Authors demonstrated that pulsed irradiation at either 10 mW/cm² or 30 mW/cm² for total 5.4 J/cm² energy dose achieved a demarcation line depth of 200–250 µm, significantly deeper than that observed with continuous light exposure (150–180 µm). However, under laboratory conditions, pulsed UV-A irradiation alone did not consistently enhance corneal biomechanical strength compared with continuous exposure [94].

Two identical phase 3 RCTs of epi-on CXL have been recently completed in the U.S with FDA approval. Both the NCT03442751 and NCT05759559 trials enrolled 201 subjects with progressive keratoconus, aged 12–55 years old (n = 280 eyes), using a 2:1 allocation to CXL versus sham treatment. The primary endpoint was the change in K_max at 12 months, reporting an average flattening of 0.4 D in the treatment group compared with an average steepening of 0.7 D in controls. Severe adverse events occurred in 0.5% of treated eyes, and overall safety was generally acceptable. While peer-reviewed publications are not yet available, the study results are publicly reported on clinicaltrials.gov. Two more phase 3 RCTs of epi-on CXL are ongoing in the U.S; the APRICITY trials (NCT06100939 and NCT06100952) are two identical multicenter, double-masked, sham-controlled studies in patients aged 8–45 years old. Up to 800 patients have been enrolled across 26 clinics, randomized 1:1 to active treatment or sham, with bilateral treatments permitted. Primary endpoint focuses on 12-month visual acuity. Both trials are ongoing, with no published results at the time of review.

As described in Sect. "Theranostic-guided corneal UV therapy", emerging iontophoresis-assisted approaches integrated with theranostic technology enable spatially controlled delivery of riboflavin into the stroma, with the potential to improve treatment customization based on corneal geometry and biomechanical characteristics.

Comparative Clinical Outcomes Epi-On vs. Epi-Off CXL Protocols

Epi-on CXL is safer but historically less effective, with higher progression rates (up to 75%) in early studies [4, 176, 181, 209]. However, modern optimized protocols (corneal iontophoresis, enhanced formulations, oxygen, pulsed UV) are reducing this gap [17, 39, 81, 105, 119, 120, 145, 146, 198, 200]. The meta-analysis by D’Oria et al. [40] pooled data from 17 comparative studies, demonstrating that when optimized formulations and enhanced delivery methods are used, the rate of keratoconus progression does not differ significantly between epi-on and epi-off CXL protocols. Visual, topographic, pachymetric, and endothelial parameters were comparable at 12- and 24 months, while the incidence of postoperative corneal haze was significantly lower in epi-on CXL procedures.

Taken together, the current body of evidence supports that optimized transepithelial CXL protocols, particularly those incorporating 0.2% riboflavin formulation, iontophoresis delivery, oxygen supplementation, and pulsed high-fluence UV-A irradiation, can achieve comparable disease stabilization to epi-off CXL protocols in adult patients with progressive keratoconus. However, younger patients (< 18 years) or those with advanced disease stage may still benefit from the epi-off CXL approach, which provides the highest and most reproducible stiffening effect, including long-term stabilization of K_max, until further evidence demonstrates the equivalence of emerging epi-on and data-driven personalized approaches (Table 3).

Table 3.

Summary of safety and efficacy outcomes of clinical studies comparing epithelium-off vs. epithelium-on CXL protocols for the treatment of progressive keratoconus

Parameter	Epi-off CXL (Conventional/Accelerated/Theranostics)	Epi-on CXL (Transepithelial/Iontophoresis/Oxygen/Theranostics)
ΔKmax (mean change)	− 1.5 D (Soeters et al.) at 1 year − 1.0 D (Vinciguerra et al.) at 1 year − 2.1 D (theranostics, Roszkowska et al.) at 1 year − 1.3 D (Çerman et al.) at 1.5 year − 1.5 D (Lombardo et al.) at 2 years − 1.5 D (Rush et al.) at 2 years − 1.7 D (Elving et al.) at 2 years − 2.1 D (Al Fayed et al.) at 3 years	+ 0.3 D (dextran-enriched riboflavin, Soeters et al.) at 1 year − 0.3 D (iontophoresis-assisted, Vinciguerra et al.) at 2 years − 1.3 D (theranostics, Roszkowska et al.) at 1 year − 0.1 D (dextran-enriched riboflavin, Çerman et al.) at 1.5 years − 1.0 D (iontophoresis-assisted, Lombardo et al.) at 2 years − 0.5 D (HMPC-enriched riboflavin, Rush et al.) at 2 years − 1.7 D (oxygen supplementation, Elving et al.) at 2 years  + 1.0 D (dextran-enriched riboflavin, Al Fayed et al.) at 3 years
Progression rate (≥ 1.0 D Kmax)	0% (Soeters et al.) at 1 year 0% (theranostics, Roszkowska et al.) at 1 year 9% (Lombardo et al.) at 2 years 0% (Al Fayed et al.) at 3 years	26% (dextran-enriched riboflavin, Soeters et al.) at 1 year 0% (theranostics, Roszkowska et al.) at 1 year 10% (iontophoresis-assisted, Lombardo et al.) at 2 years; 55% (dextran-enriched riboflavin, Al Fayed et al.) at 3 years
ΔCDVA (mean change)	Stable or improved (0.0/-0.1 LogMAR)	Stable or improved (0.0/-0.2 LogMAR)
Demarcation line depth	267–320 µm	125–339 µm
Corneal haze incidence	Transient corneal haze ≥ 30% Persistent corneal haze ≤ 3%	Lower incidence than epi-off CXL Persistent corneal haze not reported
Severe adverse events incidence	Cumulative ≤ 1.3%: infectious keratitis, scarring, endothelial decompensation	Cumulative ≤ 0.1%: none with permanent sequelae
Endothelial safety	“Minimal endothelial cell loss”, described as not clinically significant	“No significant ECD change”
Pain/recovery	“Post-op pain and delayed re-epithelialization (> 5 days)”	“Minimal discomfort”

Overall, treatment choice should be individualized based on age, disease severity, and corneal thickness, as technological advances progressively narrow the clinical efficacy gap between epi-on and epi-off CXL.

Corneal Cross-Linking in Pediatrics

Keratoconus in the pediatric population poses a unique therapeutic challenge due to its rapid progression and higher risk of vision loss compared to adults. Numerous clinical studies have confirmed the safety and efficacy of CXL in pediatric patients, consistently demonstrating its ability to halt disease progression with a low incidence of adverse events, comparable to those observed in adult cohorts. Standard CXL proved to be a safe treatment for progressive KC in pediatric patients with an anterior curvature progression rate of up to 8% at ≥ 10 years of follow-up [195]. Accelerated CXL was also an effective and safe long-term strategy to prevent progression of KC in pediatric patients, with a 10-year follow-up re-progression rate of 9.14% [11]. A 24% regression rate could be contemplated in the pediatric patients who were aged 15 years and younger at the time of inclusion in the treatment protocol [133].

Collectively, these findings provided robust support for the clinical efficacy and safety of CXL in children (< 12 years old) and adolescents (< 18 years old), with treatment performed at a median age of 15 years [27, 45, 95, 133, 150, 151, 167]. Across meta-analysis studies and several long-term series, the mean change in K_max was reported as − 0.8 ± 0.4 D, which is almost equivalent to adult data outcome, while endothelial cell density (ECD) did not significantly change (− 39 ± 51 cells/mm²) over a mean follow-up of two years, and corrected distance visual acuity (CDVA) showed to stabilize, with a mean change of − 0.04 ± 0.05 LogMAR in the same period. These results were notably in contrast to untreated controls, where K_max and visual acuity worsened in the same period. Specifically, Padmanabhan et al. [151] reported outcomes in 336 patients followed for up to seven years after epi-off CXL, with a mean K_max decrease of − 1.20 ± 3.55 D and stable CDVA. Only 24% experienced progression, compared with an estimated 88% progression risk in untreated pediatric patients [27]. Or et al. [150] corroborated these findings in a 5-year retrospective analysis of 88 eyes, showing disease stabilization in treated eyes compared to continued progression in untreated fellow eyes.

Meta-analyses have also clarified protocol-specific outcomes [2, 7, 78, 106, 169]. Several authors showed that standard (3 mW/cm²) and accelerated (10 mW/cm²) epithelium-off CXL protocols were equally effective up to four years postoperatively, confirming both as viable therapeutic options [27, 78, 106, 150, 169]. The authors further demonstrated the feasibility of accelerated CXL in children, highlighting its shorter treatment time and comparable efficacy, with stable uncorrected distance visual acuity (UDVA) and CDVA maintained for up to 5 years.

Taken together, the accumulated evidence strongly supports the use of CXL in pediatric keratoconus. Across multiple protocols, CXL has demonstrated durable stabilization of corneal biomechanics, preservation of visual acuity, and protection against the high risk of degenerative corneal progression and need for corneal transplantation [95]. Severe adverse events are rare, while moderate complications are generally non-vision-threatening and self-limiting. These findings reinforce CXL as the treatment of choice for halting keratoconus progression in children and adolescents. In light of natural history studies indicating rapid pediatric progression and the high benefit/risk profile [52, 57], authors recommended consideration of CXL in high-risk patients (positive family history, allergy, atopy, skin diseases, eye rubbing habit) even before documented progression (i.e., at first detection) [59, 63, 113].

Safety of Corneal Cross-Linking

Literature data shows that in most cases CXL treatment-related adverse events are mild and transient, with corneal haze being the most common. Severe complications following CXL are rare [37, 97, 174, 188]. It should be noted that complications were not consistently or systematically reported across clinical studies; therefore, the true incidence of adverse events may be higher than reported. Below is a summary by type and incidence (%) of adverse events after CXL:

• Severe adverse events (cumulative incidence: ≤ 1%).

These include bacterial keratitis, herpetic keratitis, corneal melting, corneal scar, phototoxic endotheliitis, corneal burn, which may result in severe vision loss.

• Moderate adverse events (cumulative incidence: ≤1%).

These include persistent corneal haze, uveitis, recurrent or persistent epithelial defects, sterile corneal infiltrates, transient epithelial erosions, corneal hydrops, and limbal neovascularization, and epithelial islands in LASIK flap (iatrogenic corneal ectasia).

• Mild adverse events: (cumulative incidence: 28%).

These include transient eye pain, transient ocular pruritus, transient corneal haze, transient corneal edema, punctata keratopathy, transient visual clouding, dry eye sensation, conjunctival hyperemia. In more than 90% of cases, these conditions are resolved within 6 months after surgery. Notably, all adverse events with permanent sequelae were recorded only after the epi-off CXL protocols and were never reported after transepithelial CXL protocols. Epithelial removal is, therefore, the main risk factor for these adverse events in the corneal cross-linking procedure.

Severe adverse events in young patients were rare, with only a few reports of postoperative infections and corneal hydrops; moderate adverse events were infrequent (≤ 3%) and mostly included sterile corneal infiltrates and delayed epithelial healing. Finally, persistent corneal haze has been reported among the most relevant long-term adverse events, although the reported incidence varies considerably across studies [78, 106, 174].

Combined CXL Protocols

Techniques like transepithelial photorefractive keratectomy and phototherapeutic keratectomy when combined with CXL, are grouped under the so-called “CXL plus” procedures. Their aim is to improve corneal regularity, visual acuity, and contact lens tolerance. Although corneal ablation in keratoconus was once viewed with caution due to biomechanical concerns, advances in imaging and customized ablation profiles have made it a viable option in carefully selected cases, enabling limited tissue removal while preserving stability.

Corneal cross-linking can also be combined with intracorneal ring segment implantation to both halt disease progression and improve corneal shape, particularly in more advanced stages [42].

Overall, these combined strategies reflect a shift toward tailored interventions that optimize both biomechanical stabilization and visual outcomes in keratoconus management.

CXL Plus

The first report of combined CXL with sequential topography-guided PRK (photorefractive keratectomy) appeared in 2007 with the Athens protocol, later expanded by the Cretan protocol [86, 99]. Simultaneous approaches avoid ablating already cross-linked tissue, while sequential treatments (PRK performed 4–6 months after epi-off CXL) have also shown favorable visual and refractive outcomes [87, 100]. Overall, excimer laser ablation after CXL is a safe and effective option to partially correct sphero-cylindrical error and reduce higher-order aberrations in mild-to-moderate keratoconus, with outcomes largely independent of age, gender, or timing.

The simultaneous method offers a distinct advantage by preventing interference with cross-linked corneal tissue during laser ablation. However, retrospective case series examining corneal wavefront-guided trans-PRK performed 4- to 6 months after epi-off CXL also demonstrated promising visual and refractive outcomes in these patients. This paves the way for therapies wherein patients who have undergone prior CXL could benefit from topography-guided or wavefront-guided excimer laser ablation to improve their visual acuity.

Across CXL Plus protocols, postoperative corneal stability is generally comparable, though treatment parameters must be tailored to corneal thickness and cone location [42]. Common eligibility criteria across the different protocols include patient- and tissue-related parameters, as follows:

• targeted keratoconus stage: stage II (range stages I–II) according to the Krumeich classification;

• average baseline age: 18.5 years old (range 14–30 years);

• average baseline stromal thinnest point: 435 μm (range 300–450 μm);

• average stromal ablation: 50 μm (range 40–55 μm);

• average optical zone diameter 5.5–6.5 mm.

The use of mitomycin C following CXL in simultaneous topography-guided t-PTK or t-PRK has been associated with an increased risk of corneal haze and is therefore not recommended.

More recently, the Central Corneal Remodeling (CCR) protocol has introduced customized transepithelial ablation combined with accelerated CXL to achieve smoother, more regular corneal geometry through large, gradual transition zones guided by ray-tracing-based planning [6, 137]. Ray-tracing guided transepithelial PRK plus pachymetry-guided ACXL significantly improved the quality of vision in patients with KC, preventing overcorrection and minimizing tissue consumption The mean UDVA improved by 3.5 ± 1.3 Snellen lines, with 38% of eyes that gained ≥ 4 ± 1.3 lines. The mean CDVA improved by + 4.3 ± 1.3 lines with 68% of eyes that gained ≥ 4 ± 0.9 lines. No eye lost one or more lines of CDVA.

A recent study including 19 patients (34 eyes) with keratoconus treated by total wavefront-guided transepithelial PRK with maximum planned stromal ablation of 55 µm followed by accelerated CXL (30 mW/cm² for 3 min; total fluence: 5.4 J/cm²) showed significant UDVA improvement and disease stabilization [68]. Overall, these findings support the concept that moderate tissue removal may contribute to clinically significant corneal shape remodeling and visual improvement beyond conventional CXL approaches.

In summary, CXL Plus procedures are safe and effective for improving visual function, particularly in patients with contact lens intolerance or poor spectacle-corrected vision. However, careful patient selection and surgical expertise remain essential, as combined approaches may carry higher risks of delayed epithelial healing, corneal haze, and visual loss compared with CXL or PRK alone [149].

CXL Combined with Intracorneal Ring Segments

Intracorneal ring segments (ICRS), introduced in the 1980s for myopia correction, act by flattening the anterior cornea. Their use has since expanded to keratoconus, where they improve corneal symmetry, reduce myopic astigmatism, and may delay the need for corneal transplantation in selected cases [153].

Candidates are typically adults with contact lens intolerance, CDVA between 0.4 and 0.8 LogMAR, corneal thickness > 480 µm at the implantation site, alignment between refractive and keratometric axes, and no central scarring [166]. Femtosecond laser-assisted implantation has improved precision and safety compared with manual techniques; after creating an entry incision and stromal channels at 70–80% of corneal thickness, the ring segments are carefully inserted into these tunnels using specialized forceps.

Despite their proven efficacy, synthetic ICRS procedures have been associated with severe complications, such as segment migration, stromal thinning leading to melt or necrosis, segment extrusion, corneal neovascularization, and infectious keratitis [34, 43, 44, 207]. In CXL combined with ICRS implantation, the reported rate of corneal perforation caused by the ICRS is 3.6%, while ICRS extrusion or explantation occurs in 4.6% of cases. To address these risks, corneal allogenic intrastromal ring segments (CAIRS) have emerged as a promising alternative [80]. Using donor corneal tissue shaped with a femtosecond laser, CAIRS offer improved biocompatibility and flexibility, with the potential for a better safety profile while maintaining comparable refractive outcomes. They can be implanted either simultaneously with or after cross-linking. The surgical technique closely resembles that of ICRS implantation using a femtosecond laser; it involves the placement of customized, semicircular segments of human eye-bank donor corneal tissue into preformed intrastromal channels, with some recent modifications to facilitate implantation. As for ICRS, depending on the individual requirements of the patient, either single or double CAIRS segments can be employed [104]. However, unlike synthetic rings, CAIRS segments can be fashioned with a femtosecond laser to any desired arc length or positioned circumferentially within intrastromal channels. Overall, by integrating the morphological and biomechanical advantages of synthetic ICRS, CAIRS may provide a more favorable benefit–risk profile, reducing the incidence of adverse events while maintaining comparable refractive and corneal shape improvements [10, 104, 153, 190].

The crosslinking of CAIRS expands its potential indications by enhancing the mechanical rigidity of the segments, thus facilitating surgical manipulation and implantation. Moreover, crosslinking promotes decellularization, minimizing rejection, corneal melt, counteracting long-term volume loss and contributing to improved structural stability and preservation of the intended therapeutic effects over time [64, 136]. Nevertheless, CAIRS implantation remains an evolving technique that requires long-term evaluation to assess tissue integration, the durability of the refractive outcomes, and the risk of late-onset complications [144].

In summary, both ICRS and CAIRS, particularly when combined with CXL, provide valuable options for visual rehabilitation in keratoconus, especially in patients with poor spectacle-corrected vision or contact lens intolerance, though careful selection and long-term evaluation remain essential [74, 80].

Two-Step Treatment with ICRS + CXL and PRK

A few studies have assessed a combined staged regimen of ICRS implantation followed by CXL and PRK [5, 126]. This approach includes variations such as ICRS with either simultaneous or delayed CXL, followed by PRK (with concurrent CXL if not performed earlier) as a second-stage procedure, typically scheduled at least six months after the first stage to allow refractive stability. The authors have reported improvements in both UDVA and CDVA as well as in corneal topographic indices in patients with keratoconus. Regarding safety and adverse events, the literature indicates that while this combined staged approach is generally safe, it is not without risks [5, 126]. Potential complications include segment migration and extrusion, delayed epithelial healing, corneal haze, inadequate vision recovery, and progression of existing keratoconus. Therefore, careful patient selection and adherence to surgical protocols are essential to minimize these risks and optimize outcomes.

Emerging Indications of Use

CXL in Infectious Keratitis

Keratitis is an inflammation of the corneal tissue with multiple etiopathogenesis. The causes can be infectious, immune, toxic, or traumatic. Due to the increase of antibiotic resistance in the last decade, the medical community is looking for new methods of treatment of ophthalmic diseases with infectious etiopathogenesis. The CXL procedure is among the treatment options that has proven to be safe and effective as an adjuvant treatment for minimizing antibiotic use.

The photo-oxidative reaction induced by riboflavin and UV-A light exerts antimicrobial activity through ROS-mediated lipid peroxidation of microbial cell membranes and enhances stromal resistance to enzymatic degradation, thereby helping to limit infectious progression [56, 91, 115].

Historically, the antimicrobial properties of the riboflavin/UV-A light photo-oxidative reaction were first demonstrated in laboratory settings as well as in veterinary clinical applications (VET-CXL) using high power density (30 mW/cm²) after earlier attempts with conventional 3 mW/cm² irradiation proved ineffective [47, 128]. These findings paved the way for translation into human clinical practice [160, 187]. Nevertheless, evidence for CXL as an adjunct in treatment of bacterial keratitis remains limited, although several scattered open studies and a few RCTs have suggested that combining CXL with standard antibiotic therapy can accelerate epithelial healing and reduce the risk of corneal perforation [13, 35, 127]. In a randomized clinical trial (CLAIR), adjuvant CXL did not significantly improve 3-month visual outcomes but reduced culture positivity and complication rates, suggesting benefit in selected cases [155].

Fungal keratitis poses a greater challenge due to rising antifungal resistance; CXL has been explored as an adjunct in these cases; early observational and randomized studies show mixed results, with pilot studies reporting reduced ulcer size, shorter healing time, and decreased medication frequency when CXL is combined with antifungal therapy, although the procedure was more effective in superficial fungal keratitis rather than deep infections [193, 199].

In the treatment of infectious keratitis, severe adverse events after CXL were reported in about 2% of cases, including corneal melting (0.3%) and corneal perforation (1.6%). These events can often be natural consequences of the underlying microbial infection and may therefore also reflect a lack of therapeutic efficacy rather than a direct complication of the treatment. Overall, clinical evidence for CXL in infectious keratitis is heterogeneous and generally low to moderate quality; differences in study design, treatment protocol, pathogen type, depth of corneal ulceration and outcome measures complicate study comparisons [26, 189]. Well-powered multicenter randomized trials are needed to define optimal CXL protocols for treating keratitis and infectious corneal ulcers, including intraoperative measure of corneal riboflavin (a core element of theranostic-guided approaches) and subgroup analyses based on microbial etiology, potentially identifying patients most likely to benefit and improving treatment personalization.

CXL is a promising adjunctive therapy for infectious keratitis, especially in bacterial and superficial fungal ulcers. High-irradiance protocols have been confirmed to be more effective than low-intensity approaches, with potential to reduce healing time, minimize medication load, and prevent complications. Systematic reviews and meta-analyses have indicated that adjuvant CXL, when combined with standard antimicrobial therapy, may provide superior outcomes in certain cases, particularly in fungal keratitis, by promoting faster corneal healing [48].

CXL in Bullous Keratopathy

Bullous keratopathy (BK) is a corneal disease with primary (Fuchs’ endothelial dystrophy, FED) or secondary (iatrogenic) etiopathogenesis. It consists of a progressive imbibition of the stroma with progressive reduction of corneal transparency and formation of epithelial bullae, causing light sensitivity, eye pain and vision loss. The therapy is medical in early disease stage through the application of hypertonic and/or anti-inflammatory eye drops and surgical (DSAEK, DMEK) in the most advanced cases.

The application of CXL in treatment of BK has been investigated for more than ten years as adjunctive therapy aimed at alleviating symptoms and improving corneal integrity in eyes with advanced endothelial dysfunction. Histopathological analyses demonstrated that CXL can alter stromal swelling behavior by compacting the collagen matrix, reducing interfibrillar spacing and stromal water content and lowering stromal swelling pressure, which in turn improves epithelial adhesion and corneal transparency [9, 19]. By stabilizing the anterior stroma and reducing its permeability to fluid, it was hypothesized that CXL could reduce corneal edema, alleviate pain associated with epithelial bullae, and possibly restore corneal optical quality. Overall, although the effect is often temporary, CXL may represent a palliative option in patients who are unsuitable for keratoplasty, with the advantage of a relatively favorable safety profile and absence of severe adverse events.

A few RCTs and retrospective studies assessed the efficacy of CXL in pseudophakic BK and corneal edema secondary to FED. Despite the small sample sizes in individual studies, the available reports consistently indicated that pain reduction represented the most reproducible outcome, with relief in some cases lasting for several months [28, 178]. Corneal thickness often showed a transient decrease after treatment, but this effect was not maintained over time, while central corneal clarity generally remained unchanged. Visual acuity improved in more than two-thirds of treated eyes; however, such improvement was not durable, even in cases where pachymetric stabilization and symptomatic relief were consistently observed. No severe or moderate adverse events were reported, and the most common mild adverse event was corneal erosion, with an incidence of about 3% [9, 19, 28, 178]. Systematic reviews have consolidated these results [180]. The authors concluded that CXL in BK may provide temporary symptomatic relief and transient pachymetric reduction, but the effect is not sustained and does not translate into clinically significant functional visual rehabilitation. Pain relief, when present, represents the main clinical benefit.

In summary, more than a decade of clinical work has clarified the potential role of CXL in the management of bullous keratopathy. While the initial rationale was biologically sound and early reports raised hopes for a non-surgical solution, controlled clinical trials and systematic reviews have shown that the benefits are transient, case-dependent, and primarily limited to symptomatic relief. Visual rehabilitation and restoration of corneal transparency remain out of reach in advanced endothelial disease, and endothelial corneal transplantation continues to be the only definitive treatment for these conditions.

Crosslinking in Dry-Eye

CXL, traditionally regarded as a purely biomechanical intervention for halting ectatic progression in keratoconus, is increasingly emerging as a biologically active procedure capable of modulating ocular surface inflammation and tear film homeostasis. This concept is particularly relevant in patients with keratoconus and concomitant allergic ocular surface disorders and dry eye disease (DED). In these patients, ocular surface inflammation, eye rubbing, and tear film instability establish a self-perpetuating pathogenic loop that may contribute not only to symptom burden but also to ectatic progression [113]. Recent clinical evidence suggests that epi-off A-CXL may exert beneficial effects in this population of patients extending beyond stromal stiffening. In a recent study, 15 pediatric patients (30 eyes) with progressive keratoconus and concomitant allergic ocular surface disease and dry eye disease (DED), characterized by elevated tear matrix metalloproteinase-9 (MMP-9) levels, were treated with epithelium-off A-CXL. At 1 year, in addition to disease stabilization, clinical efficacy was associated with a significant reduction in tear MMP-9 levels and qualitative resolution of inflammatory infiltrates on in vivo confocal microscopy [129]. These findings support the hypothesis that CXL may exert additional biological effects beyond biomechanical reinforcement. Reactive oxygen species generated during the procedure may modulate inflammatory signaling and epithelial-stromal remodeling processes, potentially interrupting the itch-rub-inflammation cycle frequently observed in allergic and pediatric phenotypes [113]. These preliminary observations support the concept that CXL may exert effects beyond corneal stiffening, potentially influencing ocular surface homeostasis. This raises the possibility of biomarker-guided applications and warrants further investigation of CXL-based strategies in selected inflammatory dry eye conditions beyond keratoconus [129, 139, 165].

Emerging Corneal Cross-Linking Technologies and Approaches

In recent years, a few novel technologies have emerged to advance CXL beyond conventional riboflavin/UV-A light protocols. Among these, ruthenium-induced CXL and rose bengal-mediated CXL using visible light have been tested in laboratory. The former approach used a water-soluble ruthenium photo-initiator activated by blue light (430 nm) to achieve effective corneal stiffening [62]. In bovine corneas treated with this ruthenium-based method, mechanical testing revealed a significant increase in corneal stiffness and no significant cellular damage or apoptosis in corneal epithelial or limbal stem cells. In the latter approach, rose bengal is illuminated by green light (522 nm) to stiffen the cornea or increase its resistance to enzymatic digestion. Preclinical studies in animal models (e.g., porcine, rabbits, lamb) have shown that the treatment induces a similar stromal effect compared to riboflavin/UV-A light CXL [14]. While promising, the cellular toxicity and potential for corneal pigmentation caused by rose bengal, despite being studied for many years, necessitate thorough evaluation and appropriate regulatory and conformity testing before clinical use in humans [186].

A more recent innovation is theranostic-guided corneal UV therapy, which integrates real-time monitoring of riboflavin corneal diffusion, and its UV-A light-mediated photo-activation, enabling patient-specific, adaptive treatment and more reproducible clinical outcomes, independent of the treatment protocol chosen by surgeon [164]. This technology has been fully validated and CE-marked through preclinical and clinical studies, demonstrating its safety, efficacy, and reliability for routine clinical use.

Finally, IVMED-80 represents a pharmacologic cross-linking approach to treating keratoconus, distinct from riboflavin/UV-A corneal cross-linking, as it is a copper-containing eye drop designed to activate lysyl oxidase and stimulate natural enzymatic cross-links within the corneal stroma [103].

While all these approaches illustrate the breadth of technological advancement in CXL, this review will focus exclusively on emerging technologies supported by clinical data. This ensures that the discussion is grounded in robust, evidence-based practice and that efficacy and safety have been evaluated through established regulatory pathways before any emerging technology is considered for routine clinical use.

Theranostic-Guided Corneal UV Therapy

Theranostics, an approach that integrates real-time molecular imaging diagnostic feedback with targeted therapy, was introduced in the management of keratoconus to address current limitations of the procedure, particularly the variability in clinical efficacy and uncertainties among available protocols, which differ in riboflavin delivery methods and UV-A irradiance, as described in previous sections. By directly measuring the corneal biological response to riboflavin absorption and UV-A exposure, theranostic-guided corneal UV therapy enables treatment personalization, delivering precise riboflavin dosing and UV-A light-mediated photoactivation tailored to each cornea, independent of the surgeon-selected protocol [108–111, 117, 118, 121, 122, 164].

A series of preclinical studies has been pivotal in developing and validating theranostic technology for keratoconus treatment [108–110, 117, 118, 122]. These studies quantified stromal riboflavin concentration during CXL and demonstrated that riboflavin availability and UV-A-mediated photoactivation strongly correlate with biomechanical stiffening, with high predictive accuracy (94%) [79, 108, 109, 117, 184]. Real-time theranostic imaging biomarkers (riboflavin score and theranostic score) enabled reliable prediction of treatment efficacy, confirming both pre-irradiation corneal riboflavin concentration and its photoactivation as key determinants [104, 108, 109, 122]. Importantly, photochemical kinetics varied with stromal concentration and UV-A power, even at constant energy dose. The findings validated theranostics as a robust, protocol-independent tool for personalized CXL and highlighted significant variability in riboflavin penetration, with most epi-on formulations failing to reach effective stromal levels (i.e., riboflavin score was lower than efficacy threshold of 0.4), except for enhanced 0.22% riboflavin formulation [117].

The ARGO trial (NCT05457647) was a multicenter, randomized study validating theranostic-guided corneal cross-linking and the predictive value of riboflavin score and theranostic score for K_max flattening [164]. Fifty patients with progressive keratoconus underwent epi-off or epi-on CXL using a UV-A device with integrated theranostic software, without affecting procedural adherence. All eyes received 0.22% riboflavin followed by UV-A irradiation at 10 mW/cm² for 9 min (energy dose: 5.4 J/cm²). The theranostic system showed high predictive performance (91% accuracy, 95% precision) for 1-year K_max outcomes. Clinically, mean K_max flattening was 1.5 ± 1.6 D with significant visual improvements and no severe or moderate adverse events. Notably, true positives treated with epi-on CXL achieved – 1.3 ± 0.9 D K_max reduction, exceeding values typically reported in the literature and in recent FDA-approved epi-on CXL protocols (– 0.4 D K_max flattening) [4, 22, 23, 29, 39, 40, 105, 163, 164, 176, 181, 198, 200, 209]. In addition, true positives treated with epi-off CXL achieved a – 2.1 ± 2.0 D K_max reduction, outperforming the average outcomes reported for conventional CXL [73, 147, 174, 175, 202].

Recently, the integration of controlled corneal iontophoresis with a theranostic UV-A device has enabled spatially patterned riboflavin delivery, concentrating it in the weakest corneal regions to achieve targeted photo-polymerization. In a preclinical study [122], this platform delivered 0.22% riboflavin through the intact epithelium, generating a reproducible central-to-peripheral gradient (~ 40% higher centrally) by modulating current amplitude (between 0.5 mA and 2 mA), application time (between 4 min and 10 min), and aperture design. Real-time theranostic monitoring confirmed controlled stromal distribution, while ex vivo corneal topography showed consistent corneal flattening.

Preliminary clinical results at 6 months in patients with advanced progressive keratoconus supported these findings, demonstrating significant improvements in visual acuity (– 0.4 LogMAR in UDVA and – 0.3 LogMAR in CDVA, respectively), a hyperopic shift of + 1.2 D, and mean K_max flattening of 1.9 D (Fig. 2) [111].

Fig. 2.

Workflow of theranostic-guided corneal UV photo-reshaping. A Riboflavin delivery is spatially patterned and precisely controlled by the theranostic device using an interconnected corneal iontophoresis system targeting biomechanically weak regions of the cornea under Placido disk guidance. B and C The same theranostic system illuminates the cornea with UV-A light to both measure riboflavin concentration in real time and photo-activate the substance, thereby stabilizing corneal biomechanics and improving optical quality

Collectively, these preclinical and clinical investigations established theranostic-guided corneal UV therapy as a predictive, data-driven strategy that provides real-time intraoperative metrics, enabling a shift from fixed, “one-size-fits-all” CXL to individualized keratoconus treatment [121, 164]. The combination of spatially patterned riboflavin delivery through controlled corneal iontophoresis with theranostic-guided UV-A light irradiation highlight the potential for precision keratoconus treatment tailored to the individual corneal characteristics, while underscoring the need for long-term studies to confirm sustained efficacy.

Chemical Cross-Linkers and Synergistic Enhancement

IVMED-80 is a topical copper-based eye drop developed to increase lysyl oxidase activity. Lysyl oxidase is a critical enzyme that regulates the extracellular matrix and facilitates the conversion of lysine into reactive aldehydes [103]. These reactive aldehydes play a role in creating cross-links between stromal proteins, as described in Sect. "Photochemistry of CXL". A phase 1/2a randomized, double-masked, placebo-controlled trial (NCT05241145) assessed the safety and efficacy of IVMED-80 in 31 patients over a 6-month period. The results indicated a statistically significant reduction in K_max by 1.8 D in the 16-week treatment group compared to placebo (p = 0.019). No treatment-related adverse events were reported. Following these results, IVMED-80 received Orphan Drug Designation from the FDA and the therapy is currently advancing to phase 3 clinical trials, although no clinical study has yet been initiated or registered at the time of review.

Future Directions

CXL is firmly established as the standard treatment for progressive keratoconus and iatrogenic ectasia, with substantial evidence supporting its long-term safety and efficacy. However, interpretation of comparative safety and efficacy outcomes across different CXL protocols remains challenging because of the considerable heterogeneity of the available literature. Variations in riboflavin formulations, delivery methods, UV-A irradiation parameters, oxygen supplementation strategies, follow-up duration, and definitions of disease progression or treatment success may significantly influence reported outcomes. Consequently, caution is warranted when comparing efficacy results across studies, and further standardization of study designs and outcome measures is needed to facilitate meaningful comparisons among treatment approaches.

Despite its clinical success, CXL has historically been performed using largely empirical and protocol-based approaches, with considerable variability in clinical outcomes among patients and treatment protocols. Reported progression or retreatment rates still vary widely across studies, reflecting the complex interplay between riboflavin formulation, stromal penetration, UV-A irradiance, oxygen availability, corneal thickness, disease stage, and intrinsic biological variability of the corneal extracellular matrix. This heterogeneity highlights the need for more standardized, reproducible, and progressively individualized treatment strategies. In this context, a structured, evidence-based framework is required to support clinical decision-making. A decisional flow diagram is provided in Fig. 3 to guide the management of progressive keratoconus using CXL, either as a standalone procedure or in combination with adjunctive treatments. The flow diagram also highlights emerging treatment protocols for other pathological conditions.

Fig. 3.

Decision-making flowchart for the management of progressive keratoconus with single or combined corneal cross-linking (CXL) procedures. The flowchart defines patient selection and eligibility criteria, describes procedural variants and follow-up recommendations, and addresses special populations (i.e., pediatric patients). The bottom of the flowchart presents emerging indications for CXL, supported by evidence identified through the literature search performed for this review

Over the last decade, significant advances have been achieved in both epithelium-off and transepithelial approaches through the optimization of riboflavin formulations, epithelial permeability enhancers, iontophoresis-assisted delivery, pulsed-light irradiation, supplemental oxygen strategies, and high-fluence UV-A protocols. In particular, transepithelial CXL has progressively evolved from an initially less effective alternative toward a safer and increasingly reliable therapeutic option. Although epithelium-off protocols still provide the most reproducible biomechanical effect, the most recent epi-on strategies are progressively narrowing the efficacy gap while preserving the advantages of epithelial integrity [164].

At the same time, theranostic technologies and real-time analytic platforms are introducing a new paradigm in CXL by enabling intraoperative assessment of riboflavin stromal concentration, UV-A photoactivation, and photochemical dynamics. This technology aims to improve treatment reproducibility across surgeons and clinical centers by supporting adaptive modulation of procedural parameters according to individual corneal characteristics. In parallel, integration of tomographic biomarkers, epithelial mapping, inflammatory profiling, and molecular indicators may progressively support biomarker-driven approaches capable of refining patient selection, predicting treatment response, and optimizing timing and treatment settings.

Nonetheless, CXL remains a biologically mediated photochemical process occurring within a structurally heterogeneous corneal tissue. Consequently, part of the biomechanical and clinical response likely retains an intrinsically variable component that may not be entirely predictable even with advanced imaging and analytics. For this reason, future progress will likely depend not on the identification of a single “ideal” protocol, but rather on the development of adaptive and data-driven therapeutic frameworks capable of improving procedural consistency, safety, and personalization across different clinical scenarios. Beyond real-time monitoring and adaptive UV-A delivery, future CXL platforms may evolve toward spatially selective biomechanical modulation of the cornea. Advances in drug-delivery technologies could enable customized distribution of riboflavin according to individual corneal tomography, biomechanical characteristics, and disease pattern. Rather than applying a uniform treatment effect across the entire cornea, next-generation systems may allow selective enhancement of cross-linking in predefined stromal regions, potentially enabling tailored biomechanical remodeling while minimizing unnecessary tissue exposure. Such approaches could expand the role of CXL from disease stabilization toward precision-guided corneal shape modulation and functional rehabilitation. Artificial intelligence is expected to play a central role by analyzing large datasets of procedural parameters and clinical outcomes. These models could predict treatment response, optimize dosing, and reduce inter-center variability, progressively transforming CXL into an adaptive, learning-based therapy [111, 121, 164].

Such strategies may further expand the role of CXL beyond static biomechanical stabilization toward a broader platform for corneal remodeling and visual rehabilitation.

Method of Literature Search

An extensive literature search was conducted to identify relevant studies addressing CXL for keratoconus and related ectatic disorders. The search covered the period from January 2003, corresponding to the introduction of the Dresden protocol, through November 2025.

The primary source was the PubMed/MEDLINE database. Search terms included combinations of the following keywords: corneal cross-linking, corneal collagen cross-linking, riboflavin and UV light, CXL, Dresden protocol, accelerated CXL, transepithelial CXL, iontophoresis-assisted CXL, riboflavin, ultraviolet-A, keratoconus, corneal ectasia, keratitis, bullous keratopathy, oxygen supplementation, pulsed irradiation, biomechanical stiffening, intracorneal ring segments, corneal allogenic intrastromal ring segments, theranostics, chemical corneal cross-linking, PRK, and PTK. Boolean operators (“AND”, “OR”) were used to refine the results and include variations in terminology.

Articles were included if they provided an abstract and sufficient methodological detail to support critical appraisal. Priority was given to randomized clinical trials, prospective or comparative studies, systematic reviews, and meta-analyses, however, the most recent clinical data on emerging approaches were also considered. Relevant experimental and translational studies were also reviewed to provide a mechanistic understanding of the CXL mechanism of actions and ultrastructural effect.

Reference lists of the retrieved articles and key review papers were screened manually to identify additional pertinent publications. The selection process was independently cross-checked by multiple authors to ensure consistency and completeness.

Author Contributions

Cosimo Mazzotta, Marco Lombardo and Francesco D’Oria contributed to the study conception and design. All authors contributed to data collection. Data analysis was performed by Cosimo Mazzotta, Marco Lombardo and Francesco D’Oria. The first draft of the manuscript was written by Marco Lombardo and Francesco D’Oria. Cosimo Mazzotta, Marco Lombardo, Alessandro Avitabile, Giovanni Alessio, and Davide Borroni and Francesco D’Oria commented on first version of the manuscript and read and approved the final manuscript.

Funding

No funding or sponsorship was received for this study or publication of this article.

Data Availability

Data sharing is not applicable to this article as no datasets were generated or analyzed during the current study.

Declarations

Conflict of Interest

Cosimo Mazzotta is medical advisory board member of Emagine (Swiss); Marco Lombardo is co-inventor on issued patents (WO2017/130043, WO2020/261163 and WO2021009623A1) related to theranostics technology for treatment of corneal disorders and minor shareholder of Regensight (Italy); Giovanni Alessio, Alessandro Avitabile, Davide Borroni and Francesco D’Oria have nothing to disclose.

Ethical Approval

This article is based on previously conducted studies and does not contain any new studies with human participants or animals performed by any of the authors.