Article Text

Download PDFPDF

Collagen cross-linkage: a comprehensive review and directions for future research
  1. P T Ashwin,
  2. P J McDonnell
  1. Birmingham and Midland Eye Centre, Birmingham, UK
  1. Correspondence to Pammal T Ashwin, Specialist Registrar, Birmingham and Midland Eye Centre, Dudley Road, Birmingham B18 7QH UK; pammal.ashwin{at}


Individuals with keratoconus form a significant proportion of patients for a practitioner specialising in corneal diseases. Yet it is a disease where the pathogenesis is poorly understood, and until recently, there has been no treatment apart from transplantation that could be offered that was curative or even capable of slowing the progression of the disease. Collagen cross-linking treatment using riboflavin and UV light has been developed to address this need, and the initial results are promising. The purpose of this review is to critically evaluate this treatment in light of the scientific basis for cross-linking, to highlight the strengths and limitations of the evidence in terms of efficacy and long-term safety, and finally to identify areas for future research in this area with a significant potential to change the way we treat our keratoconus patients. In addition, we hope that our unbiased review for the first time would bring together, in a concise fashion, scientific information for a practitioner contemplating on offering this treatment and to help inform their patients of its potential risks and benefits.

  • Collagen
  • cross-linking
  • cross-linkage
  • keratoconus
  • review
View Full Text

Statistics from

Keratoconus is a progressive, bilateral, often asymmetrical and non-inflammatory corneal ectasia. It is usually diagnosed during the second and third decades of life. The ectasia progresses at a variable rate, although it may progress more rapidly at a younger age. Patients have myopic astigmatism and are often suspected to have the condition by their optometrists when deterioration in visual acuity occurs that is no longer correctable by spectacles.

Treatment options for keratoconus hitherto largely involved interventions that were done for tectonic, optical or refractive purposes. Collagen cross-linking (CXL) is the latest treatment in the series that may offer some promise in that it is the only intervention that can potentially slow down the progression of disease.

The purpose of this review is to elucidate the principles involved in collagen cross-linkage, describe the technique, summarise the results of cross-linking treatment and provide direction for further research in this area. We hope that this comprehensive review would encapsulate the current evidence for the practitioner and as we do not have any competing interests would serve as a fair appraisal of the technique. We have systematically searched the Medline database, textbooks and industry literature in the preparation of this review.

Collagen cross-linkage

Chemistry of cross-linkage

The primary functional role of collagen in general is to act as a supporting tissue. Aggregated forms of the collagen monomers are strengthened by intermolecular cross-links. This process happens as a part of maturation, but also in ageing and disease.

Collagen fibrils cross-link naturally as a part of their maturation process. When these fibrils are secreted, they have short segments at either end of the collagen chains (telopeptides) that do not assume the triple-helical conformation. The hydroxylysine residues in these end chains participate in cross-links formation.1 The cross-links are formed by oxidative deamination of the ε-amino group of this single lysine or hydroxylysine in the amino and carboxy telopeptides of collagen by the enzyme lysyl oxidase. The aldehyde thus formed reacts with a specific lysine or hydroxylysine in the triple helix to form divalent bonds that link the molecules head to tail. They then spontaneously convert during maturation to trivalent cross-links.2 3

A second cross-linking pathway occurs during ageing (and to a greater extent in diabetes mellitus) involving a non-enzymatic reaction termed glycation. Prolonged exposure to monosaccharides results in a spontaneous bond between the reducing sugar and the amino group of a protein. This results in a reversible nucleophilic addition to form a Schiff's base adduct (aldimine like fursine or pyridosine). This then transforms into the more stable but still reactive Amadori product (eg, haemoglobin A1c), which goes through a number of additional slower reactions with the amino groups of other proteins, forming glucose-derived, intermolecular cross-links, such as pentosidine. These are termed advanced glycation end products.4 This was originally described as the Maillard or browning reaction in 1912 due to the associated yellow-brown colour change.5 Increased levels of pentosidine (an advanced glycation end products that forms fluorescent cross-links between arginine and lysine residues of collagen) have been found in diabetic corneas in comparison to age-matched controls.6 Increased stiffness of the cornea with age has been demonstrated by studying its stress–strain behaviour.7 An increase in the cross-sectional area of the collagen molecule from approximately 3.04 to 3.46 nm2 has been shown in human corneas with age using high-angle and low-angle x-ray diffraction pattern studies and has been attributed to age-related increase in glycation.8 Type 2 diabetes has been shown to have a protective effect on the development and progression of keratoconus possibly due to increased cross-links, which stiffen the cornea.9 10

Oxidation is a third mechanism whereby cross-links are formed in collagen. This new type of cross-linking has been identified as being distinctive to the ones formed by enzymatic and glycation in type I collagen and can occur after the process of oxidation (O3 mediated) or photo-oxidation (UV mediated).11 Outside biology, photopolymerisation is a comparable process that is being used in industry to generate polymers using the free radical-generating properties of radiant energy like UV light. Photopolymerisation of multifunctional monomers resulting in highly cross-linked materials suitable for applications as epoxy coatings, optical lenses, optical fibre coatings and dental materials are in common use. A monomer substrate in the presence of a photoinitiator can polymerise by way of cross-linkage in the presence of a UV light source.

Collagen cross-linkage with riboflavin and UVA


Historically collagen cross-linkage has been adopted for several beneficial applications. Tanning of leather involves promoting cross-linking of type I collagen present in the hide. Formaldehyde tissue fixation is based on the ability of formaldehyde to bind to amino acids lysine, arginine, tyrosine, asparagine, histidine, glutamine and serine by inducing cross-links.13 Artificial heart valves are stiffened using glutaraldehyde.

Christopher Foote in 1968 published the mechanisms by which photosensitised oxidation happens in biological systems.11 Fujimori14 in 1988 revealed a third mechanism of cross-linkage in type I collagen that involved either oxidation by ozone or photo-oxidation by UV light.

Spörl et al from Dresden published in German (1997)15 followed by in English (1998)16 the results of their study to induce cross-linkage in porcine eyes. After epithelial removal, the cornea was treated with UV light (254 nm), riboflavin and UV light (365 nm), riboflavin and blue light (436 nm), sunlight, glutaraldehyde (0.1%, 10 min), glutaraldehyde (1%, 10 min) or Karnovsky's solution (0.1%, 10 min). Karnovsky's solution is a fixative used in electron microscopy containing paraformaldehyde, sodium hydroxide and glutaraldehyde. There were 10 eyes in each group and a ninth group served as control. Of the agents tested, riboflavin and UV, glutaraldehyde and Karnovsky's solutions showed an increased in stiffness of the treated cornea compared with control (p <0.05%). At the next stage, riboflavin–UVA and glutaraldehyde 0.075% were evaluated as potential candidates on in vivo studies using rabbit eyes, and riboflavin–UVA was found to show promise for the next stage of human studies.17 Pilot studies on humans using riboflavin–UVA were commenced in Dresden in 1998, and their first results were published in 2003.18

In addition to treating keratoconus, CXL with riboflavin and UV has also been used to treat keratectasia after LASIK,19 20 infectious melting keratitis21 and bullous keratopathy.22 23 The latter application uses the anti-oedematous effect of cross-linkage on the stroma. Collagen cross-linkage with riboflavin and UVA has been sequentially combined with other modalities, namely, intrastromal ring segments24 and photorefractive keratectomy (PRK)25 for the treatment of keratoconus.


The treatment is carried out in sterile conditions, preferably in the operating theatre. After topical anaesthesia (gutt proxymetacaine 0.5%), the epithelium of the central 7 mm of cornea is removed. The surface is then treated by the application of riboflavin (vitamin B2) 0.1% solution (10 mg riboflavin-5-phosphate in 10 ml dextran 20% solution) for 30 min starting 5 min before the start of irradiation. UVA radiation of 370 nm wavelength and an irradiance of 3 mW/cm2 at a distance of 1 cm from the cornea are applied for a period of 30 min, delivering a dose of 5.4 J/cm2.26 Antibiotic drops are instilled as prophylaxis and a bandage contact lens is inserted, which is then removed at the follow-up visit once epithelial healing is complete.

Variations of this protocol include the use of gutt pilocarpine 1% preoperatively, a treatment area of 9 mm and the selective use of steroids in the postoperative regimen to prevent corneal haze (Siena protocol).27 Another modification of the technique involves an epithelial removal zone of 9 mm diameter followed by riboflavin drops instillation onto the cornea every 3 min for 30 min. After confirming that riboflavin has appeared in the anterior chamber on slit-lamp examination, UV irradiation is commenced. During the irradiation time of 30 min, the cornea is rinsed with riboflavin solution and topical anaesthetic every 5 min.28

A UV lamp with modifications designed to deliver a homogenous illumination has been developed and is commercially available (UV-X, IROC, Zurich, Switzerland). The manufacturers claim that in contrast to LED lamps that are dependent on their exact distance from the cornea to ensure that ineffective treatment (too far) or endothelial damage (too close) does not happen, their device that uses the Koehler's beam path is less sensitive to changes in treatment distance. In addition, because of this optical design, the radiant energy diverges behind the cornea, reducing the risk to retina.



In vitro study on human and porcine corneas showed a significant increase in corneal rigidity after cross-linking, indicated by a rise in stress in treated porcine corneas (by 71.9%) and human corneas (by 328.9%) and in Young's modulus by the factor 1.8 in porcine corneas and 4.5 in human corneas.29 The stiffening was later found to be depth dependent with the anterior layers being stiffer.30 The collagen fibril diameter in rabbit cornea was found to be increased by 12.2% (3.96 nm) in the anterior stroma where cross-linking effect was the greatest.31 On thermomechanical testing, the hydrothermal shrinkage of the anterior stroma was observed at 75°C and the posterior stroma at 70°C.32 The observed differences at different depths are attributable to the greater degree of UVA penetration in the anterior stroma with decay in the posterior layers possibly due to riboflavin shielding.

Treated porcine corneas resisted digestion by pepsin, trypsin and collagenase longer than control eyes and the cross-linked anterior cornea was the last to be digested.33 In terms of the hydration behaviour of treated porcine corneas that were incubated in a moist chamber for 24 h, a study found that on light microscopy, the swelling pattern could be divided into three zones. The anterior densely packed zone with high hydration factor, an intermediate zone and a posterior least densely packed stroma with a high hydration factor corresponding to the varying extent of cross-linkage at different depths. This study did not find ocular coherence tomography to be useful in delineating these zones limiting their clinical usefulness.34 Using gel electrophoresis, an intense polymer band of a molecular size of at least 1000 kDa has been identified in treated eyes. This was found to be resistant to mercaptoethanol, heat and pepsin treatment.35 This, in other words, is the biochemical proof of high-molecular-weight collagen polymers that form the cross-linking bridges in collagen.

To assess the long-term stability of the treatment, rabbit eyes were studied after treatment for a period up to 8 months. A variable increase in ultimate stress at 3 months (of 69.7–106.0%) followed by decrease at 8 months (69.7), a steady increase Young's modulus of elasticity (of 78.4–87.4%) and a decrease in ultimate strain (of 0.57–78.4%) were found in that period.36


Wollensak et al37 published their first report on humans in January 2003, using riboflavin and UVA to induce cross-linking in 16 eyes of 15 patients with progressive keratoconus. A further report came out that year in May,18 reporting on 22 eyes of 24 patients with a follow-up time of between 3 months to 4 years (mean 23.2 months). In that they reported that in all treated eyes, the progression of keratoconus was at least stopped. In 16 eyes (70%) there was a regression with a reduction of the maximal keratometry readings by 2.01 dioptres and of the refractive error by 1.14 dioptres. In five patients, the K value remained stable and there was a slight worsening in one patient. Visual acuity improved slightly in 15 eyes (65%). Since that pilot study, there have been eight studies that have reported their efficacy results, the methodologies of which are summarised in table 1.

Table 1

Summary of characteristics of efficacy studies performed to date

The study methodology, in terms of inclusion and exclusion criteria, treatment parameters, outcome measures and analysis are very variable among the studies. All the studies have reported varying degrees of improvement in visual acuity and reduction in keratometry, with a progressive trend in improvement for the duration of follow-up, which was in some cases for ≥3 years. No major complications were noted, although some studies have claimed to have retreated their non-responders.

The Food and Drug Administration has recently approved clinical trials on the treatment in the USA, and at the time of writing, there are 10 trials recruiting patients around the world.38


In their pilot study, Wollensak et al18 reported that corneal and lens transparency, endothelial cell density and intraocular pressure remained unchanged. In studies done on rabbit corneas, they found that a cytotoxic dose of radiation ≥0.65 J/cm2 (0.36 mW/cm2) was reached at endothelium using the standard surface UVA dose of 5.4 J/cm2 (3 mW/cm2) when applied to corneas thinner than 400 μm. A significant necrosis and apoptosis of the endothelial cells at 24 h was noted at the cytotoxic dose level and above.39 It is therefore recommended that in corneas thinner than 400 μm, radiation should not be performed owing to the risk of endothelial injury. The same conclusion was reached following in vitro studies on endothelial cell cultures from porcine corneas.40

Mazotta et al41 in their 3-year serial follow-up of 44 patients who underwent CXL with riboflavin and UV had reported the findings of confocal micromorphological analysis in vivo. They report epithelial regeneration in 4 days with no damage observed in the limbus. There was evidence of healthy epithelial stratification. The immediate disappearance of subepithelial and anterior-midstromal nerve fibres after treatment was reversed by regeneration of subepithelial plexus in the first month and the anterior-midstromal fibres during the second and third postoperative months. The process was almost complete at 6 months restoring normal corneal sensitivity. Disappearance of keratocytes to a depth of 340 μm occurred postoperatively. Confocal microscopy demonstrated a gradual repopulation of the corneal stroma, starting between the second and third month and completing after 6 months.42

A study on rabbit eyes has showed that at the surface dose of 5.4 J/cm2, keratocyte damage was down to a depth of 300 μm from the surface.43 Cell culture studies done on porcine keratocytes have shown that the combination of riboflavin has increased the safety margin of irradiation by 10-fold compared to UVA alone.44

Infrared thermocamera measurements of the human corneal surface during cross-linking treatment have shown that the temperature to be constant during the entire procedure and remain under the threshold of thermal injury to corneal collagen.45 In their review of the safety of the procedure, Spoerl et al28 have highlighted the importance of homogenous irradiance of UVA in the field of application. They advise that hot spots should be avoided as it may mean local exceeding of safely limits causing focal endothelial damage, in spite of the average irradiance being less than the recommended 3 mW/cm2.


Corneal scarring (diffuse subepithelial opacification) that was slow to resolve has been reported in a patient.46 Stromal haze in two cases after cross-linking treatment has been reported in one case report,47 although one large study had noted haze in of varying degrees in all their patients that took up to 12 months to resolve completely.48 Bacterial keratitis has been reported 3 days after treatment.49 Acanthamoeba keratitis due to eye washing under tap water as the patient was unaware of a bandage contact lens being inserted50 and poor contact lens hygiene resulting in polymicrobial keratitis51 has been reported recently. A patient with no history of herpetic keratitis developed herpes simplex keratitis geographical ulcer and iritis 5 days after treatment.52 There is no evidence yet of any lowering of immune mechanisms of the cornea after UV treatment as speculated in some of the reports. Diffuse lamellar keratitis (stage 3) has been reported after treatment in a case of post LASIK ectasia.53

Future research

Approximately 20% of patients with keratoconus will end up having keratoplasty.54 Any intervention that would reduce the need for transplant surgery would be welcomed. The primary challenge in identifying a treatment for keratoconus is that we have not fully understood the pathophysiology of the disease. Cross-linking of tissue is a scientifically sound approach to reduce the risk of ectasia and has the potential to revolutionise the way we manage the disease in future. Once an effective treatment for the condition is found, screening for the disease might become appropriate. It is possible that the costs would be balanced by decreased morbidity and a reduction in healthcare expenditure that a successful treatment would offer. However, there are areas of concern outlined below centering around the robustness of scientific data on the safety and long-term efficacy of the treatment that need to be addressed before a more generalised uptake can be anticipated.

The treatment criteria have to be specified accurately. Even the presently recruiting randomised trials vary in terms of their eligibility and exclusion criteria. For example, the evidence of recent progression of disease if often poorly defined and varies between studies. One of the trials (Federal University of São Paulo) currently recruiting however explicitly defines progression as being “an increase in the cone apex keratometry of 0.75 D; alteration of 0.75 D in the spherical equivalent or increase of the anterior chamber depth on Pentacam in a period of at least 6 months”. The major challenge for these studies of the efficacy of CXL is the very variable natural history of keratoconus. Clinicians are aware that the corneal changes of keratoconus can vary across a wide spectrum from remarkable stability to rapid progression, and this variation in progression can be seen in the same patient at different points in the natural history and the variation in progression can even be very marked between each eye in the same patient. Study design is therefore crucial with a prospective randomised controlled trial of the treatment involving large numbers of patients followed up over many years being needed to answer all the questions concerning long-term efficacy.

The treatment's effect on the tissue is not often apparent, and research needs to shed light on when and whom to re-treat should the initial treatment result be unsatisfactory. Although there is consensus on a minimum corneal thickness of 400 μm to be present, a recent study on rabbit eyes, where riboflavin–UV cross-linking was carried out without epithelial removal, has found a lower penetration depth of 200 μm, raising the possibility for its applicability in corneas thinner than 400 μm, where the standard technique cannot be done without the risk of significant damage to endothelium.55 A more recent case series has shown the possibility of using a hypo-osmolar solution of riboflavin to hydrate the stroma to a thickness of at least 400 μm to successfully treat thinner corneas.56

The treatment protocol itself has to be standardised. The diameter of treated surface, the use of pilocarpine, the method of epithelial removal (is alcohol delamination, as some studies have used, appropriate?), the duration of riboflavin instillation before commencement of UVA, frequency of application during UVA, UVA delivery system (direct LED vs a homogenous illumination device), postoperative management guidelines and clear indications for considering retreatments need to be developed.

Although no change in corneal rigidity affecting IOP measurements have been shown in the short term,57 it is an area of potential concern in the long term. The progressive changes in keratometric indices reaching significance levels at nearly every follow-up time point identified by several studies mentioned above raises concern as to whether this treatment, which in a way induces premature ageing by accelerating the cross-linking that happens naturally with age will, over time, make the rigidity supernormal because of the effect of treatment superimposed on natural increase in cross-links with ageing. A corollary of the argument would be that when some of the patients who develop diabetes in later life would be made even worse, when factors like non-enzymatic glycosylation would come into play, augmenting the treatment effect.

There is a need for a study that addresses the aforementioned design and methodological concerns and be of sufficiently long follow-up for us to know the treatment–effect curve that would inform us of the time to stabilisation of effect (should there ever be one). None of the trials that are currently under way have sufficient follow-up duration to address this question.

The absorption spectrum of the photoinitiator has to overlap the emission spectrum of the source for the oxidation reaction to happen. By choosing a photoinitiator whose excitation wavelength lies within the emission spectrum of visible light, the polymerisation can be made to happen without the need for a radiation source and minimising the harmful effects of the exposure. For example, riboflavin's peak absorption spectrum is in the ultraviolet spectrum but can be extended into the visible spectrum by the choice of the solvent.58

There is also plenty of scope for fundamental research aiming at other ways of cross-linkage by harnessing the naturally or pathologically observed mechanisms of cross-linkage either enzymatically or by non-enzymatic glycosylation (glycation) being avenues that may offer promising solutions to the limitations of photo-oxidation. In their initial publication, where they had tried dilute concentrations of glutaraldehydes, Wollensak et al concluded that the reactions observed were not fast enough as riboflavin–UV. However, in that paper, they report successfully using Karnovsky's solution (0.1%) soaked in blotting paper applied for 5 min to reverse corneal melting that was refractory to conventional treatment.16 More recently, their group have reported successfully treating amniotic membrane tissue with glutaraldehyde 1% solution for 30 min and found that the biomechanical strength was significantly increased as was its enzymatic resistance. The treated membrane was more transparent, less wrinkled and did not dissolve for months.59 It is not apparent from published literature as to why riboflavin–UV was pursued in preference to aldehydes. Although the reactions may be faster and better controlled with the former, a gentler course of treatment (perhaps involving several applications of dilute concentrations) titrated against the tissue response may be the advantage with the latter. Other reducing sugars like fructose (fructation as opposed to glucation) are also promising candidates to induce cross-linkage as they are also present in human tissue owing to their role in the sorbitol pathway.60 These agents may be able to overcome some of the limitations of the riboflavin–UV method.

Assessment of tissue response is the one of the hardest to quantify for research and clinical monitoring of the cross-linked cornea. Keratometric changes have provided surrogate clinical evidence of rigidity. Holographic interferometry,12 electronic speckle pattern interferometry61 and dynamic corneal imaging62 are methods that have been used to measure corneal biomechanics in a clinical setting. Newer devices such as the Ocular Response Analzser (Reichert, Depew, New York, USA) have become commercially available to measure indices such as corneal resistance factor and corneal hysteresis, which may supply us with the direct possible and objective confirmation of corneal rigidity.63

This is a scientific area of exciting potential, and we foresee an explosion of clinical and basic sciences research over the next decade that may ultimately generate a safe and effective treatment method. At present, we recommend that where this procedure is carried out especially outside a controlled research setting, appropriate information in terms of its efficacy and risks (even if potential rather than actual) including the irreversible nature of the treatment is provided to all patients.


View Abstract


  • Competing interests None.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.