Crosslinking with UV-A and riboflavin in progressive keratoconus: From laboratory to clinical practice – Developments over 25 years

Changes in the biomechanical and biochemical properties of the human cornea play an important role in the pathogenesis of ectatic diseases. A number of conditions in primarily acquired (keratoconus or pellucid marginal degeneration) or secondarily induced (iatrogenic keratectasia after refractive laser surgeries) ectatic disorders lead to decreased biomechanical stability. Corneal collagen cross-linking (CXL) represents a technique to slow or even halt the progression of ectatic pathologies. In this procedure, riboflavin is applied in combination with ultraviolet A radiation. This interaction induces the production of reactive oxygen species, which leads to the formation of additional covalent bonds between collagen molecules and subsequent biomechanical corneal strengthening. This procedure is so far the only method that partially interferes etiopathogenetically in the treatment of ectatic diseases that slows or stops the process of corneal destabilization, otherwise leading to the need for corneal transplantation. Besides, CXL process increases markedly resistance of collagenous matrix against digesting enzymes supporting its use in the treatment of corneal ulcers. Since the discovery of this therapeutic procedure and the first laboratory experiments, which confirmed the validity of this method, and the first clinical studies that proved the effectiveness and safety of the technique, it has been spread and adopted worldwide, even with further modifications. Making use of the Bunsen-Roscoe photochemical law it was possible to shorten the duration of this procedure in accelerated CXL and thus improve the clinical workflow and patient compliance while maintaining the efficacy and safety of the procedure. The indication spectrum of CXL can be further expanded by combining it with other vision-enhancing procedures such as individualized topographically-guided excimer ablation. Complementing both techniques will allow a patient with a biome-chanically stable cornea to regularize it and improve visual acuity without the need for tissue transplantation, leading to a long-term improvement in quality of life.


Introduction
Keratoconus (KC) is a progressive ectatic, clinically noninflammatory corneal disease involving conical bulging and thinning of the corneal stroma due to a biomechanical weakening of the tissue.This leads to an irregular astigmatism and results in a loss of vision.The clinical manifestation is often found in younger patients, where the disease has a dramatic impact on their quality of life (Kymes et al., 2008;Maeno et al., 2000).The available treatment options failed to halt the rapid progression of keratoconus until the 2000s.Due to its nature as a progressive ectatic disease, KC was one of the most common reasons in the indications for corneal transplantation within the treatment options for this type of affliction, which is due to its nature as a progressive ectatic disease (Cunningham et al., 2012;Maeno et al., 2000;Ting et al., 2012).The exact cause of the disease, which leads to biomechanical changes of the corneal tissue, is unknown.The biomechanical properties of the cornea are determined by the structure of collagen, the composition of proteoglycans and their binding to collagen fibers.The three-dimensional arrangement of collagen lamellae is one of the important factors that determine corneal durability (Winkler et al., 2011).Biochemical and immunohistochemical studies of extracellular proteoglycan matrix demonstrated differences between normal corneas and those affected by KC (Rehany et al., 1982).Keratoconic corneas have also been shown to change at the enzyme level, including an increased expression of lysosomal and proteolytic enzymes and decreased levels of protease inhibitors (Kao et al., 1982;Sawaguchi et al., 1989;Zhou et al., 1998), decreased number and thickness (Rehany et al., 1982), and changes in the arrangement of the collagen lamellae of the stroma (Daxer and Fratzl, 1997;Radner et al., 1998).
Although KC is classically defined as a non-inflammatory condition, enzymes, cytokines and free radicals play an important role in the pathogenesis of the disease.Enzymes of the matrix metalloproteinase (MMP) family mediate the degradation of extracellular matrix proteins in response to stress or injury.They are upregulated during matrix remodeling, which is a characteristic feature of KC.MMP-9 is the major matrix-degrading enzyme produced by the human corneal epithelium.Its activity is regulated by a number of inflammatory cytokines, and these proteases are inhibited by tissue inhibitors of MMP enzymes.An imbalance between protease enzymes and protease inhibitors leads to thinning of the stroma in KC (Mazzotta et al., 2018d;McMonnies, 2015).The total collagen content of the afflicted cornea does not differ significantly from that of the normal cornea, however the corneal strength is reduced by an average factor of 0.7 (Andreassen et al., 1980) In addition, twice as much hydroxyproline, which significantly stabilizes the collagen triple helix, can be extracted from keratoconus-affected corneas with pepsin than from normal corneas (Andreassen et al., 1980).These two factors suggest that crosslinking within the collagen (at the level of the tertiary and quaternary structure) and/or between individual collagen molecules must be impaired.Crosslinking disorder has, among others, been discussed as one of the causes of KC, but no experiments have been performed to correct this pathological condition (Cannon and Foster, 1978).
In the natural sciences, crosslinking describes the formation of chemical bridges as a result of chemical reactions between proteins or other molecules.Cross-links can be formed by chemical reactions that are initiated by chemicals, heat, pressure, or radiation.These reactions result in a change in the physical properties of the crosslinked material.During the natural ageing process, both enzymatic and non-enzymatic cross-linking reactions take place in different parts of the human body.Among the observed phenomena, the observation of which led to the development of the concept of crosslinking for the treatment of ectatic corneal disease, is that corneal ectasia does not progress in patients with diabetes mellitus due to naturally occurring non-enzymatic crosslinking of the tissue (Bailey et al., 1998;Malik et al., 1992;Spoerl et al., 2004c).
Corneal crosslinking (CXL) with riboflavin and ultraviolet light (type A, UVA) is a new treatment option for progressive corneal ectatic diseases.It creates new covalent bonds (cross-links) between collagen fibers aiming to increase the strength of the cornea in order to restore its sufficient mechanical stability.The history of this treatment dates back to 1996, when it was first presented at the annual congress of the Association for Research in Vision and Ophthalmology (ARVO, Fort Lauderdale, USA), by a scientific team from Dresden, Germany, led by Professor Theo Seiler (Seiler et al., 1996).That presentation was followed by a number of publications by the same research group analyzing the first results of animal experiments (Spoerl et al., 1998;Spoerl and Seiler, 1999;Sporl et al., 1997Sporl et al., , 2000)).A revolution in the treatment of ectatic corneal disorder, namely keratoconus, was sparked in 2003 when Wollensak et al. from the Dresden group published the first clinical results (Wollensak et al., 2003a).The method was initially not entirely well accepted by the general ophthalmological community, but after the publication of several clinical studies with very good results, including minimal complications and side effects in reputable ophthalmology journals (Caporossi et al., 2010;Raiskup-Wolf et al., 2008;Vinciguerra et al., 2009;Wittig-Silva et al., 2008), it became globally accepted as the treatment method of choice for progressive ectatic corneal disease.Later, several clinical trials were also initiated as part of the Food and Drug Administration (FDA) approval process in the USA (Greenstein et al., 2010;Hersh et al., 2011), in which they confirmed the efficacy of this treatment with an excellent safety profile leading to the FDA approval in April 2016 (Belin et al., 2018) for keratoconus (Hersh et al., 2017b) and ectasia after refractive surgery (Hersh et al., 2017a).
With the introduction of new technologies in ophthalmology, the understanding of the in vivo changes induced by CXL at the microscopic, macroscopic and geometric levels of the cornea has become clearer.These technologies were introduced for example in vivo confocal microscopy (Mazzotta et al., 2015), Scheimpflug imaging (Koller et al., 2009a), and air-puff induced dynamic Scheimpflug imaging (Vinciguerra et al., 2017).
From an economic point of view, Godefrooij et al. stated in their costeffectiveness analysis of the CXL procedure that it was of lower financial cost to the national healthcare system when compared to corneal transplantation.They further hypothesized that if the CXL procedure were to provide stability over the course of 15 years, the treatment would become even more cost-effective (Godefrooij et al., 2017c).Recently, our group showed the effectiveness of the CXL treatment concerning long-term stability of corneal morphology without major complications or the need for subsequent re-interventions with a follow-up of 15 years, confirming the cost-effectiveness theory by Godefrooij et al. (Raiskup et al., 2023).

Mechanical stability of collagen-containing tissues
The mechanical stability of the cornea is primarily determined by the structure of the collagen molecules and their spatial arrangement, i.e. long collagen fibrils that branch out and simultaneously stretch from limbus to limbus.The resulting junctions (cross-links) stabilize the mechanical state and protect the collagen fibrils in the curved cornea from disintegration.Undiscovered pathological tissue changes occur as a result of an increase (e.g. in diabetes mellitus or scarring) or decrease (e. g. in Ehlers-Danlos syndrome) in the degree of crosslinking (Boote et al., 2011).
For this reason, the maintenance of the physiological function of the degree of crosslinking must be very well regulated.With increasing age, the number of cross-links and thus the rigidity of structures increases (Elsheikh et al., 2007;Knox Cartwright et al., 2011).This can be observed in the cornea, skin, lens of the eye, blood vessels, and articular cartilage.Analogous changes are caused by sunlight and smoking (Kennedy et al., 2003;Madhukumar and Vijayammal, 1997).

Possibilities to increase the degree of cross-links
Under physiological conditions, collagen molecules are enzymatically crosslinked in the extracellular space after leaving the cell (i.e.post-translationally) by the enzyme lysyl oxidase (Kagan and Trackman, 1991).In this way, collagen acquires its natural strength and stability F. Raiskup et al. and tissue-specific elastic properties.Lysyl oxidase transforms groups of certain amino acids into aldehyde groups, which can either spontaneously react with adjacent aldehyde groups in a reaction called aldol condensation or with ε-amino acid groups to form aldimine and covalent crosslinking.In the case of Ehlers-Danlos syndrome, lysyl oxidase deficiency is present; in the case of keratoconus, it is thought that a defect in the gene for lysyl oxidase is present (Li et al., 2006), or that lysyl oxidase activity is impaired by the elevated tear fluid pH observed in keratoconus (Avetisov et al., 2011), and conversely, in the case of keloids and scars, the activity of the enzyme is increased (Uzawa et al., 1998).
In addition, non-enzymatic crosslinking can occur, when it is caused by chemical crosslinking agents, such as glutaraldehyde, formaldehyde, diphenylphosphoryl, or genipin, or by sugar aldehydes (advanced glycation end products (AGEs).Chemical agents are preferentially used to modify the properties of collagen-containing tissues in tissue engineering processes (Spoerl and Seiler, 1999;Spoerl et al., 2004b).On the other hand, with increasing age and especially in the case of diabetes mellitus, the amount of these carbohydrate conjugates increases.AGE junctions have a protective effect against the development of keratoconus (preventing or reducing the severity of keratoconus) (Kuo et al., 2006;Seiler et al., 2000).The very first investigations of the group from Dresden by Spoerl and Seiler in this field showed the effectiveness of photo-oxidative crosslinking using ultraviolet-A light and riboflavin as photosensitizer, which was furthermore expected to be clinically relevant (Spoerl and Seiler, 1999).

Photo-oxidative crosslinking with riboflavin and UVA radiation
The aim of corneal crosslinking is to restore the reduced tensile strength of the cornea in keratoconus as well as to increase the resistance against degrading enzymes in order to stop further progression of the disease and thus, a deterioration of visual acuity.The cornea is freely accessible, so crosslinking via UV light radiation can be performed locally and in a precisely limited area, this means in turn, that the subsequent photochemical reaction takes place only where the light is absorbed.
The photo-oxidative crosslinking method using riboflavin and UVA radiation was chosen to strengthen the cornea because of its localized effect; a short treatment period is sufficient to achieve the desired effect and the use of the treatment does not alter corneal transparency (Spoerl and Seiler, 1999;Wollensak et al., 2003b).Riboflavin is a vitamin (vitamin B2) that is also used for food coloring (e.g. in vanilla pudding), it is non-toxic, and it is also available as a medical preparation.The effect of increasing the corneal stiffness can be achieved using the photo-oxidative crosslinking method and the following components: riboflavin, UV light, oxygen, and saturation of the crosslinking effect (Fig. 1).

Riboflavin solutions
A water-soluble photosensitizer that was non-toxic and had a low molecular weight (400 Da) to diffuse sufficiently into the cornea was chosen for crosslinking.Riboflavin is poorly water-soluble, but riboflavin phosphate (vitamin B2) meets these requirements.However, due to its hydrophilic nature and negative charges, riboflavin phosphate has difficulty crossing tight junctions and the lipid layer of the epithelium.Therefore, the epithelium had to be removed.Riboflavin has an absorption peak at both 365 nm and 455 nm wavelengths.For further use on the cornea, UV light of 365 nm was used to avoid toxic blue light.The UVA light can be generated very well with appropriate light emitting diodes (LED, Fig. 2).The concentration of 0.1 % riboflavin solution was chosen in such a way that a cornea of 400 μm thickness impregnated with riboflavin transmits almost no UVA light and the UV exposure for the endothelium is below the toxic threshold (Wollensak et al., 2003d).Meanwhile, the originally proposed riboflavin solution ((0.1 % riboflavin: 10 mg riboflavin-5-phosphate in a 10 ml solution including 20% dextran T-500 (Sporl et al., 2000;Wollensak et al., 2003a),) has been further developed, due to the availability of different recent CXL protocols.
The isoosmolar riboflavin solution prevents swelling of the stroma, which is a combination of 0.1% riboflavin phosphate solution and a 20% dextran solution.However, in clinical application, it was found that this caused excessive stromal de-swelling and thinning to below 400 μm after 30 min of dripping (Kymionis et al., 2009b).To circumvent this drawback, dextran was then replaced by 1.1% hydroxymethyl cellulose (HPMC).A mixture of riboflavin and HPMC even achieved faster diffusion into the stroma than is the case with dextran (Ehmke et al., 2016).Therefore, the application of riboflavin solution with HPMC can be shortened to 10 min without an effect of lower concentration (Ehmke et al., 2016) or higher risk of endothelial damage (Marcovich et al., 2020) The hypoosmolar riboflavin solution was developed for treatment of thin corneas below the threshold of 400 μm after epithelium removal.
This solution does not contain dextran or HPMC, however, it includes 0.5% sodium chloride (NaCl).This allows a thin cornea to swell up to a thickness of >400 μm prior to irradiation (Hafezi et al., 2009;Raiskup and Spoerl, 2011).Experimental investigations have shown that the biomechanical effect between isoosmolar and hypoosmolar are similar due to CXL that acts in the anterior part of the cornea and is not affected by the swelling (Wollensak and Sporl, 2019).Another riboflavin solution is used for transepithelial CXL (TE-CXL) treatments.The reason for the specific development of such a solution is F. Raiskup et al. that the hydrophilic riboflavin phosphate insufficiently passes through the lipophilic epithelium which in turn prevents passage through the paracellular pathway via intercellular tight junctions.In addition, there is a repulsive force between the negatively charged anion and the negatively charged corneal surface.Benzalkonium chloride (BAC) (Kissner et al., 2010), ethylenediaminetetraacetic acid (ETDA), tocopherol (vitamin E) (Caruso et al., 2016), sodium iodide (NaI), cyclodextrins, as well as iontophoresis (Cassagne et al., 2016;Mazzotta et al., 2018a;Vinciguerra et al., 2021) have been reported to improve the permeability of riboflavin through the epithelium.A higher concentration of riboflavin up to 0.25% could improve the accumulation in the stroma.

Bleaching effect
Under UV irradiation, riboflavin decomposes into lumichromes and lumiflavin.This photodegradation (for irradiance 3 mW/cm 2 about 0.4-1% every 30 s) leads to a reduced riboflavin concentration in the stroma (Diakonis et al., 2012).With increasing irradiance, photodegradation also increases (for irradiance 30 mW/cm 2 about 8% every 30 s) (O'Brart et al., 2018a).Therefore, either riboflavin must be added during irradiation or a higher riboflavin concentration (0.25%) must be used (Mazzotta et al., 2021a;O'Brart et al., 2018b).In a study of Abdshahzadeh et al., the authors demonstrated that the repeated application of riboflavin (without dextran or HPMC) during UV-light irradiation does not improve the corneal stiffening effect in comparison to an application of a phosphate-buffered saline solution (Abdshahzadeh et al., 2022).

Ultraviolet radiation
In the CXL process, a large amount of the radiation energy is converted at a high absorption in the tissue, not into heat but into chemical energy.When riboflavin and UV radiation are applied together, the result is the transformation of radiation energy into different forms of energy: into fluorescence radiation, into chemical energy (cross-linking energy) and only a small part into heat (increase of only 2-3 • C) (Mencucci et al., 2007).This temperature increase is sufficiently low, below the thermal damage threshold of collagen, thus, that the thermal effect is negligible in the case of the photochemical crosslinking method.
The wavelength of light, which occurs at 365 nm, corresponds to the absorption peak of riboflavin.Wavelengths below 300 nm are not acceptable due to potential DNA damage (Seiler et al., 1988), and above 400 nm there is a blue light hazard to the retina (Wu et al., 2006).Knowledge of irradiation parameters (irradiance, irradiation time, and irradiation dose) is essential for understanding and assessing treatment protocols.With an irradiance (intensity) E = power/area in mW/cm 2 (e. g. 3 mW/cm 2 ) and the irradiation time t, the irradiation dose (also called total energy per area or fluence) is calculated as follows From the studies of the dependence of the biomechanical effect on the irradiation time in the range of 5-60 min at an irradiance of 3 mW/ cm 2 , an optimal value of 30 min was obtained (Ahearne et al., 2008;Sporl et al., 2000).
In order to accelerate the treatment, an attempt was made to shorten this irradiation time under the assumption of the Bunsen-Roscoe law.The Bunsen-Roscoe law is well known from photochemistry and states that the photochemical reaction in a thin photo layer depends only on the irradiation dose regardless of the irradiation time.However, the Bunsen-Roscoe law is not applicable to the crosslinking of thick tissues such as the cornea, where oxygen diffusion and the limited number of crosslinking sites on the collagen are further factors.Shortening the therapy time is desirable, but does not bring the same biomechanical effect.The effects of accelerated protocols will be discussed later.

Oxygen
Besides riboflavin and UV light, a third factor is required for the sufficient CXL effect, namely oxygen.The formation of oxygen radicals (hydroxyl radicals OH, superoxide radicals O 2 − , hydrogen peroxide H 2 O and singlet oxygen O) requires oxygen.Thus, the CXL effect is oxygen dependent.McCall was able to suppress the crosslinking reaction by adding reactive oxygen species (ROS) inhibitors (McCall et al., 2010).In the following years, numerous studies confirmed the necessity of oxygen for crosslinking.For example, crosslinking does not occur in an oxygen-free environment (helium atmosphere) (Richoz et al., 2013) and no formation of hydroxyl radicals could be detected in an oxygen-free environment (argon atmosphere) (Sel et al., 2014).Under reduced oxygen conditions, such as contact-lens assisted CXL, the biomechanical effect was found to be lower (Kling et al., 2015;Wollensak and Sporl, 2019).Additionally, a lower stiffening effect was observed in iontophoresis-assisted TE-CXL, in which the authors considered oxygen to be a limiting factor (Torres-Netto et al., 2018).On the other hand, by supplementing oxygen during the CXL procedure the efficacy concerning corneal stiffening can be achieved particularly well in accelerated epi-on protocols with high UV-irradiance (Hill et al., 2020).
Oxygen diffuses from the atmosphere into the cornea.Under normal conditions, the partial pressure of oxygen in the air is 150 mmHg (760 mmHg x 0.2 = 150 mmHg), at the basal epithelium, it is 120 mmHg and decreases to 55 mmHg by the time it reaches the endothelium.Oxygen diffusion is a time-and site-dependent process.During UV irradiation, oxygen is required and the oxygen available in the tissue is consumed very quickly (Kamaev et al., 2012), e.g., oxygen is depleted at 199 μm depth after 15 s.CXL happens as a result of ROS, which then create covalent bonds between collagen and proteoglycan molecules.
Different ways that enable a higher oxygen content in the cornea during crosslinking have been investigated such as an oxygen-rich environment or the use of ozone.In a laboratory study on oxygen kinetics in CXL, Seiler et al. showed that the oxygen concentration decreases with increasing corneal depth and is consumed very quickly during UV irradiation.However, after cessation of UV irradiation, oxygen concentration returns to baseline very rapidly.While at normal oxygen levels (normoxid environment) of 3 mW/cm 2 oxygen is available down to 300 μm depth, at higher irradiance levels, it is only sufficiently available at shallower depths.With increased environmental oxygen (95%), a CXL effect can also be achieved down to 300 μm depth with irradiances between 9 and 18 mW/cm 2 .However, a CXL effect with mW/cm 2 is only possible down to 100 μm (Seiler et al., 2021).Another possibility for oxygen enrichment is the use of ozone, which is an unstable trioxygen molecule (O 3 ).It decays very rapidly (10 min) into free monatomic reactive oxygen (is one of the ROS) and low-reactive molecular oxygen (O 2 ), thus serving as an oxygen supplier.This additional oxygen fraction could enhance the CXL effect.Dogan et al. applied 2 ml of ozonated water (20 μg/ml), which was generated using an ozone generator, to the de-epithelialized cornea for 2 min before irradiation.Increased hyperreflectivity and more compact collagen fibrils were found in this treatment group (Dogan et al., 2020).

Limited number of free crosslinking sites
If UV light, riboflavin, and oxygen are available in sufficient quantities for crosslinking, there will not be unlimited crosslinking but a saturation effect due to the limited number of crosslinking sites on the collagen or proteoglycans in the cornea.
It is known from chemical studies that collagen has only a limited number of free reactive amino acid residues for UV crosslinking (Weadock et al., 1995).Thus, the crosslink density can only increase up to a certain value -the saturation value.The saturation effect can be explained as follows: in the upper corneal layer, the potential crosslinking sites are crosslinked when UV, riboflavin and oxygen are sufficient, and the crosslinking process reaches a saturation effect.The oxygen diffusing through the cornea which does not find free crosslinking sites, migrates to deeper layers and can contribute to F. Raiskup et al. crosslinking there if the UV irradiance is high enough.As a result, the area up to the demarcation line is almost homogeneously (saturated) crosslinked.This is also supported by experimental studies that demonstrated no further biomechanical effect in repeated crosslinking (Beshtawi et al., 2014;Tabibian et al., 2017;Zhang et al., 2023).Thus, the superficial layer of the cornea cannot become more strongly crosslinked because all crosslinking sites are already occupied.Increasing the total energy for S-CXL (3 mW/cm 2 ) by extending the time does not enhance the biomechanical stiffening effect, although oxygen is readily available at a depth of 300 μm (Lanchares et al., 2011).Only when higher irradiances and longer irradiation times (larger energy doses) are used, then a crosslinking effect can also be achieved in deeper layers (Kymionis et al., 2016;Mazzotta et al., 2016b;Seiler et al., 2016).
Another consequence from the saturation of the crosslinking density is that the CXL effect in older corneas is lower.With age, the collagen in the cornea is naturally crosslinked by UV or advanced glycation end products.Thus, only a smaller part of free crosslinking sites is available for additional artificial crosslinking.That is the reason, why the biomechanical effect of CXL in older corneas is lower (Alenezi et al., 2022).

Biochemical mechanism of CXL
In summary of the previously mentioned factors that affect the photo-oxidative crosslinking process, the biochemical mechanism proceeds in several steps (Fig. 3).In the beginning, riboflavin absorbs energy from UV light and is transferred from an excited singlet ( 1 RF*) to an excited triplet riboflavin ( 3 RF*).In this respect, type I and type II reactions are to be distinguished.In type II reactions, oxygen forms singlet oxygen ( 1 O 2 ), which results from an interaction of triplet oxygen ( 3 O 2 ).If the necessary oxygen is depleted by UV light, then the type I reaction prevails.In the CXL method, both reactions take place (Sporl et al., 2008).A direct reaction of the triplet riboflavin with reactive groups of the collagen could not be detected.In the presence of oxygen, this energy is transferred from the excited riboflavin to oxygen whereby oxygen radicals (ROS, hydroxyl radical (.OH), singlet oxygen (1O 2 ), hyperoxide anion (O 2 -), hydrogen peroxide (H 2 O 2 )) are formed and the excited riboflavin subsequently returns to the ground state.Mainly singlet oxygen occurs as an intermediate.This further reacts with the amino acids (histidine, hydroxyproline, hydroxylysine, tyrosine, threonine, phenylalanine) of the collagen side chains, with reactions occurring mostly at the carbonyl groups (C --O groups).Amino groups (NH 2 -) do not play a significant role in UV crosslinking (McCall et al., 2010).These photochemical reactions cause further covalent interfibrillarly and intrafibrillarly bonds at the collagen and proteoglycans as well as between the fibrils and proteoglycans, thereby stabilizing the collagen fibers and improving the collagen structure (Zhang et al., 2011).It appears that interfibrillar linkages can also be formed but CXL does not increase the number of interlamellar cross-linkers (Wollensak et al., 2011).
Hayes and coworkers investigated the swelling behavior (incubation in saline solution) of corneal buttons and found no differences in the swelling rate of riboflavin and UV treated, riboflavin only, and untreated buttons.Another important finding was that the intermolecular spacing of the collagen was not different in treated and untreated corneas.Thus, the authors concluded that CXL does not occur between collagen and proteoglycans, instead, CXL primarily takes place within and between molecules on the fibril surfaces, and within proteoglycan core proteins in the interfibrillar space (Hayes et al., 2013).Another ex vivo study showed a significant reduction in stromal pressure in porcine and human corneas after CXL, suggesting that this procedure may reduce corneal stromal pressure in vivo, thereby reducing edema and improving vision (Sondergaard et al., 2013b).The same research group from Denmark also investigated the effect of CXL on stromal shear moduli ex vivo.They found higher shear moduli in the anterior compared to the posterior stromal sheets (Sondergaard et al., 2013a).

CXL protocols 2.5.1. Application of riboflavin for stromal diffusion
A high concentration of riboflavin is necessary for a strong biomechanical crosslinking effect in the stroma.There are different methods that can be used to achieve the desired concentration of riboflavin in the stroma.Until present, the discussion regarding epithelium-off (epi-off) or epithelium-on (epi-on) CXL is ongoing.
The originally proposed method for riboflavin to soak into corneal stroma was to de-epithelialize the cornea with alcohol or surgical instrument.The advantage of this method is that there is no barrier limiting the diffusion of riboflavin or oxygen into the stroma.However, the removal of the epithelium carries some risks as the layer is an important penetration barrier for microorganisms that could cause infections (DelMonte and Kim, 2011), causes pain after the surgery, which may last up to 3 days, and leads to increased risk for infiltrates, haze, and scarring.
F. Raiskup et al.In view of these negative side effects and possible vision-threating complications, physicians are trying to find out a treatment that does not require epithelium removal, the so-called "epithelium on" or "transepithelial" CXL (TE-CXL).During the TE-CXL, a stromal penetration of riboflavin should be enabled despite of the hydrophobic behavior of the epithelium.There are several methods that can be used to increase the penetration of riboflavin through the epithelium such as altering epithelial permeability or alterations of the physicochemical properties of the riboflavin solution.BAC increases epithelial permeability by loosening its tight junctions.This pharmacological modification of corneal epithelial permeability represents an alternative method that allows the omission of epithelial debridement in CXL.Therefore, there are some efforts to modify the standard protocol and perform treatment without epithelial debridement using BAC, tetracaine, or pilocarpine, which contains BAC and EDTA.The first experimental study regarding the epi-on CXL procedure was performed by Wollensak et al., who tested isoosmolar riboflavin (with 20% dextran) + 0.005% BAC in rabbit eyes and found a weak biomechanical effect (Wollensak and Iomdina, 2009).Kissner from our group tested several concentrations of BAC in a solution of 0.1% riboflavin in 0.44% NaCl on rabbit eyes and found good UV absorption and a satisfactory biomechanical effect using a solution containing 0.02% BAC in experimental setting (Kissner et al., 2010).
An osmotic gradient between the apical and basolateral sides increases paracellular conductance without ionic selectivity.Basolateral hypoosmolarity increases paracellular conductance; also, apical hypoosmolarity increases paracellular conductance moderately.The hydrostatic pressure gradient acts as a driving force for water movement.The osmotic gradient also acts as a driving force for water movement, i.e., if the permeability of the water is sufficiently higher than the permeability of the solution, the concentration difference of the solution acts as a driving force for water movement.Dextran inhibits the growth of paracellular conductance.This inhibition has shown to be dependent on the concentration of dextran, regardless of the site of administration.For this reason, the transepithelial riboflavin solution for the TE-CXL procedure should not contain dextran but 0.01% BAC and 0.44% NaCl to increase epithelial permeability, thereby achieving high riboflavin concentrations in the corneal stroma (Raiskup et al., 2012).
The iontophoresis CXL procedure is another technique in which negatively charged riboflavin molecules are passed through the epithelium to the stroma with an electric current (1 mA).A negative electrode is placed onto the cornea and the positive electrode is clamped onto the forehead.The riboflavin solution is added with special supplements such as BAC or EDTA, or contains a higher concentration of riboflavin (Vinciguerra et al., 2021)

UV irradiation
The wavelength of the light beam during the CXL procedure is 365 nm due to riboflavin having its absorption maximum at this wavelength.In contrast to the beginnings of CXL, when the procedure was performed with two UV diodes only, today's devices can irradiate with a homogeneous and variable light cone (Fig. 4).There is the top-hat profile and the beam optimized profile (Cummings et al., 2016;Herber et al., 2018).Some devices even offer individual irradiation patterns (Seiler et al., 2016).Moreover, the UV intensity can also be increased by reducing the irradiation time according to the Bunsen-Roscoe law.Those protocols are referred to accelerated CXL.

a) Standard CXL protocol -S-CXL
The standard CXL (S-CXL) is performed in outpatient facilities providing day surgical care.Thirty minutes before the procedure, the patient is given painkillers and, if necessary, sedatives.The procedure is performed under sterile conditions in the operating room.Recently, a crosslinker device has started to be used to perform the procedure on a slit lamp microscope in a sitting position in the outpatient clinic (Hafezi et al., 2021b).After application of local anesthesia, an eyelid speculum is applied, and the corneal epithelium is removed from an area of 8.0 mm in diameter.The 0.1% riboflavin solution is applied to the cornea every 2 min for 20 min to achieve sufficient penetration of the solution.When using HPMC-based solutions, the imbibition time can be reduced.After epithelium removal (and also at each step of the procedure), the corneal thickness is measured using ultrasound pachymetry or optical coherence tomography measurements.Based on the results of this measurement, a decision is made as to which riboflavin solution will be used to saturate the cornea, i.e. whether isoosmolar or hypoosmolar.Before starting UV irradiation, it should be verified that the thickness of the irradiated cornea is greater than 400 μm to protect the corneal endothelium.Otherwise, the cornea can be treated with an appropriate protocol for thin corneas (Hafezi et al., 2021a).An area of 8.0 mm in diameter in the central part of the cornea is irradiated with an intensity of 3 mW/cm 2 for 30 min.Before treatment, the intensity of the radiation is checked using a calibrated UV light meter.Local anesthetics are added during the procedure as needed.Postoperatively, the patient receives topical antibiotics and lubricants.A soft therapeutic contact lens is applied until the re-epithelialization process is complete.After corneal re-epithelialization, topical steroid treatment is given for 4 weeks.Patients are followed daily until epithelialization.After epithelialization, follow-ups are performed at 1, 3, 6 and 12 months and then every year.
Once the cornea has stabilized, approximately 6-8 months after CXL, a new contact lens can be fitted.
The disadvantages of the treatment are the long treatment time of overall 60 min, epithelial removal and therefore increased risk for infectious keratitis and sterile infiltrates, pain, and the limitation to corneal thickness above 400 μm.

b) Accelerated CXL
The accelerated protocol (A-CXL) was developed based on the Bunsen-Roscoe reciprocity law to shorten the duration of the treatment procedure.This law states that the degree of photochemical reaction is directly proportional to the total energy, regardless of the time in which the dose is applied.In A-CXL, the irradiance is increased, while the irradiation time is decreased so that the dose of 5.4 J/cm 2 remains constant.The most widely used A-CXL protocol uses an intensity of 9 mW/cm 2 for 10 min (Raiskup and Herber, 2022).It can also be combined with pulsed irradiation, in which the beam is switched on and off for 1 s, increasing the treatment accordingly to a total fluence of 5.4 J/cm 2 (Mazzotta et al., 2014a(Mazzotta et al., , 2014b)).The intraoperative procedure and the postoperative treatment with eye drops is similar to the S-CXL.Due to the different parameters that can be varied, such as intensity, time, and type of riboflavin, a standardized terminology and protocol nomenclature was proposed by Randleman et al., in which the parameters should be listed in a table and abbreviated as follows: e.g.A-CXL (intensity*time), A-CXL(9*10) (Randleman et al., 2017).

Effects and proof of crosslinking in the cornea
Photochemically induced crosslinking in the cornea cannot currently be directly in vivo visualized by staining methods or other microscopic techniques yet.Crosslinking causes changes in many physicochemical properties of collagen-containing tissues, from which it can be indirectly inferred that crosslinking has occurred.Some of the changes in the corneal stroma that are caused by crosslinking are described below.

Resistance against deformation -ex vivo
CXL aims to increase the strength of tissue in biomechanically weakened corneas such as keratoconus.Thus, a major part of the procedure is to measure the (bio-) mechanical strength (increase in modulus of elasticity) of the tissue.According to mechanical science, there are two ways to determine the stiffness that can be used for applications of the cornea: stress-strain extensometry (Fig. 5) and inflation tests (Elsheikh and Anderson, 2005).
Previous studies have found that corneas that have undergone riboflavin and UV light CXL are stronger than normal corneas by a factor of 1.7, and thus, the strengthening effect is sufficient to compensate the reduced stability observed in keratoconus cases (Kohlhaas et al., 2006;Spoerl et al., 1998;Wollensak et al., 2003b).The evidence for CXL was provided by histological findings by detecting cross-links in the anterior part of the cornea with a Karnovsky solution in the initial experiments of our group (Spoerl et al., 1998).
According to the photochemical law of reciprocity (Bunsen-Roscoe law), the same photochemical effect can be achieved with reduced irradiation time and correspondingly increased irradiation intensity so that the total dose remains the same.In biomechanical studies, the results are slightly inconsistent.Wernli et al. found a significant stiffening effect up to 45 mW/cm 2 that corresponds to an irradiation time of 2 min (Wernli et al., 2013).Other studies reported similar stiffening effects up to 15 mW/cm 2 compared to the S-CXL (Krueger et al., 2014;  4) Stress-strain measurement using an extensometer.
F. Raiskup et al.Schumacher et al., 2011).Contrarily, Hammer et al. found less biomechanical effect when using the 18 mW/cm 2 for 5 min protocol, which showed no statistical significantly effect in comparison to control eyes (Hammer et al., 2014).Moreover, the S-CXL protocol was superior to the A-CXL(9*10) and A-CXL(18*5) protocol.Bao et al. demonstrated a gradually decreasing tangent modulus when higher irradiances (and shorter times) were used.The authors treated rabbit corneas and applied inflation tests after euthanasia.The elastic modulus at physiological strain (normal IOP conditions) was statistically significantly increased in S-CXL and A-CXL(9*10) compared to untreated control eyes (Bao et al., 2018).When investigating corneal stiffness at 2% and 4% strain using stress-strain measurements, the S-CXL and A-CXL(9*10) protocol led to a significant increase in stiffness compared to controls and A-CXL(9*18) treated eyes (only comparison with S-CXL) (Herber et al., 2020a).The results of Bao et al. and Herber et al. were in accordance with those of Hammer et al., indicating that CXL is effective up to 9 mW/cm 2 .Higher irradiances with lower period of time (total energy of 5.4 J/cm 2 ) seemed to be less effective.Another conclusion is that the Bunsen-Roscoe law has its limitations on complex photochemical applications such as CXL, where other factors such as oxygen and limited number of crosslinking sites are involved.

Resistance against deformation -in vivo
Since the introduction of corneal crosslinking, the procedure has been tested on several occasions, by using destructive methods such as stress-strain measurements under ex vivo or in vitro conditions.These methods cannot be applied in vivo, whereby a non-invasive and an easy to use measurement is needed to clinically demonstrate the stiffening effect.For this purpose, a force applied to the tissue is required to evaluate the biomechanical properties.The principle of air-puff tonometry meets these criteria because the air-puff exerts an external force when it hits the front surface of the cornea.This leads to a deformation process of the cornea.
Two commercially available devices that measure the biomechanical properties of the cornea are: the Ocular Response Analyzer (ORA, Reichert, Ophthalmic Instruments, Depew, NY, USA) and the Corvis ST (CVS, Oculus GmbH, Wetzlar, Germany).In both, an air-puff is applied to the cornea and the inward and outward movement is detected or recorded, respectively.From the deformation behavior of the cornea, conclusions can be drawn about its biomechanical properties.However, the two devices differ concerning the technique of detecting and recording the deformation process.The ORA is a dynamic measurement method using a bidirectional applanation process that predominantly characterizes the mechanical inertia and viscous properties of the cornea (basic substance, collagen-matrix interaction) (Luce, 2005;Terai et al., 2012).Besides the intra-ocular pressure (IOPg) and the corneal-compensated intraocular pressure (IOPcc), the main biomechanical parameters are the corneal hysteresis (CH) and the corneal resistance factor (CRF).
Conversely, the CVS records the corneal deformation in a twodimensional image of an 8 mm cross-section using an ultra-high speed Scheimpflug camera.Shortly after the clinical introduction, the device was able to measure intra-ocular pressure, corneal thickness, and deformation corneal response (DCR) parameters such as deformation, time, and velocity until the first and second applanation (Hon and Lam, 2013).Later, novel parameters were introduced, for instance, the integrated inverse radius (IIR), which is the sum of the inverse radius (reciprocal of the radius of the concave state of the cornea) between 1st and 2nd applanation and the deformation amplitude ratio (DAR2, DAR1), which is the ratio of the central deformation amplitude and mean peripheral deformation at ± 2 mm or ± 1 mm from apex.Further, the biomechanically corrected IOP (bIOP) obtains IOP estimates that are less affected by the main deformation corneal response parameters, corneal thickness, and age (Joda et al., 2016;Vinciguerra et al., 2016b).Furthermore, the stiffness parameter (SPA1) is defined as the difference of inward pressure (pressure of the air-puff) and the bIOP divided by the corneal displacement (inward direction) at the 1st applanation (Roberts et al., 2017).In this context, two indices were introduced for ectasia and keratoconus detection (Ambrosio et al., 2017;Vinciguerra et al., 2016a).Several studies confirmed the ability of the parameters of both devices to discriminate between healthy and keratoconus eyes (Herber et al., 2019;Sedaghat et al., 2018a), with the parameters of the CVS being preferable in the evaluation of very early ectasia (Herber et al., 2022a).The dynamic behavior of a keratoconus cornea is characterized by a lower CRF, CH, SPA1, and increased IIR, DAR2, and DAR1 compared to normal eyes (Herber et al., 2019).Other studies confirmed that the amount of DCR parameters are in relation to the severity of KC, which means the higher the stage of KC, the weaker the corneal behavior is against force (Herber et al., 2020b;Koh et al., 2020).Furthermore, the DCR parameters are highly repeatable and reproducible in keratoconus eyes, which is an important requirement to assess these parameters in longitudinal studies (Herber et al., 2020c).
Regarding the CXL treatment, it can be stated that the main parameters (CRF and CH) of the ORA do not change postoperatively (Goldich et al., 2009;Greenstein et al., 2012b;Spoerl et al., 2011b).The reasons for this observation could lie in the ORA method itself and not due to the absent effect of CXL, since the CH and the CRF represent more viscoelastic properties of the cornea (Roberts, 2014).In contrast, the DCR parameters of the CVS are closely related to elastic properties of the cornea, so a change after CXL is more likely (Fig. 6).In an ex vivo study, the relationship between the change of DCR parameters and the increase in corneal stiffness by stress-strain measurement was found in different CXL protocols (Herber et al., 2020a).These results confirmed the ability of the CVS to detect biomechanical changes induced by CXL.Simultaneously, several in vivo studies showed that DCR parameters were altered by CXL after one month and up to 4 years (Hashemi et al., 2019;Jabbarvand et al., 2021;Sedaghat et al., 2018b;Vinciguerra et al., 2017), with IIR showing the best results to detect corneal stiffening.An overview of observed changes in DCR parameters is given in Table 1.
The results were confirmed by a study, which investigated a cohort of patients in a follow-up of 1 month, 6 months, and 12 months, showing that the CXL effect measured by DCR parameters was the most pronounced after one month (Herber et al., 2023a).Interestingly, the IIR value decreased statistically significantly after one month, indicating corneal stiffening, although corneal thickness also decreased.The contralateral eyes of these patients, which were untreated, did not show any changes during the follow-up period (Fig. 7).The results should be interpreted as follows: The IIR value is increased when the cornea is weaker (keratoconus), a decrease in the value after CXL treatment indicates stiffening of the cornea (Herber et al., 2023a).In another study, the repeatability of specific DCR parameters were investigated before, and one month after CXL treatment.It has been shown that the change of IIR was more pronounced pre-and postoperatively than the variation or measurement noise itself (Herber et al., 2023b).This study supported the previous findings leading to the conclusion that specific biomechanical related parameters of air-puff tonometry are able to detect corneal stiffening in vivo after CXL.

Stronger crosslinking effect in the anterior cornea
The photo-oxidation method does not achieve homogeneous crosslinking throughout the total corneal depth.Due to the strong absorption coefficient, the UV intensity decreases with increasing depth, resulting in a decreasing degree of crosslinking.In addition, the oxygen decreases with increasing depth, whereby both factors cause the stronger effect in the anterior part of the cornea.Other observations were a smaller increase in fiber diameter in the posterior region (Wollensak et al., 2004c), higher enzymatic degradation (Schilde et al., 2008;Spoerl et al., 2004c), and a greater shrinkage effect (Spoerl et al., 2004a) in the posterior crosslinked cornea.In an experimental study using enucleated porcine eyes, it was possible to demonstrate that the stiffening effect is present only in the upper third (approximately 200 μm) of the corneal stroma (Kohlhaas et al., 2006).In these studies, two 200-μm-thick lamellae F. Raiskup et al. were excised from the cornea (Fig. 8).The strength of the stiffening of the upper lamella that was treated was significantly higher than the lamella, that was lower positioned (deeper) and the one of the matched controls.The strength of the second treated lamella was only slightly higher than that of the second untreated lamella.Analogous results were also observed in biochemical studies of crosslinking in stromal layers from different depths (Schilde et al., 2008).The decreasing crosslinking effect with increasing depth can be explained by the fact that the riboflavin concentration decreases linearly with increasing corneal depth, in accordance with the diffusion gradient (Cui et al., 2011).The exponential decrease in UV intensity is much greater which is in accordance with the Beer-Lambert law.Therefore, 65% of the UV radiation is absorbed in the upper 200 μm.Since the enhancement effect is primarily dependent on the UV intensity and to a lesser extent on the riboflavin concentration, the enhancement effect is highest in the first superficial 200-250 μm.

Summary of structural and cellular changes (modifications) after CXL
As a result of the crosslinking-induced changes of the corneal lamellar structure (Fig. 9), the increase in the diameter of the fibrils (Chang et al., 2018;Wollensak et al., 2004c), the decrease in the distance between the fibrils (interfibrillar spacing) (Bao et al., 2018), the denser packing of the collagen fibers, and the decrease in the interlamellar spacing (Subasinghe et al., 2021) and the increase of the biomechanical strength of the cornea were observed (Wollensak et al., 2003b).CXL causes the collagen fibers to shorten and tighten, decreasing the waviness or collagen fiber curvature.As a result, the fibers are more densely (compactly) arranged (Hepfer et al., 2021).In a recent study, molecular alterations within the corneal tissue were examined by surface enhanced Raman spectroscopy, and molecular bonds were found to be involved in the crosslinking process, but without the appearance of bonds between individual collagen chains.(Melcher et al., 2023).Therefore, the corneal stiffening effect might not be primarily driven by forming covalent bonds between side groups of the collagen chains, and seemingly other processes are occurring that are not associated with structural alterations (Melcher et al., 2023).
Moreover, CXL causes an apoptosis of keratocytes in the anterior stroma.Wollensak et al. found an irradiation threshold at 0.5 mW/cm 2 for a cytotoxic effect on keratocytes, when the cells were treated with riboflavin and UVA light.The threshold was 10 times higher when only UVA light was applied (Wollensak et al., 2004a).Further, the depth of the apoptosis of keratocytes depended on the fluence of UVA irradiation (Wollensak et al., 2004b).In human eyes, an apoptosis up to 300 μm was clinically expected using 3 mW/cm 2 for 30 min (5.4J) (Wollensak et al., 2004a(Wollensak et al., , 2004b)).After the treatment, a re-population of keratocytes proceeded gradually up to 3 months, which was shown ex vivo (Chen et al., 2021) and in vivo using confocal microscopy (Mazzotta et al., 2007b).
Further, the transition zone between the cross-linked anterior corneal stroma and untreated posterior corneal stroma is visible in vivo as so-called stromal demarcation line, which was first observed during slit-lamp examinations (Seiler and Hafezi, 2006).This demarcation line may result from differences in the refractive index and/or reflection properties of untreated versus cross-linked corneal stroma (Seiler and Hafezi, 2006;Wittig-Silva et al., 2013) and correlates to the treatment induced keratocytes apoptosis depth (Mazzotta et al., 2019).In addition, on clinical examination after CXL, typical corneal opacification is observed that is most served after one month, reaches the plateau phase after 3 months, and decreases significantly between three and twelve months (Greenstein et al., 2010).This transient effect can be explained by the fact that the wound healing process is accompanied by an Fig. 6.Cross-section Scheimpflug image of the cornea preoperative (red) and postoperative (green).A less deformable cornea is observed after the CXL treatment (green cross-section of the cornea).

Micromorphological changes of the corneain vivo
In vivo confocal microscopy (IVCM) enables the examination of the entire living human cornea at the cellular level, which has provided a new understanding of postoperative corneal changes after CXL (Mazzotta et al., 2007a(Mazzotta et al., , 2007b(Mazzotta et al., , 2008(Mazzotta et al., , 2015)).Due to the detailed resolution and high magnification of the individual corneal layers, it is possible to examine corneal changes with temporal resolution (early and late phase) (Efron and Hollingsworth, 2008;Mazzotta et al., 2012).It could be shown that the keratocytes disappear from the anterior and middle stroma due to apoptosis in the early (1st to 6th month) and late postoperative (after 6th month) phase, whereby repopulation can already be observed between the second and third postoperative month (Croxatto et al., 2010;Mazzotta et al., 2008;Touboul et al., 2012).This occurs centripetally from the non-irradiated peripheral area (over 8-9 mm in diameter) and from the deeper stromal layers.The repopulation is commonly completed after 12 months, which is documented by a Fig. 7. Change of the mean integrated inverse radius (IIR) from baseline to each follow-up (above).Direct comparison between treated eyes and untreated contralateral eyes of the mean change at 1 month, 6 months, and 12 months compared to baseline (below) (Herber et al., 2023a).(Kohlhaas et al., 2006).gradual reduction of the edema and a progressive reappearance of activated nuclei (Mazzotta et al., 2015).While in S-CXL keratocyte apoptosis occurs up to a depth of 340 μm, in TE-CXL this apoptotic phenomenon is less obvious and more uneven up to a maximum depth of 100 μm.The reason for this is the limited saturation of the stroma with riboflavin.In addition, neither relevant changes in the density of the surface nor relevant changes in the extracellular matrix density and microtriate reflections were observed (Mazzotta et al., 2015).
The stromal edema usually remains up to three months after surgery and decreases up to the 6th postoperative month.Postoperative stromal edema was associated with significant keratocyte loss in the first month and a dense network of hyperreflective extracellular tissue described as "trabecular patterned stroma" (Mazzotta et al., 2008).A variable increase in extracellular matrix density is typical after conventional CXL, especially in IVCM up to 6 months postoperatively, and is seen as hyper-reflective tissue around keratocyte nuclei (Mazzotta et al., 2008).
As previously stated, the demarcation line seen after CXL is a result of light scattering (reflectivity changes) caused by different tissue densities, which explains the transition from an edematous region without cells to an area unreachable by UV radiation and regularly populated by cells (Mazzotta et al., 2015).
Other significant corneal changes have been observed in the subepithelial nerve plexus, where an immediate loss occurs after CXL.From the surrounding non-irradiated areas of the cornea, the subepithelial nerve fibers regenerate between the 2nd and 3rd postoperative month, which is almost complete after the 6th month and correlates with the recovery of corneal sensitivity.No neurodystrophic effects were observed.However, the original subepithelial nerve plexus structure is only restored after approximately 2 years (Mazzotta et al., 2008).
The corneal endothelium remains unchanged in its morphology and function, which has been studied according to different CXL protocols, confirming the safety of the procedure (Mazzotta et al., 2015;Touboul et al., 2012).
After TE-CXL, IVCM examination revealed transient changes in corneal epithelial cells.The toxic effect on the epithelium is related to photonecrosis and UVA-induced apoptosis, which affect and delay epithelial cell turnover during the first three months after treatment.In addition, other toxic components present in TE riboflavin solution such as anesthetics, surfactants, tension enhancers and preservatives result in diffuse punctate epitheliopathy requiring postoperative therapeutic soft contact lens wear (Mazzotta et al., 2015).

CXL safety
The safe use of UV light requires some precautions.Before irradiation, the measurement of corneal thickness is an essential step.With a stromal thickness greater than 400 μm, approximately 90% of UV radiation is absorbed within the cornea, so there is no risk to the endothelium (Wollensak et al., 2003a), lens (Vinciguerra et al., 2011), or retina (Lazaridis et al., 2020;Spoerl et al., 2007a).
At the beginnings of the CXL procedure, many studies and calculations were performed by our group in order to ensure sufficient safety for the eye.The health and safety standards and regulations allow a daily UVA radiation intensity of 1 mW/cm 2 without the use of a photosensitizer for the unprotected eye (Sliney et al., 2005).Approximately 35% of this amount is absorbed by the cornea, so it can be assumed that the endothelium is exposed to a radiation intensity of 0.65 mW/cm 2 .In animal experiments, the threshold for endothelial cell damage with min of UVA exposure without a photosensitizer was found to be mW/cm 2 (Wollensak et al., 2003c).In the case of CXL, where riboflavin (photosensitizer) and UVA light (3 mW/cm 2 ) is used, 94% of the UVA radiation is absorbed in an approximately 400 μm thick stromal layer treated with riboflavin, i.e., the endothelium is exposed to only 0.18 mW/cm 2 (Spoerl et al., 2007a(Spoerl et al., , 2011a)).This value is below the damage threshold of 0.35 mW/cm 2 established in animal experiments for the endothelium (Wollensak et al., 2003c), as well as below the 0.65 mW/cm 2 defined in health and safety regulations (Fig. 10).Safety for the retina can be further improved by the choice of irradiation equipment.Due to the short distance of the radiation source from the eye and the divergent irradiation (Koehler principle), the UV radiation is not focused on the retina.Therefore, only very low-density radiation, which is well below the threshold of harmfulness, reaches the lens or retina.Using Scheimpflug imaging, no significant differences were found in crystalline lens density before CXL and 12 months after CXL (Grewal et al., 2009;Vinciguerra et al., 2011).No morphological changes in the retina were observed after CXL (Romano et al., 2012).
In order to limit the intensifying effect to the corneal stroma only, it must be ensured that most of the UV radiation is absorbed in this area, as radiation has an effect only at the point where it is absorbed and where it transfers energy to the tissue (1st photochemical law).
Recently, measurements of riboflavin concentration at the endothelium showed values of 0.015 ± 0.005%, compared with the theoretically calculated value of 0.025% -it means, treatment could provide safety even up to 350 μm (Seiler et al., 2019a).In clinical routine, the "400 μm safety rule" has proven to be effective.If the measured corneal thickness is lower than 400 μm, then one should either use hypoosmolar riboflavin or choose the irradiation parameters for thin corneas.
During irradiation, the distance between the irradiation source and the corneal surface specified by the manufacturers must be regarded.Furthermore, it must be ensured that the irradiation area does not reach the limbus (Vimalin et al., 2012).

Keratoconus -characteristics
Keratoconus is a progressive corneal disease that begins in 84% of cases between the ages of 20 and 49 years (Zadnik et al., 1998) and typically involves bilateral (80-85%) conical corneal bulging and stromal thinning.The manifestation of the disease is highly variable.It can range from mild to high irregular astigmatism and myopia resulting in a severe deterioration of visual acuity due to increased corneal bulging and subepithelial scarring.

Epidemiology
The data regarding incidence and prevalence of KC differ between former studies from the 1980s to current studies from the last decade (Hashemi et al., 2020;Kennedy et al., 1986).The discrepancy arises due to several reasons such as the technical improvement of diagnostic devices, availability of health care, differences in the design and methodology of the studies, and local or ethnic variations (Cozma et al., 2005;Pearson et al., 2000), thus, KC cannot be considered a rare disease anymore.
A retrospective database study from the United States tracked the occurrence of keratoconus between 1935 and 1982 and reported an overall prevalence rate of keratoconus of 54.5 per 100,000 population (Kennedy et al., 1986).Data on the incidence of KC based on this epidemiological study have reported the number of newly diagnosed cases to be 2.0 per 100,000 population per year.The Dutch epidemiological study reported an incidence of KC with 13.3 new cases per 100, 000 population per year.The prevalence of keratoconus in the general population in this study was estimated at 265 cases per 100,000 population (Godefrooij et al., 2017a).The recent Danish epidemiological study found that the current incidence rate has increased 2-3-fold in the last 10-15 years.The authors suggest that this increase may be a consequence of more complete data due to the availability of CXL, which has changed the referral pattern from the primary sector, better diagnostic tools and immigration may also play a role (Bak-Nielsen et al., 2019a).The reasons for the different results are based on the different study designs and the recording of the KC diagnosis in the national health insurance registry database (Bak-Nielsen et al., 2019a).However, these current data are very important for the health care system to assess the real impact of this disease on the incurrence of costs.

Risk factors
In 2015, the panelist of the "Global Consensus on Keratoconus and ectatic Diseases" defined KC as a "multifactorial disease"; however, it also identified important risk factors for the development and progression of keratoconus: Down syndrome, positive family history, ethnicity, Leberʼs congenital amaurosis, connective tissue disorders (Marfan syndrome, Ehlers-Danlos syndrome), atopy, and mechanical influences such as eye rubbing and "floppy eyelid" (Gomes et al., 2015).Although the exact pathophysiological correlations are not yet fully understood, eye rubbing in particular is considered to be an important predisposing factoran important information based on a detailed medical history and the only modifiable risk factor.A thorough medical history with focus on habitual eye rubbing and allergic disposition is therefore essential due to it becoming a serious concern when it is performed too frequently or too intensively (McMonnies, 2009a), especially in corneas susceptible to developing ectasia.A clear association with the development of keratoconus has been shown in case-control studies (Bawazeer et al., 2000;McMonnies, 2009b).
There are several pathophysiologic mechanisms that may explain the association between eye rubbing and the development or progression of keratoconus, namely (bio-) mechanical and biochemical processes.Eye rubbing leads to increased corneal temperature, which is associated with increased activity of collagenases, which in turn leads to increased degradation of the collagen matrix leading to a deterioration of biomechanical properties of the cornea.During eye rubbing, it has been shown that both intraocular pressure (up to 10-fold) (McMonnies, 2008) and hydrostatic pressure is increased, which then leads to damage of keratocytes, especially those in the anterior stroma (Bron, 2001).The hydrostatic pressure causes compression and shear stress in the cornea, especially in the epithelial cells and keratocytes.This leads to altered or accelerated metabolism in these cells (Nakamura et al., 2006) and increases the activity of certain enzymes, including MMPs (McMonnies, 2009b) and causes a rearrangement of the cytoskeleton, changing the shape of the cell.The mechanism of cell damage by increased hydrostatic pressure has already been described for astrocytes and plays an important role in the pathogenesis of primary chronic open-angle glaucoma (Hernandez, 2000) as well as in the increased activity of MMPs (Salvador-Silva et al., 2004).Consequently, especially keratoconus patients should be urged to avoid pressure increases that may result from eye rubbing, weightlifting, certain positions in yoga, or playing wind instruments (McMonnies, 2010).That playing wind instruments in particular leads to an intermittent IOP spike has also been shown in glaucoma patients and was associated with a significant increase in visual field defects (Schuman et al., 2000).Another pathomechanism in the development of keratoconus is seen in the sliding of collagen fibrils, which can lead to their rearrangement and formation of a cone (Cristina Fig. 10.Schematic presentation of the theoretically calculated toxic threshold for endothelial cells regarding the CXL procedure (Wollensak et al., 2003c(Wollensak et al., , 2003d)).Kenney and Brown, 2003).This process is mainly promoted by increased activity of proteases.Mechanical stress in the form of rubbing, especially across the arrangement of such "sliding" collagen fibrils, promotes the development of corneal ectasia and can lead to tears in both Bowman's and Descemet's membranes (McMonnies, 2009b;Raiskup et al., 2018).
The sole process of shear forces within the cornea as a trigger of biomechanical weakening is in contrast to recent experimental findings.Netto-Torres et al. investigated the elastic modulus in ex vivo porcine eyes immediately after 10,500 rubs compared to a non-rubbed control group resulting in no difference of corneal stiffness between the groups (Torres-Netto et al., 2022).However, the shear forces act on the corneal cells and trigger a biochemical cascade with the result that the biomechanical strength of the cornea is weakened (Dou et al., 2022), which could take a long time (Zhang et al., 2021).

Evidence of progression
When CXL was introduced into clinical practice, there was no consistent definition of progression that provided an explicit indication for the performance of this treatment.In their first study, Wollensak et al. described that progression was confirmed by the medical history of the patients.Some of them had shown progression in topography readings within 6 months preoperatively (Wollensak et al., 2003a).Later on, Caporossi and coworkers described the progression of their patients as clinically and instrumentally documented significant changes in specific parameters within 6 months (Caporossi et al., 2010).With increasing evidence of positive results of the treatment, a standardized definition of the progression of the disease became necessary.The CLEK study revealed that the keratoconus progresses naturally by demonstrating a gradual increase in corneal curvature and a decrease in visual acuity over a follow-up of 8 years.However, only 24% of eyes had an intensive progression of more than 3 diopters (D) (Wagner et al., 2007).Therefore, not all keratoconus patients have a progression and are suitable for CXL, which was confirmed by a longitudinal study, where only 26.5% of eyes progressed by 1.5 D (Choi and Kim, 2012).In the following, various definitions based on changes in visual acuity, keratometry readings, and spherical as well as cylindrical values were used (Raiskup-Wolf et al., 2008;Vinciguerra et al., 2009;Wittig-Silva et al., 2014b).
In Dresden, progression indicating the need for CXL treatment was defined as an increase of 1 D in the maximal keratometry (Kmax) value within one year, or deterioration of visual acuity, or the need for a new contact lens fitting more than once every two years (Raiskup-Wolf et al., 2008).Vinciguerra et al. defined progression of keratoconus as a change of either myopia and/or astigmatism of at least 3 D in the last 6 months or an average change in central keratometry of at least 1.5 D, or a decrease in average central corneal thickness of at least 5% observed in 3 consecutive topographies or tomographies performed in the last 6 months, respectively (Vinciguerra et al., 2009).
In the three major randomized, controlled trials, progression was determined as a deterioration of visual acuity and one of the following criteria such as an increase of at least 1 D in steepest keratometry (K steep), an increase in astigmatism of a manifest refraction of at least 1 D, or a decrease in radius of back curvature of the contact lens of 0.1 mm within 12 months (Wittig-Silva et al., 2014b); as changes in K steep of at least 1 D, an increase in manifest refraction of at least 1 D, or an increase in manifest refraction spherical equivalent of at least 0.5 D within 24 months (Hersh et al., 2017b); as an increase in K steep or Kmax of at least 1.5 D during a continuous 3 month interval (Larkin et al., 2021).
The current standard of care is to await the progression, in which topographic or tomographic measurements are repeated within a follow-up interval of 3 up to 6 months in dependence of age and risk factors.The detection of true progression as a change in corneal morphology depends on the measurement noise of the devices.It can be assessed by test-retest repeatability (TRT) studies, in which a threshold for clinical progression can be calculated.Certain studies showed that this threshold depends on the type of the device and on the stage of the keratoconus, thus, a generalized definition is somewhat difficult.For instance, the TRT of Kmax of the Scheimpflug based tomographer exceeds the threshold of 1 D in moderate and advanced keratoconus indicating that an increase of more than 1.5 D would be more suitable in those cases (de Luis Eguileor et al., 2021;Kreps et al., 2020).In accordance to those intraday repeatability studies, Gustafsson et al. investigated the interday repeatability in which two different time points were considered for the analysis.This approach makes the evaluation of the measurement noise more reliable.It was found that an increase in (Scheimpflug-based) Kmax of 0.67 D and 1.42 D for Kmax <49 D and Kmax >49 D, respectively, was associated with progression of keratoconus (Gustafsson et al., 2021).Thresholds are slightly lower when measurements are repeated in each session, for example 4 times (0.44 D for Kmax <49 D and 1.08 D for Kmax >49 D).
Recent studies investigated the TRT of Kmax in optical coherence tomography (OCT) based tomographers and found that the threshold in moderate keratoconus is lower than 1 D (Herber et al., 2022b;Seiler et al., 2022).Therefore, this technology might be more appropriate for investigating progression with a change of 1 D in higher stages of keratoconus.
Over the years, the paradigm changed from a qualitative subjective to a quantitative objective assessment of progression to be more reliable and reproducible.This was also recorded in the "Global Consensus on Keratoconus and Ectatic Diseases".The panel proposed the definition of progression as a change in anterior corneal surface (steepening), posterior corneal surface (steepening), and corneal thickness (general thinning or increase in the rate of corneal thickness change from the periphery to the thinnest point) (Gomes et al., 2015).This led to a specific software analysis of the Scheimpflug based tomographer, the Pentacam HR (Oculus, Wetzlar, Germany), called "ABCD progression display" (Belin et al., 2020) Despite of these definitions, progression of the disease in young patients (younger than 18) is much higher and more rapid than in adults (Chatzis and Hafezi, 2012;Ferdi et al., 2022).This should be considered in clinical practice when young patients are presented for the first time, e.g., to re-examine them at a shorter follow-up interval between 1 and 3 months in order to detect the progression.This is also appropriate in patients who have a lot of stress.In a study by Lenk et al. higher hair cortisol levels were found in progressive KC, indicating higher stress than in stable KC (Lenk et al., 2017).

Medical history
It is important to assess the history of patients with keratoconus so to identify whether the patient falls into a high-risk group where rapid progression should be expected within a short period of time and focus on very close follow-ups, or whether rapid progression is not to be expected, i.e. the patient is classified as 'low risk', where regular annual follow-ups are sufficient.
The following parameters should be considered when assessing the patient's medical history during the first time visit and classifying him/ her as high or low risk for progression of corneal ectasia: age, sex, regular medication, allergies, pregnancy, sports, hobbies, and smoking habits.
The Italian group found that in the period up to the age of 18 years the ratio of male to female patients is 6:1 and that at this age, keratoconus is more aggressive and the possibility of progression is higher than in other age groups (Caporossi et al., 2011).In patients with atopic eczema, there is a tendency for rapid progression of ectasia, which should be "additionally" facilitated by prolonged use of steroids, which alter the biomechanical properties of the cornea and provide a reduction in corneal strength by their effect (Spoerl et al., 2009).Regular use of other steroids in patients with chronic systemic inflammatory diseases or regular intake of hormones such as estrogen (contraceptives) and anabolics (professional athletes, bodybuilders) may trigger progression of ectasia in patients predisposed to the corneal disease (Spoerl et al., 2007b).Hormonal changes during pregnancy can negatively affect corneal biomechanics; therefore, pregnant women diagnosed with keratoconus or after corneal refractive surgery should be carefully monitored, and in the event of progression, the patient should undergo CXL after delivery (Bilgihan et al., 2011).In our institution, we do not perform CXL during pregnancy because of the potential for postoperative complications (e.g., infection, persistent corneal epithelial defect) that would require systemic therapy or adjunctive procedures requiring general anesthesia.We should inform female patients that despite performing CXL, further progression of ectasia may occur due to changes in hormonal levels during the next possible pregnancy (Hafezi and Iseli, 2008).
Certain sports and leisure time activities leading to regular prolonged elevation of intraocular pressure (bodybuilding or weightlifting, yoga with head-down positions, playing wind instruments) may also be a risk factor for progression of existing ectatic corneal disease.On the other hand, it could be noted that there are also some general conditions or habits of the patient that induce natural crosslinking in the tissues, such as diabetes mellitus or smoking, and patients with these conditions or habits are classified in a group in which rapid progression of keratoconus is not expected or no progression at all, and in these cases, close follow-up is not necessary.The protective effect of manifest diabetes may be explained by the induction of crosslinking in the stroma, which prevents biomechanical weakening of the cornea (Kuo et al., 2006;Seiler et al., 2000).A different finding was reported by Danish authors in a large registry-based study who found no statistically significant association between keratoconus and diabetes.However, they did find that these patients had a higher risk of allergic rhinitis, asthma, atopic dermatitis and depression (Bak-Nielsen et al., 2019b).Cigarette smoking also has an effect on keratoconus progression.Stiffening in the skin and blood vessels is observed in smokers (Mahmud and Feely, 2003;Morita, 2007) and the toxic substances in cigarette smoke can induce chemical crosslinking of the cornea.Recently, a university-based study investigated the effects of smoking and alcohol consumption on biomechanical response of the cornea using air-puff tonometry (Corvis ST).An association between smoking and a stiffer corneal response was found (Liu et al., 2023).Because of the multiple health risks of tobacco smoking, cigarette smoke exposure cannot be recommended for patients with keratoconus, and the relationship between cigarette smoking and keratoconus must not be misused to make treatment suggestions or recommendations for irreversible side effects and extensive organ damage (Spoerl et al., 2008).

Table 2
Overview of randomized controlled studies concerning the study design, main outcomes, and conclusions.

Clinical trials
In the last fifteen years, several studies have been conducted in various corneal centers in Australia, Europe and the USA to provide clinical data supporting the efficacy and safety of CXL in patients with keratoconus.The pilot study of Wollensak from Dresden was conducted as prospective non-randomized trial.Patients were treated when a progression based on medical history was present (Wollensak et al., 2003a).Therefore, the main outcome of this study was the halt of progression, which was finally achieved in all eyes, with an observed reduction of the maximal keratometry (K max) reading by 2.01 D. Due to an initial progression of the disease, the observed success cannot really be associated with the phenomena of the regression to the mean.However, addressing the problem, several randomized controlled trials were conducted in Australia, Germany, the U.K., and the U.S. The design of those studies enabled the assessment of efficacy and safety of the procedure.

Results of randomized controlled trials
An overview of the main randomized controlled trials using the standard protocol are given in Table 2.The study by O'Brart et al. demonstrated the efficacy of CXL in comparison to untreated contralateral eyes of the same patient for the first time (O'Brart et al., 2011).In the treated eyes, the K mean and the K value at the cone decreased statistically significantly by − 0.6 D and − 0.9 D, respectively, whereas no change was observed in the untreated eyes after 18 months follow-up.Corneal thickness and visual acuity did not change between the groups (O'Brart et al., 2011).Another study from Australia showed the efficacy of CXL over a follow-up of 36 months between 50 treated and 50 untreated eyes (Wittig-Silva et al., 2014b).The topography findings of this study reveal a statistically significant decrease of K max by − 1.03 ± 0.19 D in the CXL group and an increase by +1.75 ± 0.38 D in the control group.Corneal thickness remained unchanged when measured with ultrasound pachymetry, however, a lower thickness was measured in both groups which was more pronounced between 3 and 12 months in the CXL group.Uncorrected (UCVA) and best-corrected visual acuity (BCVA) improved in the CXL group, where the control group deteriorated or remained unchanged (Wittig-Silva et al., 2008).Lang et al. reported the lowest alterations (K max decreased by − 0.35 ± 0.58 D per year) in topography after CXL but these were different between treated and untreated eyes.They found no differences in corneal thickness and visual acuity (Lang et al., 2015).The largest trial was conducted by Hersh et al. andled to the FDA approval in 2017 (Hersh et al., 2017b).The authors found a decrease of K max by − 1.6 ± 4.2 D and an increase by +1.0 ± 5.1 D in treated and untreated eyes, respectively.Moreover, the best corrected visual acuity had improved more in the CXL group than in the control group (Hersh et al., 2017b).A recent study investigated the treatment efficacy in young patients aged between 10 and 16 years using an accelerated protocol (9 mW/cm 2 for 10 min) (Larkin et al., 2021).The treatment group was superior to the control group by a 3.0 D lower keratometry value of the steepest meridian (K2).The visual acuity improved in the treatment group compared to the control group as well (Larkin et al., 2021).Overall, all randomized controlled studies reveal superiority of the CXL treatment compared to the untreated eyes regarding efficacy and safety.In addition, progression occurred in 14%-52% of eyes in the control groups within the study period, while re-progression ranged from 0% to 7% in the treatment group, leading to the conclusion that CXL is effective in halting further progression of keratoconus.

Results of non-randomized longitudinal studies
In 2008, our group described the largest published series to date, involving 241 eyes in 130 patients who were followed in Dresden for up to 6 years after CXL.This uncontrolled, retrospective study confirmed previous findings of statistically significant improvement in astigmatism, BCVA, and K max (Raiskup-Wolf et al., 2008).Reports from several other centers have yielded similar results.In 2009, Vinciguerra et al. conducted a prospective, nonrandomized, monocentric clinical trial and described improvements in UCVA, BCVA, and a reduction in corneal and total wavefront aberrations at 1 year postoperatively (Vinciguerra et al., 2009).The prospective, nonrandomized, open-label Siena Eye Cross Study analyzed outcomes in 44 eyes and confirmed stability of keratoconus after CXL with no relevant side effects for up to 60 months after surgery (Caporossi et al., 2010).In another prospective study, significant improvement in topographic and aberrometric parameters was observed between 7 years of follow-up and preoperative values, indicating corneal stability persisted up to 7 years after corneal crosslinking (O'Brart et al., 2015).
Study results of at least 10 years show promising long-term efficacy of the standard CXL protocol in progressive keratoconus.Vinciguerra et al. published data of their study cohort with a follow time of up to 13 years and found that the treatment is safe and effective (Vinciguerra et al., 2020).For the corneal topographical and tomographical results, the authors found a statistically unchanged K max value over the study period demonstrating a stabilization of the disease.Instead, there was even a reduction in an index representing the anterior curvature in a 3-mm area around the thinnest point, indicating an improvement in the steeper area of the cornea (Vinciguerra et al., 2020).In contrast, another study from Romania confirmed a continuous flattening of central keratometry values up to 10 years (Nicula et al., 2019).Moreover, UCVA, BCVA, refractive cylinder, and spherical equivalent improved significantly as well.Furthermore, a reduction of central keratometry (K1 and K2) was found to be significant after 10 years (Salman et al., 2022).
In our retrospective study including the patients of the pilot study from Wollensak et al., we found a decrease of the K max from 61.5 D to 55.3 D after 10 years, which was statistically significant.Central keratometry values also decreased statistically significantly by 3.6 D and 2.0 D for K1 and K2, respectively.The mean change of BCVA was 0.14 logMAR indicating a statistically significant improvement by 1 line (Raiskup et al., 2015).
A direct comparison between the studies is difficult as various measuring devices were used and different study cohorts were investigated.Nowadays it is known that changes in corneal morphology induced by CXL depend on several preoperative factors such as the severity of the disease (Koller et al., 2011) as well as age (Alenezi et al., 2022).Studies have shown that corneal flattening occurred stronger in advanced cases (Koller et al., 2011;Padmanabhan et al., 2021;Sloot et al., 2013).
In a recent study by our group, we demonstrated long-term stabilization of progressive keratoconus, even after 15 years with the advantage of a minor loss to follow-up (Raiskup et al., 2023).The results showed significantly reduced K mean and K max values (Fig. 11) as well Fig. 11.Long-term efficacy of S-CXL concerning keratometry readings (Raiskup et al., 2023).
as an improved BCVA with a simultaneous reduction of corneal astigmatism.However, neither tomography nor visual acuity changed between 10 and 15 years.These findings suggest that an ongoing remodeling process, such as corneal flattening, may not be endless.A re-progression (increase of K max of 1 D) was found in only 8 % of eyes, however, this did not proceed within one year and no re-treatment was necessary (Raiskup et al., 2023).We assume that the stabilization process was supported by the natural stiffening of the cornea (Knox Cartwright et al., 2011;Seiler et al., 2019b) due to only adults being treated at the beginning of the procedure.In summary, the experiences from the 15 years of data confirm the long-term stability of CXL without major complications or the need for subsequent re-interventions such as keratoplasty.A cost-effectiveness analysis of the CXL procedure showed a lower financial cost to the national healthcare system compared to corneal transplantation.Proven 15 year stability confirmed high cost-effectiveness of this procedure in progressing KC (Godefrooij et al., 2017c).

Standard CXL versus accelerated CXL
Accelerated CXL protocols are based on the Bunsen-Roscoe law and aim to shorten the treatment time by increasing UV-light intensity, while not exceeding a total energy of 5,4 J.The most common UV-light intensities are 9 mW/cm 2 (for 10 min), 18 mW/cm 2 (for 5 min), and 30 mW/cm 2 (for 3 min).There are some prospective clinical trials that have investigated the efficacy and safety between S-CXL and A-CXL(9*10) protocols.In these studies, the outcomes were similar in terms of BCVA, keratometry readings, and corneal thickness after 12 months (Cummings et al., 2016;Hagem et al., 2017;Sadoughi et al., 2018;Shetty et al., 2015).This indicates that the A-CXL protocol with an intensity of 9 mW/cm 2 for 10 min is as effective as the S-CXL protocol, which is confirmed by the experimental results of a similar stiffening effect compared to untreated eyes (Bao et al., 2018;Hammer et al., 2014;Herber et al., 2020a).In their experimental study, Hammer et al. showed that the A-CXL(18*5) has no significant stiffening effect in comparison with untreated eyes, indicating that the Bunsen-Roscoe law cannot be fully applied to the CXL process because of influencing factors such as oxygen, riboflavin, and the saturation or the availability of free crosslinking sites (Fig. 1).Shetty et al. demonstrated a corneal flattening (K flat and K steep) effect after 12 months for S-CXL and A-CXL(9*10) as well as for A-CXL(18*5) only in K steep.The amount of the reduction in keratometry reading was more pronounced for S-CXL (K flat and K steep) and A-CXL(9*10, only K flat) compared to shorter treatment protocols with 5 min (18 mW/cm 2 ) or 3 min (30 mW/cm 2 ).The authors also measured a deeper demarcation line in S-CXL and A-CXL(9*10) than the for other protocols (Shetty et al., 2015).However, very fast protocols (≤5 min) are non-inferior in terms of halting progression after 12 months (Lang et al., 2019;Shetty et al., 2015;Toker et al., 2017) and longer (Hashemi et al., 2015;Kang et al., 2020)."The Siena Eye-Cross Study 2" assessed the 5-year results of epi-off A-CXL(9*10) and demonstrated statistically significant and stable improvements in UCVA and CDVA, K-values, and corneal higher-order aberrations over this long-term follow-up period (Mazzotta et al., 2021b).
Combining those studies in meta-analysis, the S-CXL protocol leads to a greater reduction in keratometry readings, a greater reduction of corneal thickness, and a deeper demarcation line in comparison to A-CXL protocols (Shajari et al., 2019;Wen et al., 2018).Overall, both methods similarly stopped the disease progression (Kobashi and Tsubota, 2020;Shajari et al., 2019;Wen et al., 2018).

CXL in thin cornea with keratoconus
The criteria for CXL, which require a minimum corneal thickness of 400 μm after epithelial removal, are primarily intended to protect the endothelium.To ensure this unavoidable condition, an intraoperative pachymetric measurement during each procedure is essential.However, in advanced keratoconus, stromal thickness is often less than 400 μm due to progressive corneal thinning.In the case of patients whose good best-corrected vision (e.g. with contact lenses) is maintained, allowing them to manage their daily activities at an adequate quality level without significant limitations, we could modify the standard treatment protocol and preoperatively swell the thin cornea to a stromal thickness of at least 400 μm using a hypoosmolar riboflavin solution.In 2009, Hafezi et al. published the first study in which thin corneas were treated with a hypo-osmolar riboflavin-dextran-free solution to swell the corneal stroma to a minimum thickness of 400 μm.Standardization has not been possible because the transient swelling effect varies widely from cornea to cornea and stromal hydration causes separation of collagen fibrils available for CXL with reduced oxygen diffusion (Hafezi et al., 2009).However, this treatment modification represents a safe extension of the indication spectrum for CXL to thin corneas that would not otherwise be considered for treatment.The biomechanical stiffening effect is comparable to that of a 400 μm cornea (Wollensak and Sporl, 2019) because the anterior stroma swells only slightly, whereas the posterior stroma swells significantly (Muller et al., 2001).Accordingly, the localized treatment effect of CXL in the anterior stroma (approximately 200 μm) is not affected by the posterior swelling effect of the hypoosmolar riboflavin solution.
The results of our study using a hypoosmolar riboflavin solution to crosslink thin corneas demonstrated stability of keratoconus one year after CXL based on BCVA and K max.Application of a hypoosmolar riboflavin solution protected the crosslinked cornea from stromal scar formation (Raiskup and Spoerl, 2011).
However, even this modification has limitations, as reported by Hafezi in a case report describing CXL failure after preoperative stromal swelling with hypoosmolar riboflavin solution in an extremely thin cornea (268 μm) (Hafezi, 2011).Although swelling of the corneal stroma was effective and resulted in excess thickness over 400 μm before radiation application, the increase in biomechanical resistance was not sufficient to halt disease progression (Hafezi, 2011).
Other modalities, according to Jacob et al., were contact lensassisted CXL (CA-CXL), in which an isoosmolar riboflavin-soaked contact lens is used to artificially "thicken" the cornea, although the contact lens may reduce oxygen diffusion.This technique was effective and safe in performing crosslinking in corneas less than 400 μm after epithelial abrasion and appeared to be effective based on the depth of the stromal demarcation line (Jacob et al., 2014;Mazzotta et al., 2016a).In an ex vivo study, the effect of CA-CXL on regional corneal stiffness changes was observed and compared with the S-CXL protocol using Brillouin microscopy.Both techniques induced significant corneal stiffening, mainly concentrated in the anterior cornea.CA-CXL resulted in less stiffening compared to the standard protocol (Zhang et al., 2019).Another technique described stromal expansion of thin and ultrathin corneas in KC patients using refractive stromal lenticule from patients undergoing small-incision lenticule extraction for myopic correction.The stromal lenticule was placed and spread over the host cornea after epithelial debridement in the thinnest area of the keratoconus, and the CXL procedure was performed routinely.Long-term results demonstrated safety and disease stability following this lenticule-assisted CXL (Sachdev et al., 2015(Sachdev et al., , 2024)).The "epithelial island cross-linking" was described by Mazzotta et al. in which a small circle of epithelium is left over the thinnest area as a shield.This technique may represent a further surgical development in the treatment of ectatic thin corneas, providing a safe and more effective alternative to transepithelial CXL due to its limited penetration, biomechanical and clinical efficacy (Mazzotta and Ramovecchi, 2014).However, the epithelium shields riboflavin and UV light penetration and thus causes an unequal demarcation line between epithelialized and de-epithelialized areas.
Although this technique has achieved good clinical results, the addition of contact lenses or refractive stromal lenticule is not always accessible to surgeons and makes the procedure more complex and less economical.Therefore, another possibility to provide treatment of thin corneas is the reduction of the total energy (fluence) by adjusting the time below the Bunsen-Roscoe equivalent, e.g. 3 mW/cm 2 for 24 min F. Raiskup et al. when stromal thickness is 360 μm (Fig. 12).The treatment time can also be calculated in dependence of stromal thickness by an online calculator (https://jscalc.io/calc/VmanUJD6yQ13VQQ6)(Shetty et al., 2020).
The team of scientists from Switzerland grouped under Prof. Hafezi newly developed the "Sub400 Protocol" for thin corneas, where an individualized fluence, adjusted to the corneal thickness, guarantees a safety distance of 70 μm between the demarcation line and the endothelium (Hafezi et al., 2021a).This protocol individualizes the total fluence delivered to each cornea based on its minimum thickness, with the fluence decreasing as the cornea thickness decreases.Mazzotta et al. compiled an "M-nomogram" from in vivo confocal microscopy and OCT measurements from which the depth of the treatment zone for various treatment protocols can be determined (Mazzotta et al., 2018b).This protocol is a pachymetry-based, individualized, and accelerated approach that predicts the depth of the demarcation line based on the minimum preoperative corneal thickness.Regardless of the baseline minimum corneal thickness, this method maintains a constant fluence of 5.4J/cm 2 as in the standard protocol and adjusts the UV-A power from 9 mW/cm 2 to 30 mW/cm 2 and the exposure time from 10 to 6 min, by using continuous and pulsed UV-light mode of exposure.Comparison of the demarcation line depths showed a high correlation between the measured and calculated depths based on the mathematical models that are the mainstay of the M-nomogram protocol for "all thickness" ectatic corneas.The novelty of these approaches is that thin ectatic corneas in the range of 200-400 μm can be crosslinked by epi-off treatment with a higher chance of success and a high level of safety (Mazzotta et al., 2023b).

Transepithelial CXL
S-CXL or epithelium-off techniques are technically simple and safe procedures, but epithelial debridement can cause severe pain and vision loss during the first few days after surgery until the epithelium has completely recovered.Infectious complications and cases of persistent corneal epithelial defect, even melting with perforations have been reported in association with epithelial removal and the use of a bandage contact lens (Kymionis et al., 2007b;Pollhammer and Cursiefen, 2009;Rama et al., 2009;Zamora and Males, 2009).Thus, the need for epithelium-on procedures has increased in the past.
The animal experimental studies that are mentioned in 2.5.1 revealed good results concerning the biomechanical stiffening induced by TE-CXL using enhancers such as BAC and EDTA.Based on these results, various clinical studies have been conducted by maintaining the S-CXL parameters (riboflavin solution using dextran and a total fluence of 5,4 J/cm 2 ) as well as adding the penetration enhancers (BAC and EDTA).The first studies reported inconsistent results.Filippello et al. demonstrated the TE-CXL effect in a bilateral study, in which the contralateral progressive eye was not treated and used as a control.The treated eyes showed improvements of UVCA, BCVA, keratometry reading and highorder aberrations.In contrast, the untreated eyes worsened during the follow-up period (Filippello et al., 2012).Conversely, other studies reported a lower treatment effect compared to S-CXL (Caporossi et al., 2013;Koppen et al., 2012;Leccisotti and Islam, 2010).In particular, the studies by Caporossi et al. and Koppen et al. even showed progression after the TE-CXL in their study cohort for K max and thinnest pachymetry, after 12 and 24 months (Caporossi et al., 2013;Koppen et al., 2012).TE-CXL was also found to be appropriate for younger patients to avoid pain during and after surgery.However, corneal tomography parameters deteriorated in 50% of younger patients (Caporossi et al., 2013).In addition, a randomized controlled study confirmed the lower effectiveness of TE-CXL in comparison to S-CXL, where the flattening effect of K max occurred statistically significantly stronger in the S-CXL than in the TE-CXL group.The re-progression was 23% in the TE-CXL compared to 0% in the S-CXL group (Soeters et al., 2015).
The insufficient results led to a further improvement of the TE-CXL procedure, modifying the riboflavin solution by increasing the concentration to 0.25 % (0.01% BAC) and excluding dextran.In an international multicenter study, in which our institute participated as well, Gatzioufas et al. investigated the success of this treatment (also the intensity was changed to 9 mW/cm 2 for 10 min, same total fluence of 5.4 J/cm 2 ) and found no statistical improvements in K max, UCVA, or BCVA.Moreover, 46% of the treated eyes reprogressed by more than 1 D within 12 months postoperatively, indicating an ineffective treatment option (Gatzioufas et al., 2016).The authors suspected two reasons for the failure of their TE-CXL treatment.Firstly, the epithelial cell breakdown caused by BAC was not high enough to allow riboflavin to adequately penetrate into the stroma.Secondly, the epithelial cells also provided a barrier to the oxygen, which again inhibits diffusion into the stroma (Gatzioufas et al., 2016).This could be concluded from the appearance of the demarcation line depth after using the treatment modality.The demarcation line was much shallower (135 μm from the epithelial surface) than in epithelium off procedures (Caporossi et al., 2012b;Spadea et al., 2018).These findings indicated that cross-link formation by this epithelium-on procedure was likely to occur in the upper third of the corneal stroma as opposed to the "standard" procedure where collagen cross-links were found in much deeper layers of the stroma.
Later on, oxygen was supplemented during the irradiation of UV light to compensate the limited oxygen diffusion by the epithelium with special goggles that over-saturate the environment of the eye.Mazzotta et al. demonstrated the efficacy of TE-CXL by supplementing oxygen and with a customized irradiation pattern as well as pulsed UV-light (Mazzotta et al., 2020b).The authors found a flattening effect of the cornea, an improvement in high-order aberrations and visual acuity, leading to the conclusion that the supplementation of oxygen overcame the problem of the epithelial barrier.Additionally, the overall depth of the demarcation line was deeper than a comparable epi-off treatment with an equal UV fluence, indicating a potentially deeper treatment volume (Mazzotta et al., 2020b).
Another important step in the further development of the TE-CXL procedure was the use of an enhanced fluence (7 J/cm 2 ) to overcome the limiting factor of the UVA radiation absorption by the epithelium (Mazzotta et al., 2020a).This, in combination with iontophoresis-assisted imbibition of riboflavin enhancing the diffusion into the stroma by the electric device, is the current protocol of TE-CXL.Long-term studies conducted by experienced Italian researcher teams from Siena and Milano have confirmed the efficacy of this treatment modality with lower rates of treatment failures (up to 26 %) than reported in the beginning of TE-CXL (Mazzotta et al., 2020a;Vinciguerra et al., 2022).The results of novel TE-CXL protocols showed clinically acceptable success rates in halting the progression of keratoconus and therefore might be a promising option compared to epi-off protocols.
The use of iontophoresis for riboflavin loading during CXL requires additional equipment, which is costly, and a minimally invasive approach due to the need for a suction ring requires good patient cooperation.Enhanced fluence pulsed light without iontophoresis and without supplemental oxygen (EFPL-M-TE-CXL) uses new generation biphasic chemically enhanced 0.25 and 0.22% hypotonic riboflavin solutions for a 10-min soaking time, dextran free, with trometamol (TRIS) and EDTA.Only 21% room air oxygen is used in this procedure.The postoperative spectral-domain corneal OCT performed one month after treatment showed a mean demarcation line depth at 272 ± 43 μm.
The results are is similar to TE-CXL with iontophoresis, but it is less invasive and may therefore be better indicated in poorly compliant patients (Mazzotta et al., 2022(Mazzotta et al., , 2023b)).

CXL in children
Childhood age at the time of KC diagnosis is a negative prognostic factor in terms of rapid keratoconus progression and is associated with an increased likelihood of requiring corneal transplantation.Younger patients in particular, represent a population at high risk for more rapid disease progression (Reeves et al., 2005).The "Siena CXL Pediatrics" study analyzed findings in 152 progressive keratoconus patients aged between 10 and 18 years (Caporossi et al., 2012a).In this study, the authors found a male-to-female (M:F) incidence ratio of 4:1, which differs from the epidemiologic findings in the literature of 2:1 (Ertan and Muftuoglu, 2008;Rabinowitz, 1998).The results of this study demonstrated significant and rapid functional improvement in pediatric patients with progressive keratoconus after undergoing CXL.No severe adverse effects were recorded.This study demonstrated the efficacy of CXL in terms of slowing keratoconus progression in pediatric patients and improving functional performance in 80% of patients up to 3 years after the procedure, regardless of preoperative corneal thickness.Deterioration of functional results was observed in 4.6% of patients during follow-up, which was rationalized by the higher aggressiveness and progression of keratoconus in this fragile population (Caporossi et al., 2012a).
Later on, the same group demonstrated the long-term efficacy of S-CXL (Siena protocol, a modified version of the S-CXL) in a pediatric cohort for a duration of up to 10 years.The main outcome was that the CXL effect remained successfully stable in corneal topography and tomography up to 7 years.In the period between 7 and 10 years, the reprogression rate increased to 21% of the treated eyes (Mazzotta et al., 2018c).The authors assumed a loss in efficacy due to the corneal collagen turnover resuming in a renewed instability of the cornea.Other long-term studies reported an improvement in visual acuity and keratometry reading without severe adverse events that led to a halt of the diseases progression.However, they also showed a re-progression rate of approximately 22 % (Godefrooij et al., 2016b;Henriquez et al., 2018) These results suggest that CXL should be the first-choice therapy for the treatment of progressive keratoconus in pediatric patients, as the disease is more aggressive in this age group.In addition, there should be intensive follow-ups performed on these patients to detect an early reprogression and avoid a loss of vision.

Complications after CXL
All of the above studies demonstrate that CXL is an effective treatment option for halting the progression of keratectasia and stabilizing the architecture of the corneal stroma.The treatment is also considered to be a safe procedure, however, side effects and complications may occur immediately after the operation, mainly due to the removal of the epithelium.
Koller et al. conducted a prospective study involving the treatment of 117 eyes in 99 patients with primary keratectasia and evaluated the rate of complications and CXL failure during the first postoperative year (Koller et al., 2009b).The re-progression of keratectasia was verified by repeated Scheimpflug imaging over a minimum of 6 months (range, months to 2 years) with an observed increase in Kmax of more than 1 D. The complication rate, which was defined as the percentage of eyes with a loss of 2 or more lines on the Snellen chart at a corrected visual acuity examination during the 1-year follow-up compared with the preoperative condition, was 2.9 %.An age of more than 35 years and preoperative BCVA better than 20/25 were found to be significant risk factors for complications.No morphological or optical reasons for vision loss could be established in this study.The failure rate, which was defined as the percentage of eyes with an increase in Kmax of more than 1 D from the preoperative value, was 7.6 %.A high preoperative Kmax value was determined to be a significant risk factor for failure.Sterile infiltrates were observed in 7.6% of eyes and central stromal scarring in 2.8% of eyes (Koller et al., 2009b).
Our group analyzed the development of permanent stromal scarring after CXL in a retrospective study (Raiskup et al., 2009).The cohort included 163 eyes of 127 patients with Amsler-Krumeich stage 1-3.At year after CXL, 8.6% of the treated eyes developed significant corneal scarring.The development of scarring occurred mainly in eyes characterized by high Kmax values (mean 71.1 ± 13.2 D) and thin corneas (mean 420.0 ± 33.9 μm).These results allowed us to postulate that advanced keratoconus should be considered as a stage of the disease with an increased risk of corneal scarring after CXL due to low corneal thickness and high corneal curvature (Raiskup et al., 2009).
In a recent retrospective study, our team revealed that S-CXL and A-CXL(9*10) does not differ regarding the failure rate over a time period of 36 months (Lenk et al., 2021).We investigated 230 eyes of 173 patients that underwent either S-CXL or A-CXL(9*10).The success rates were F. Raiskup et al. 92.5% and 86.4% for S-CXL and A-CXL(9*10), respectively (Fig. 13), which was in line with the results of Koller et al. (92.4 %) using similar criteria for re-progression (Koller et al., 2009b).We detected higher preoperative Kmax values (cut-off value was 60.4 D) and the presence of neurodermatitis combined with other atopic diseases as significant risk factors for progression of keratoconus after CXL (Lenk et al., 2021).

Effect of CXL on corneal transplantation metrics
Penetrating keratoplasty is by far the most common tissue transplantation in humans worldwide.Data from the Australian Corneal Transplant Registry, one of the largest corneal transplant registries in the world, show that 91% of grafts successfully survive one year postoperatively and 74% survive five years postoperatively, for all diagnoses typically indicated for corneal transplantation.The success rate of corneal transplantation has been attributed to advances in microsurgical techniques, development of eye banks, corneal preservation techniques, and surgeons' ability to recognize and treat postoperative complications (Mendes et al., 2003).However, many patients with successful (clear) corneal transplantation may have low visual acuity postoperatively due to irregular astigmatism or other complications.The success of keratoplasty is often evaluated by visual acuity, corneal graft clarity, and keratometry.While these tests are useful in measuring important parameters of visual function, they do not always accurately correlate with the patient's ability to perform normal activities of daily living.The surgeon's definition of a successful corneal graft may not accurately reflect how the patient perceives success, as some patients report that they are dissatisfied with their visual acuity and are limited in their "activities of daily living" despite good CDVA and a clear graft.These outcomes can often be attributed to different (realistic or unrealistic) postoperative expectations, inadequate preoperative counseling, and/or limitations in patients' ability to adapt (Mendes et al., 2003).Patients with KC have the best prognosis after corneal transplantation (Mendes et al., 2003); therefore, KC was the cause of 15.1% of the more than 30, 000 corneal transplants performed in the United States in 2004 (Eye-BankAssociationofAmerica, 2005).However, sometimes corneal transplantation is not the best option because patients with this disease are often young and very active, which affects their career choices, ability to work, leisure time, and overall quality of life.Surgical treatment, especially corneal transplantation, is a complicated, time-consuming, and also economically demanding procedure that places a great burden on affected patients.Due to the limited lifespan of the transplanted tissue and potential complications such as tissue rejection or otherwise unsuccessful grafting, CXL represents a promising treatment alternative for this ectatic corneal pathology.Corneal collagen cross-linking has clear clinical, economic, and psychological advantages for this indication: it is associated with fewer postoperative complications, can be performed on an outpatient basis, is minimally invasive, and is economically cost-effective.The lifelong economic burden on KC patients and the healthcare system is not negligible.With an average disease duration of 37 years, treating a keratoconus patient costs US$24, 168 more than treating a patient with myopia, for example.More than half of this cost is the estimated cost of corneal transplantation and treatment of associated postoperative complications.In addition, the expected cost of routine ophthalmic or optometric care (regular visits and nonsurgical treatment) in the United States is 42.3% of the total cost of care (regular care and surgical treatment) (Rebenitsch et al., 2011).Because KC usually manifests in adolescents and young adults, refractive errors in the younger age group should also be considered.The annual cost of refractive care for all young adults in the United States is US$3.6 billion.The average annual cost for routine individual vision care is US $200, but for keratoconus patients it is US$653 (Rebenitsch et al., 2011).In the context of the economic impact of other diseases associated with visual impairment, the economic impact of major adult vision disorders, such as age-related macular degeneration, cataract, diabetic retinopathy, glaucoma, and refractive error, was estimated at US$51.4 billion in the United States, in 2007 (Rein et al., 2006).The introduction of the CXL procedure into the standard treatment algorithm for progressive keratoconus has made it possible to reduce the need for corneal transplantation for this condition (Godefrooij et al., 2016a;Sandvik et al., 2015).This fact leads to a better redistribution of corneal transplants for other corneal pathologies.The economic impact of introducing this treatment is also important.The fact that it is a relatively simple outpatient procedure, the efficacy of which we have been able to confirm in our long-term results, leading to a reduction in the need for corneal transplants, reduces the financial costs for patients with this diagnosis, which has a significant positive economic impact on the national healthcare system (Godefrooij et al., 2017c;Leung et al., 2017;Lindstrom et al., 2021).Our recently published longitudinal 15-year results of this procedure have only confirmed its high Fig. 13.Kaplan-Meier analysis of S-CXL and A-CXL over a time period of 36 months (Lenk et al., 2021).

Combining CXL with refractive interventions to improve vision
Corneal cross-linking has been shown as an effective treatment for keratoconus by halting the progression of the disease in pediatric and adult population.Many studies have reported significant improvements in keratometry readings and visual acuity.However, the improvement of visual acuity was not better than one line after S-CXL and it was also not predictable (Wittig-Silva et al., 2014a).Apart from the decrease in Kmax values by more than 1 D, high-order aberrations also improve, although these have no effect on visual acuity (Godefrooij et al., 2017b;Greenstein et al., 2012a).
A promising method for a specific improvement of vision is the combination of CXL and photorefractive keratectomy (PRK).Both procedures can be performed simultaneously, and is named as which is called "CXL+" (Hafezi et al., 2023).Such a procedure was first described by Kanellopoulos and Binder, the so-called Athens protocol (Kanellopoulos and Binder, 2011).The long-term results (10 years) showed that CXL+ is a safe and effective treatment option that stabilizes the ectatic cornea and improves visual acuity (Kanellopoulos, 2019).However, ablation of corneal tissue by the laser leads to a further reduction in corneal thickness, which can have a negative effect on corneal stability (Kanellopoulos, 2019).Comparing the visual outcomes of both CXL and CXL+, statistically significant improvements were found for CXL+ (Alessio et al., 2013;Iselin et al., 2020).In addition, other studies reported similar results of the CXL + treatment (Ahmet et al., 2018;Gore et al., 2018;Rechichi et al., 2021).Fig. 14 presents a 1-year postoperative follow-up of a patient, who receives a CXL + procedure without correcting sphere or astigmatism.
To avoid a large reduction in corneal thickness due to laser ablation of the refractive profile, treatment planning should be optimized to correct only higher order aberrations and regularize the corneal surface.The goal of this procedure is to improve visual acuity using spectacle lenses or soft contact lenses if the patient cannot tolerate rigid gas permeable lenses.In transepithelial photorefractive keratectomy (transPRK), the actual thickness of the epithelium can be changed as a treatment parameter because it is often reduced in the apical region of the cone (Pircher et al., 2020;Rocha et al., 2013).The epithelium thickness can be measured by novel OCT tomographers for the anterior segment of the eye (Vega-Estrada et al., 2019).As a general rule, no more than 50 μm should be ablated in the central area, where the cornea is the thinnest (Kymionis et al., 2009a).
The introduction of ray-tracing technology in the treatment planning allowed the calculation of the refractive contribution of the posterior corneal surface and epithelium, thus avoiding undesirable overcorrection often reported in the literature with topography-guided techniques, and allowed to reduce the stromal tissue ablation (Mazzotta et al., 2024).The refractive outcome was an improvement in UCVA and BCVA of more than two lines in 95% and 100% of the cases, respectively, with no reported loss of visual acuity or re-progression of KC, indicating superior treatment compared to other protocols.A gain of more than 4 lines of BCVA was reported in 68% of cases (Mazzotta et al., 2024) It enabled to overcome the inability to calculate more precisely the refractive contribution of the posterior corneal surface in topography-guided excimer laser treatments by ray-tracing-guided customized ablations improving functional results, sparing stromal tissue and regularizing the stromal gradient as a new concept of a beneficial effect of "selective corneal remodeling" in irregular keratoconic corneas (Mazzotta et al., 2024;Roberts et al., 2024).Corneal remodeling in terms of regulating the corneal curvature gradient could interrupt the cycle of ectasia decompensation by redistributing the biomechanical stress and decreasing the difference between maximum and minimum stress, shifting from the apical region to the periphery (Roberts et al.,Fig. 14.Upper image: Sagittal topography of an 18 years old patient receiving CXL + for the correction of higher order aberrations preoperatively and up to 12 months.The BCVA improved from 0.6 preoperatively to 0.7 postoperatively after 12 months.Root mean square of total aberrations and higher order aberrations improved from 9.43 to 3.34 μm and from 2.24 to 1.04 μm, respectively.Corneal thickness was reduced by a maximum of 25 μm in the area of the thinnest point (lower image).
F. Raiskup et al. 2024).This approach still needs to be investigated in the future.
In addition, recent developments in corneal tomography and corneal biomechanics have made it possible to detect ectasia at a very early stage, allowing better follow-up of patients with suspicious corneas and earlier detection of disease progression (Ambrosio et al., 2017(Ambrosio et al., , 2022;;Herber et al., 2022a;Vinciguerra et al., 2016a).The earlier the ectasia and/or progression is detected, the better is the use of the CXL + procedure (Hafezi et al., 2023).

Enhanced fluence and customized irradiation patterns
To also achieve the crosslinking effect in deeper layers of the cornea (deeper demarcation lines), the dose was increased via heightened irradiance, i.e. the irradiation time was prolonged.Mazzotta et al. increased the fluence from 5.4 J/cm 2 to 7 J/cm 2 by increasing the time to 6.28 min using an intensity of 18 mW/cm 2 .Finally, they used pulsed irradiation, therefore, the total time was 12.56 min (Mazzotta et al., 2018a).Kymionis et al. treated with 18 mW/cm 2 and a dose of 7.5 J/cm 2 and found a similarly deep demarcation line to the S-CXL, indicating a comparable CXL effect with the use of a shorter time.No complications such as endothelium damage were observed (Kymionis et al., 2016).At a higher irradiance, the extended irradiation time allows more oxygen to diffuse into deeper layers where it is available for crosslinking.However, irradiance higher than 3 mW/cm 2 must be applied so that sufficient UV energy is still available in deeper layers, otherwise no improvement is obtained.This approach is essential for customized CXL, in which the cornea is irradiated with individual patterns with varying fluence, resulting in improved corneal flattening (Seiler et al., 2016).Various irradiation patterns of individualized CXL have been investigated, such as concentric circles (Sachdev et al., 2021;Seiler et al., 2016;Shetty et al., 2017) or sectors (Cassagne et al., 2017;Nordstrom et al., 2017) based on the individual topography of the patients.In all studies, the flattening effect was more pronounced compared to uniform irradiation (S-CXL or A-CXL) and thus offers better regularization of the cornea.In addition, the higher fluence of 15 J/cm 2 in the inner area of the individual irradiation patterns did not lead to a loss of endothelial cell density or shape (Nordstrom et al., 2017).Despite the consistently good results in terms of topography changes, the refractive and visual changes for the individual patient should be viewed with caution, requiring further improvements to this technique (Nordstrom et al., 2017).
On the other hand, as mentioned before, the oxygen depletion during the irradiation is much higher with intensities above 9 mW/cm 2 , which may lead to the consequence that oxygen must be supplemented (Seiler et al., 2021).
Recent laboratory studies have proven that higher fluences increase the biomechanical impact of CXL (Fischinger et al., 2023).The progressively higher fluence protocols enhance epi-off and epi-on efficacy, as demonstrated in past and recent literature.The increased fluence variance may be the key to optimizing the photodynamic process of epi-on CXL (Mazzotta et al., 2023a).Preliminary data of the new pachymetry-based progressive epi-on nomogram showed good functional results (ectasia stabilization) without complications and could be considered a new paradigm of corneal photodynamic therapy for early progressive KC (Mazzotta et al., 2023a).
New horizons for the direct application of this treatment are also opening up in the field of corneal transplantation, where, for example, an ACXL can be applied manually in a deep anterior lamellar manner in a peripheral or annular "doughnut" shape to both the donor and the recipient cornea at different times to reduce peripheral keratolysis and consequent thinning with wound displacement and secondary progressive irregular astigmatism after corneal transplantation (Arafat et al., 2014;Huang et al., 2017;Ziaei et al., 2020).This procedure stabilizes the progression of the peripheral ectatic process in the recipient cornea after lamellar transplantation and also induces a cross-linking-dependent biochemical and biomechanical strengthening of the entire donor-recipient junction.This finally stabilizes the wound and prevents its late dehiscence and the relative progressive irregular astigmatism that often occurs with the slow continuous thinning of the peripheral ectatic recipient cornea.ACXL should be performed on the recipient one month prior to DALK surgery, while the peripheral donor corneas are crosslinked on the same day prior to DALK surgery (Zagari et al., 2022).In addition, CXL reduces the tendency for graft rejection since procedure-induced apoptosis (decellularization) lowers the immunogenicity "readiness" of corneal tissue (Zagari et al., 2022).In the new indication spectrum, it could be also used the lymphangioregressive effect of the CXL procedure as a pretreatment in order to promote graft survival in cases of high-risk corneal transplantations (Schaub et al., 2021;Wiedemann et al., 2024).

Conclusion
Corneal cross-linking with riboflavin and UVA radiation was proposed 25 years ago as a therapeutic approach to improve the biomechanical properties of the cornea.This method is one of the most promising breakthroughs in the treatment of progressive corneal ectasia that interferes in the ethiopathogenesis of disease itself.Currently available conservative and surgical treatment options can only temporarily correct optical visual impairment but cannot stop the development and progression of the ectatic process.Clinical studies have demonstrated that by increasing the biomechanical stability of the cornea based on CXL, the progression of corneal ectasia can be halted with very low incidence of side effects or complications.In addition to the clinical benefit, the technique has considerable economic and psychosocial advantages.CXL can be easily performed on an outpatient basis, and it is a minimally invasive, cost-effective treatment with minimal burden on patients.Currently, none of the ectatic corneal diseases are curable.However, CXL can halt their progression.Therefore, it is important to perform CXL with progressive keratoconus as early as possible.In addition, newer applications of this method, including combinations with refractive procedures such as topo-guided or wavefront-guided corneal surface excimer ablations using laser technology improve visual acuity and shorten patients' visual rehabilitation.

Financial disclosure
This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Fig. 4 .
Fig. 4. Evolution of UV light sources.(a) Self-made CXL device with two diodes (b) UV-X optimized beam light source by former IROC Innocross AG (c) clinical application of the self-made UV light diodes (d) clinical application of the UV-X device.

Fig. 8 .
Fig. 8.Comparison of anterior and posterior corneal flaps with and without treatment of CXL.Anterior flap is stiffer after CXL compared to the treated posterior flap and both untreated flaps(Kohlhaas et al., 2006).

Fig. 9 .
Fig. 9. Summary of microscopic, macroscopic, and clinical changes after CXL treatment of corneal tissue.

Fig. 12 .
Fig. 12. CXL in thin corneas, reducing irradiation time in dependence of stromal thickness before irradiation for the S-CXL and A-CXL(9*10) protocol (According to experimental calculations of Prof. Dr. E. Spoerl).

Table 1
Comparison of changes in DCR parameters in ex vivo experiments and in vivo studies induced by CXL.