Repeated corneal crosslinking for progressive keratoconus: efficiency and safety

Abstract

Corneal crosslinking (CXL) was first introduced in clinical practice in 2003. Since then, this procedure has been established as the first-line treatment in the management of progressive keratoconus. Over the last years, many different protocols have emerged, each one of them with variable clinical outcomes and safety profile. Progression of keratoconus after primary CXL is very rare, but it has been reported in the literature. This review summarises the existing data on repeated CXL after primary failure, emphasising on clinical efficacy and safety.

Introduction

In 2003, Wollensak and Spoerl were the first to introduce corneal crosslinking (CXL) with the Dresden protocol (epi-off procedure) for the management of progressive keratoconus. ¹ Since the first clinical application of conventional CXL, many different protocols and CXL modalities with significant variations have emerged, including CXL without abrasion of epithelium (epi-on procedures), accelerated CXL, pulsed CXL, customised CXL, aiming to increase the efficacy and safety of the procedure.^{2 –5}

CXL has been employed for the management of progression in numerous corneal ectatic diseases, such as keratoconus, iatrogenic post-LASIK ectasia, pellucid marginal degeneration, Terrien’s marginal degeneration, and more.^{6 –8} Progressive keratoconus represents, however, the most common indication for CXL. ⁹

CXL leads to corneal stiffening and stability, halting the progression of keratoconus. ¹⁰ At the same time, CXL is inducing corneal flattening, which is contributing to the remodelling process occurring in the cornea, along with biomechanical alterations in corneal tissue. ¹⁰ This remodelling process is responsible for topographic changes recorded even years after treatment. ¹¹

However, progression of keratoconus can still occur after CXL. Repeated following unsuccessful primary CXL procedure, was first reported in 2014 by Hafezi et al. ¹² Later on, several cases of repeated CXL treatment after failure of primary CXL, have been reported.^{13 –21}

Hereby we aim to provide an overview of the literature on repeated CXL for progressive keratoconus after failure of primary CXL, emphasising on available data regarding efficacy and safety.

Definition of progression and pseudoprogression, after primary CXL

Numerous earlier studies and randomised control trials have demonstrated that CXL halts the progression of keratoconus and even induce corneal flattening.^{22 –24} The protective effect of CXL can last long term. ²⁵ However, the failure rate of CXL has been estimated to be approximately 7.2%. ²⁶ Progression of keratoconus despite CXL treatment is defined as failure of CXL. ²⁷ Differences in the reported CXL failure rate among various studies, depend on the definition criteria of progression, the selection of patients, as well as the implemented treatment protocol. Progression of the disease has been detected in as many as 23% of the cases among eyes with advanced keratoconus treated with CXL. ²⁸

Most authors define progression as an increase of more than 1 diopter, in maximal keratometry (Kmax), over a period of 12 months.^26,28 Visual acuity deterioration, increase of astigmatism or decrease in corneal thickness have also been included in the criteria of progression in many studies on repeated CXL after primary CXL failure. ²⁹ The importance of multivariable estimation of progression, can be highlighted by the 5-year results of Mazzotta et al.; after accelerated CXL approximately 8.33% of eyes exhibited topographically defined progression based on Kmax readings, following initial improvement. ¹¹ However in none of these eyes was visual function negatively affected, thus no further re-treatment was required. ¹¹ Despite potential limitations, Kmax remains the most commonly used parameter for detection of keratoconus progression and evaluation of CXL efficacy. ³⁰

Xanthopoulou et al., described, in an analysis of 151 corneas after CXL, corneal steepening with decrease in BCVA 6 weeks after primary CXL. However, further follow-up examinations revealed a progressive flattening most apparent 1 year post-operatively. ³¹ This observation is in accordance with the study of Antoun et al., which demonstrated major topographic changes occurring in the first 6 months post-operatively, with concomitant changes in VA and refraction; thereafter, any change in keratometry was considered as true progression. Vinciguerra et al., also reported an increase in steepest meridian keratometry and the simulated cylinder (AK Values) at the first month after CXL, followed however by regression of those topographic values with flattening of corneal curvature after the third month and amelioration of BCVA und UCVA during the first 6 months after CXL. ³²

Those published results are in agreement with the Siena Eye-Cross Study 2, which manifested a transient deterioration of Kmax during the first months after CXL and statistically significant improvement from the sixth month until the fifth year after CXL. ¹¹ The initial, yet temporary progression of corneal steepening following CXL, is defined as pseudoprogression.^31,33 The exact duration of this period of instability, remains unclear, with reports varying in the literature, depending on the design of each study. However it has been shown that no major changes occur after the first 6 months post-CXL. ¹³ Thus, it is recommended that the baseline post-operative evaluation for the detection of keratoconus progression, takes place at least 6 months post-operatively, while a more conservative approach would set the appropriate time point at 9–12 months post-operatively.^13,14,33 On the other hand, the observed flattening effect of CXL commences approximately at the sixth month post-operatively, being more apparent after 12 months though. ³⁴ Apart from that, an additional, ongoing course of flattening may be noticed during the next years after CXL. ³⁴

Pseudoprogression might be attributed to initial transient haze and the beginning of CXL-induced corneal remodelling. ¹³ It is also important to consider that the repopulation of keratocytes responsible for the long-term corneal remodelling has not been completed until 4–6 months after CXL^25,35 Additionally, corneal thickness fluctuations are observed during this early period of instability. An early edematous phase is immediately followed by a decrease in corneal thickness due to the packing of collagen fibres, responsible for the temporary steepening effect. ²⁹ This process is mainly observed during the first month. ¹¹ It is however believed that this decrease in corneal thickness does not represent a negative effect of corneal cross-linking; hence, it should not be interpreted as true progression. ³⁶ In fact, a second sequel of slight increase in corneal thickness is observed afterwards, which tends to restore the preoperative corneal thickness, without fully achieving it.^11,36 Consequently, the pseudoprogression effect starts to fade away. It is assumed that this slight increase in corneal thickness, may be caused by the production of proteoglycans by the new keratocytes that have repopulated the anterior corneal stroma. ³⁶

The long-term stabilisation and the ongoing flattening of the cornea associated with corneal remodelling, can last a long-term, albeit it is not an endless process. Indeed, the protecting effect of CXL can gradually disappear as shown by Raiskup et al. ¹⁴ Collagen regeneration process can partially explain this phenomenon. ¹⁴ Although the exact rate of collagen turnover is not known, it is estimated to occur every 2–7 years.^29,37 Moreover, CXL per se is supposed to extend collagen half-life by halting the activity of stromal collagenases.^35,38 Consequently, it is speculated that 10 years after initially successful CXL, progression of keratoconus can reoccur.^21,29 These estimations are based both on clinical observations and basic research, without definitive conclusions obtained yet. ³⁹ For instance, progression has also been noted even 5 years after a long period of post-CXL stabilisation period in two cases presented by Kymnionis. ⁴⁰ Lenk et al. detected CXL failure occurring at different time points during a 10-year follow-up period post-operatively. ⁴¹ Further evidence for the collagen half-life hypothesis is provided by the higher relapse rates in paediatric keratoconus, as well as the known natural stiffening of cornea in older persons, both of which are possibly related to different rates of collagen turnover in each age group.^21,39

Factors that influence the efficacy of primary CXL

Numerous studies have identified potential risk factors for failure of primary CXL.^{17,21,28,41,42} Younger age seems to represent an important factor for primary CXL failure, as mean age of cases exhibiting progression of disease after primary CXL was significantly lower compared to cases without CXL failure. ¹⁷ Higher collagen turnover rate in younger patients, may be responsible for keratoconus progression after primary CXL, while it also correlates with more severe forms of keratoconus. ²¹ In a study by Caporossi et al., assessing the post-operative results of epi-on CXL, nearly half of the paediatric patients exhibited keratoconus progression requiring a secondary procedure. ⁴² Moreover, high preoperative Kmax reading (>58) has been reported to be a significant risk factor for progression of keratoconus after CXL.^28,41 High preoperative astigmatism (>4.3D) has also been linked to failure of primary CXL. ⁴³

A plethora of studies have documented that patients with atopic diseases like neurodermitis, carry an extra risk for further progression of keratoconus.^13,41,44,45 The combination of neurodermitis with other atopic diseases has been shown to significantly affect the progression after initial CXL. ⁴¹ The same is true for allergic conjunctivitis and eye rubbing. ¹³ Eye rubbing has been associated with loss of keratocytes through increase in temperature and the consequential action of collagenases, which increase the degradation of the collagen matrix. ⁴⁴ Another mechanism suggestive of loss of keratocytes through eye rubbing includes increase of the intraocular and hydrostatic pressure. ⁴⁴ Metabolic changes that occur during pregnancy have also been proposed as possible risk factors for keratoconus progression prior or after CXL treatment. ⁴⁵

In terms of the applied CXL protocol, overall failure rate has been reported to be higher after transepithelial procedures (epi-on).^46,47 Preliminary findings demonstrate that the Dresden epi-off CXL protocol achieves deeper crosslinking, as evaluated by demarcation line depth measurement, in comparison to accelerated CXL.^48,49 Thus, studies with longer follow-up and larger samples are required to demonstrate possible difference in failure rates between classical Dresden and accelerated protocols.

Demarcation line and efficacy

Demarcation line depth has been widely used as an indicator to evaluate efficacy of CXL.^2,50 Mainly identified in clinical practice with the aid of anterior segment OCT (AS-OCT), demarcation line represents the transition zone between the anterior crosslinked and the posterior untreated stroma, resulting in collateral areas of different refraction indexes. Likewise, keratocytes population density also differs between the two zones, due to the UVA mediated cytotoxic effect of riboflavin to the anterior stroma above the demarcation line. ³⁶ The depth of treated stoma is of paramount importance as, not only is the anterior part of the cornea considered to be largely responsible for corneal stiffness in general but also UVA absorption itself is mainly absorbed within the first 200 μm of cornea. ⁵¹ Actually, Mazzota et al. modulated UVA power settings accordingly to achieve a desired demarcation line depth of 250 μm at least, further highlighting possible clinical applications, as a credential of efficacy. ¹¹

There are no reports in the literature assessing demarcation line depth in cases of repeated CXL, with the exception of the first cases presented by Hafezi, who found no significant differences between primary and secondary procedures. ¹²

Experimental basic science studies on repeated CXL

The biomechanical effect of repeated CXL was investigated by Beshatawi et al. in 30 cadaver human corneas in vitro with the aid of scanning acoustic microscopy (SAM), showing an additive effect, albeit not statistically significant and not as extensive as that of the primary treatment. ³⁷ Thus, further cross links may be induced even after shortly delivered secondary CXL; however, the additive biomechanical effect has be found to be insignificant. Tabibian et al. also assessed the biomechanical effect of repeated CXL treatments in mice cornea by ex vivo stress-strain extensiometry. ⁵² The results demonstrated a limited effect of the secondary procedure, in accordance with the aforementioned in vitro experiments. Zhang et al., analysed the biomechanical effect of repeated CXL procedures in cat corneas by ex vivo stress-strain assessment 1 month post-operatively. The results were in agreement with those of Beshtawi et al. and Tabibian et al. in terms of the limited effect on additive corneal stiffening. ⁵³

The intervals between the CXL procedure and the biomechanical tests of these experimental studies may not be long enough to demonstrate the long-term remodelling taking place in human corneas, which are observed in the clinical practice, during the following months to years after primary or repeated procedures. ⁵² The migration and repopulation of keratocytes which may be responsible for this ongoing process, cannot be consecutively reactivated to achieve additive stiffening in such short periods and ex vivo.^35,53 Not only the effects of the long-term remodelling but also the rate of collagen renewal and turnover may also partially explain the additive effect of repeated CXL in clinical practice, but not in basic research experiments.

Outcomes of repeated CXL after failure of primary procedure

We searched in Pubmed for literature on repeated CXL, between 2004 and 2024: (redo OR retreatment OR repeated) AND (crosslinking OR CXL OR cross-linking). The results of our search are summarised in Table 1.

Table 1.

Summary of clinical studies including cases of repeated CXL.

Author/Year	Protocol	Mean time difference between 2 CXL	Failure rate of primary CXL	Efficacy rate of repeated CXL
Hafezi et al. 2014 12	Accelerated epi-off (10 min @ 9 mW/cm2)	4 years	N/A Case report	N/A Case report
Antoun et al. 2015 13	Dresden protocol (30 min of 3 mW/cm2)	7 years	7 out of 221 eyes (3.17%)	7 out of 7 eyes (100%)
Mazzotta et al. 2018 21	Modified Dresden protocol (30 min of 3 mW/cm2)	N/A	6.3% (4 eyes)
Turhan et al. 2020 15	Accelerated epi-off (10 min @ 9 mW/cm2)	19.3 months	12 out of 800 eyes (1.5%)	10 out of 12 eyes (83.3%)
Wu et al. 2021 16	Epi-on (double iontophoresis) (10 min @ 9 mW/cm2)	N/A	21 out of 1027 eyes (0.02%)	21 out of 21 eyes (100%)
Jamali et al. 2022 19	Accelerated epi-off (10 min @ 9 mW/cm2)	49.71 months	N/A	7 out of 7 eyes (100%)
Grentzelos et al. 2022 20	Dresden protocol (30 min of 3 mW/cm2)	6 years	N/A Case report	N/A Case report
Raiskup et al. 2023 14	Dresden protocol (30 min of 3 mW/cm2)	9 years	6 out of 42 eyes (14%)	6 out of 6 eyes (100%)
Maskill et al. 2023 17	Accelerated epi-off (4 min @ 30 mW/cm2)	47.1 months	21 out of 1535 eyes (1.37%)	20 out of 21 eyes (95%)
Vorobichik et al. 2023 18	Accelerated epi-off (10 min @ 9 mW/cm2)	2.7 years	N/A	10 out of 10 eyes (100%)

Hafezi first published a case of a 32-year-old man with progression of keratoconus in one eye (Kmax increase of 1.2D), 4 years after epithelium-off primary CXL on both eyes, according to the classic Dresden Protocol. ¹² Repeated accelerated epi-off (10 min @ 9 mW/cm²) CXL procedure was performed, leading to stabilisation of the disease with decrease in Kmax of more than 2 D over 2 years post-operatively. ¹²

Antoun et al. reported on repeated CXL in 7 eyes of 5 patients, between 2010 and 2013. ¹³ Keratoconus was the underlying disease in 6 out of 7 eyes, whereas post-LASIK ectasia in the remaining one. Repeated CXL was necessary only in 3.17% of eyes, which underwent primary CXL (7 out of 221 eyes). Progression detection after primary CXL was based on a combination of Kmax, K1 or K2 increase of more than 1D along over a period of at least 6 months post-operatively. Both primary and repeated CXL procedures followed the same classic epi-off Dresden protocol (30 min of 3 mW/cm²). Repeated CXL was successful in all eyes over 12 months after treatment, without any complications.

Raiskup et al., in a series of 42 eyes with 15-years follow-up after primary CXL (performed between 2001 and 2006), described 6 eyes that necessitated repeated CXL (failure rate of primary CXL was 14%). ¹⁴ Progression of Kmax more than 1D at least 9 months post-operatively was considered as indication for repeated CXL. The mean time between initial CXL and the secondary procedure was 9 years, with the exception of 1 eye requiring retreatment after only 9 months. Thus, the importance of continuous follow-up in the long run is highlighted by the authors. Additionally, repetition of the CXL procedure was uneventful, while not only did it decrease Kmean and Kmax in all eyes but also improved the visual acuity in three cases, confirming the efficacy and safety of retreatment, similar to the aforementioned repeated CXL studies.

Turhan et al. in a study with 12 keratoconic eyes from 10 patients, reported on repeated accelerated epi-off CXL (10 min @ 9mW/cm²), with a mean interval of 19.3 months after the first procedure between 2011 and 2018. ¹⁵ Failure rate after primary CXL was 1.5%. Increase of Kmax more than 1D or astigmatism increase of 1D or more constituted the inclusion criteria, with an average Kmax increase of 2.3D observed among the eligible cases, detected 7–42 months after primary procedure. After repeated CXL and during the follow-up period of 36 months on average, regression of Kmax was observed in 8 eyes, stabilisation in 2 eyes and progression in 2 eyes, without any post-operative complications noted.

Effect and safety of redo-crosslinked assisted by transepithelial double-cycle iontophoresis (DI-CXL) in progressive keratoconus after primary CXL failure was assessed by Wu et al. and published in 2021. ¹⁶ This study included 21 eyes of 12 patients who underwent repeated CXL out of 498 patients previously operated between 2011 and 2017 (primary CXL failure rate was 2.41%). Progression criteria included at least two of the following: Kmax increase of more than 1D, visual acuity decrease of one or more lines and change in astigmatism of more than 1D. All cases demonstrated decrease in Kmax post-operatively, without any complications, during a 24-month period.

Furthermore, another study was published by Maskill et al. and included 21 eyes of 20 patients with keratoconus, who underwent repeated accelerated epithelium-off CXL (4 min @ 30 mW/cm²), during a 10-year period (2012–2022) with a mean follow-up period of 29.9 months. ¹⁷ Progression was defined as increase in Kmax of more than 1.5D or increase in corneal astigmatism of more than 1.5D or loss of two lines or more of BCVA. Selection criteria were applied in 1535 keratoconic eyes, which previously underwent primary CXL (failure rate of primary procedure was 1.37%). The authors reported a mean interval between the primary and secondary procedure of 47.1 weeks. Stabilisation rate of over 95 % after repeated CXL was recorded (20 out of 21 eyes). Finally, it has to be noted that eight eyes in total had an intracorneal ring segment implanted prior to the secondary CXL procedure.

A retrospective study by Vorobichik et al. was published in 2023, including 10 eyes which underwent repeated accelerated epi-off CXL (10 min @ 9mW/cm²). ¹⁸ The failure rate of primary CXL was 8%. Criteria for repeated CXL included increase in Kmax of more than 1D and 10 μm decrease in thinnest pachymetry, within 1 year after primary CXL. Stabilisation of keratometric readings was reported in 8 out of 10 eyes in this study.

In another recent study by Jamali et al., seven cases of repeated accelerated epi-off CXL (10 min @ 9 mW/cm²) were evaluated. ¹⁹ Primary CXL procedure was also based on the same protocol; however, no failure rate was reported. After the secondary keratometric improvement was noted in all eyes included, with a mean follow-up time of 18.71 months (12–41 months).

In addition, a case report of repeated CXL procedure for progressive keratoconus, showed an interesting decrease of 16D in Kmax during a follow-up period of 6 years. ²⁰ Both primary and secondary CXL procedure were performed according to Dresden protocol.

Finally, a prospective cohort study by Mazzotta et al., included 62 eyes of 47 paediatric patients, who underwent primary standard CXL for progressive keratoconus. ²¹ Authors reported that 4 eyes of 2 children required repeated CXL after 36 months. Repeated CXL was successful in all eyes. ²¹

Safety of repeated CXL

Regarding the safety profile of repeated CXL, the study of Antoun et al., extensively addressed this topic. ¹³ No major complications were observed until 1 year post-operatively.

Overall, repeated-CXL is considered to be a safe approach for management of cases with progressive keratoconus after primary CXL failure. ¹⁸ However, corneal endothelial cell count (ECD) was not measured, neither by Antoun nor by the rest of the repeated-CXL clinical studies, which is an important limitation of the safety assessment. However, it is worth mentioning the case report by Grentzelos et al. which displayed a stable ECD after the repeated procedure. ²⁰

Furthermore, attention should be given regarding the risk of excessive corneal flattening following repeated CXL. One case demonstrated extreme corneal flattening (up to 16.00 D in Kmax) within the first 5 years after repeated CXL, followed by stabilisation during the sixth year of follow-up. ²⁰ In a study by Henriquez et al., 15.5% of patents (7 out of 45 cases) experienced excessive flattening after initial CXL. Notably, the authors observed that excessive corneal flattening is not necessarily disadvantageous for the patients; it is often associated with improvement in CDVA and UDVA (in 5 of 7 cases) rather than with a decline in CDVA (in 2 of 7 cases). ⁵⁴

Protocol selection is also of paramount importance. Minor adverse events related to CXL can occur at as many as 26.7% of epi-off cases in contrast to only 3.3% in epi-on CXL. ⁵⁵ Consequently, epi-on protocol may constitute a safer option for repeated-CXL, as repeated epithelial removal may theoretically increase the risk of post-operative complications. ¹⁶ In addition, Maskil et al. reported significant reduction of central corneal thickness after repeated CXL, in comparison to the primary procedure. ¹⁷

Conclusion

Progression of keratoconus following primary corneal crosslinking constitutes a rare, but not neglectable complication that all ophthalmologists should be aware of. Primary CXL failure rates have been reported in the literature, to vary between 2% and 10%, but can also be as high as 23%, depending on the study design criteria. Risk factors known to affect primary CXL are yet to be conclusively determined. It is however of paramount importance to clearly define the criteria of progression, as well as the interval between CXL and initial post-operative assessment, in order to prevent confusion with pseudoprogression.

Repeated CXL mechanism of action has been proposed to be associated with the effect on residual unlinked corneal stroma collagen after the primary treatment, along with the following long-term remodelling, after the repopulation of keratocytes.^19,37,21,20 Additionally, the direct crosslinking effect on the newly produced collagen, during the process of collagen regeneration, appears to play an important role. ¹⁴ Although basic research does not demonstrate any efficacy of repeated CXL, clinical studies undoubtedly confirm not only the efficacy but also the safety of repeated CXL procedures.^{12 –15,17 –19,37,52,53}

According to the current literature data, repeated CXL represents an effective and safe procedure, resulting in the stabilisation of progressive keratoconus. No major complications have been reported post-operatively, while minor ones, such as mild corneal haze, do not differ than those occurring after primary CXL. It has been proposed however that epi-on CXL could be theoretically a safer approach for repetitive CXL treatments, despite concerns about reduced efficacy in comparison to the epi-off protocols. Hence, measurement of demarcation line depth could facilitate a valid evaluation of CXL efficacy, especially in the early post-operative period, until long-term data can enable definitive conclusions.

The main restriction in available studies on repeated CXL is the heterogenicity of the cases in terms of severity of the disease, definition of progression, CXL protocol parameters, etc. Moreover, the short post-operative follow-up after repeated CXL does not allow conclusive proof of both efficacy and safety. Further randomised clinical trials with longer follow-up period are required in order to better validate the efficacy and safety of repeated CXL, as well as further identify risk factors for progression after primary CXL.