Comparative Results of “Epi-Off” Conventional versus “Epi-Off” Accelerated Cross-Linking Procedure at 5-year Follow-Up

Purpose The aim of our study was to compare the long-term efficacy and safety of “epi-off” conventional and “epi-off” accelerated corneal cross-linking (CXL) in patients with progressive keratoconus. Methods “Epithelial-off” (“Epi-off”) CXL using the conventional technique (3 mW/cm2, 30 minutes) was performed in 93 eyes of 93 patients (S-CXL group) and “epi-off” accelerated method (9 mW/cm2, 10 minutes) in 76 eyes of 76 patients with progressive KCN (A-CXL group). Cases with different stages of keratoconus and topographic evidence of progression were included. Main outcomes comprised refraction, keratometry measurements, uncorrected (UCVA) and best-corrected visual acuity (BCVA), and topographical indices. Micromorphological analysis was assessed by anterior segment ocular coherence tomography (AS-OCT). The follow-up period was 5 years. Results In both groups, Kflat presented similar results: decrease at 1 year (p=0.465), at 2 years (p=0.672), at 3 years (p=0.198), at 4 years (p=0.32), and at 5 years (p=0.864). In both groups, Ksteep presented a similar decrease at 1 year (p=0.709), at 2 years (p=0.455), at 3 years (p=0.43), at 4 years (p=0.57), and at 5 years (p=0.494), with no statistically significant difference. Decrease in Kavg was similar in both groups at all analyzed time points (p=0.18 at 1 year, p=0.093 at 2 years, p=0.57 at 3 years, p=0.154 at 4 years, and p=0.247 at 5 years). Kmax had a similar decrease in both groups at 1 year (p=0.06), at 2 years (p=0.09), at 3 years (p=0.126), at 4 years (p=0.113), and at 5 years (p=0.114). There was no statistically significant difference between the cylinder decrease in both groups (p=0.349 at 1 year, p=0.6782 at 2 years, p=0.299 at 3 years, p=0.0943 at 4 years, and p=0.144 at 5 years). The BCVA values were statistically significantly higher than the preoperative values in both groups at all time points (p < 0.05). Topographical indices such as thinnest corneal point (TP), corneal volume (CV), index vertical asymmetry (IVA), index of vertical asymmetry (ISV), index of height asymmetry (IHA), index of height decentration (IHD), Belin/Ambrosio Enhanced Ectasia Display (BAD_D), and Ambrosio retinal thickness (ART Max) were significantly statistically decreased compared with baseline at all time points, in both groups. Conclusion “Epi-off” accelerated and conventional CXL have the same efficacy in terms of improvement in visual and topographic outcomes.


Introduction
Keratoconus (KCN) is a bilateral ectatic corneal disorder, frequently asymmetric, with a progressive thinning of the cornea resulting in protrusion, progressive irregular astigmatism, and visual deterioration [1]. In the majority of cases, this condition affects young patients with an early age of onset as a negative prognostic factor for evolution and corneal transplantation [2]. at is why, an early detection of KCN by using the topographical and tomographical evaluation of the Pentacam device is of major importance [3,4]. During the past years, corneal collagen cross-linking (CXL) became the standard procedure for KCN therapy. CXL has been used in dentistry, orthopedics, and dermatology for many years, and since 1998, it has been tested and subsequently introduced in the treatment of KCN together with riboflavin (a nontoxic photosensitizing agent) and ultraviolet irradiation (UVA) [5]. e final effect of the CXL technique is represented by the strengthening of the cornea, and the goal is to slow down or stop the progression of KCN, thus avoiding or delaying the necessity of keratoplasty.
CXL technique consists in the photopolymerization of the stromal fibrillar tissue, in order to increase their stiffness and resistance to corneal ectasia and proteolytic enzymes (collagenase) [6,7], reducing corneal permeability [8] and formation of large collagen molecular aggregates [9] through the combined action of the photosensitizing substance (riboflavin-B2) and ultraviolet A (UVA) light irradiation performed with an illuminator in a solid state of UVA kind [5,9]. e conventional (standard) technique ("epi-off" technique) named as the Dresden protocol uses riboflavin which is exposed to a measured dose of long wavelength UVA radiation (370 nm) at 3mw/cm 2 for 30 minutes, applied after epithelial removal and resulting in a total energy dose of 3.47 or a radiant exposure of 5.47 cm [2,9,10]. Several studies demonstrated the safety and long-term efficacy of the conventional "epi-off" cross-linking in stabilizing progressive KCN [11][12][13][14][15][16][17][18][19].
In order to reduce the time of treatment and the patient discomfort and to avoid the excessive corneal dehydration and thinning that can occur for a period of 30 minutes, the accelerated "epi-off" CXL was introduced. By applying a higher intensity (9 mw/cm 2 ) for a shorter period of time (10 minutes), the same level of radiant exposure as in the conventional CXL can be achieved [20,21]. is is according to Bunsen-Roscoe's law of reciprocity that showed that an increase of UVA irradiation associated with a reduced exposure time theoretically delivers a total energy dose to the tissue similar to that in the conventional treatment, with the same biological effect [20,21]. Previous ex vivo experiments on porcine corneas performed with high energy and short irradiation time settings have revealed similar results on the biomechanical properties compared to the standard protocol [20]. e potential advantages of the accelerated CXL technique include decreased exposure time, improved comfort of the patient, and inferior infection risk [22]. Furthermore, there are studies that demonstrated the efficiency of "epi-off"accelerated CXL [20,[23][24][25]. However, Hashemi et al. [26] compared the two-year changes in dynamic corneal response evaluated by Corvis ST, between 18 mW/cm 2 (5 minutes) and 9 mW/cm 2 (10 minutes) CXL protocols.
Evaluation of the demarcation line is considered as a measurement of the depth of CXL treatment into the stroma [35]. Corneas that underwent CXL treatment present an inferior rate of enzymatic collagenase degradation when compared to an untreated cornea [35]. Some studies [36,37,39] showed the role of the demarcation line after CXL depth as representative of CXL effectiveness. e aim of our study was to assess the long-term efficacy and safety of "epi-off" conventional and accelerated CXL by comparing the keratometry measurements, spherical and cylinder equivalents, visual acuity, and topographical indices in patients with progressive keratoconus up to 5 years, following the procedure, and also to determine the qualitative corneal changes evaluated by anterior segment optical coherence tomography (AS-OCT) after both procedures.

Materials and Methods
A retrospective single-center study was performed at the Oculens Private Clinic in Cluj-Napoca, Romania, after the unanimous approval of the clinic ethics committee (5/2020) and adhered to the tenets of Helsinki Declaration (1964). Two groups of patients were included in our study. e conventional group included 93 eyes of 93 patients, with progressive KCN who underwent CXL by the conventional (standard) "epi-off" technique (S-CXL group). e accelerated group included 76 eyes of 76 patients, with progressive KCN who underwent CXL by the accelerated "epi-off" procedure (A-CXL group). All patients included in the study were affected by KCN with a documented clinical and instrumental worsening at least in the last three to six months of observation: myopia and astigmatism changes >3D, a mean change of central K value > 1.5D in three consecutive corneal topographic measurements, increase in the maximum keratometry (Kmax) in topography of more than 1D, or a mean decrease in central corneal thickness >5% in three consecutive tomographic measurements [39]. e "epi-off" conventional CXL was performed between January 2011 and January 2013 and the "epi-off" accelerated CXL between February 2013 and January 2015. All cases were followed up postoperatively at 1 month, 6 months, and 1, 2, 3, 4, and 5 years, following the procedure. e inclusion criteria for both groups included age greater than 18 years, any gender, progressive KCN of different stages (according to the Amsler-Krumeich classification), and an average corneal thickness of at least 400 µm at the thinnest corneal location. Patients with previous intracorneal ring placement, corneal pachymetry less than 400 µm, Vogt striae presence, history of herpetic keratitis or other corneal infections, dry eye syndrome, aphakia, central corneal scar, autoimmune illnesses, nystagmus, and pregnancy were excluded.
After signing the specific informed consent, the two groups of patients underwent the "epi-off" conventional or "epi-off" accelerated CXL procedure. e CXL procedure was performed in the operating room in sterile conditions. e "epi-off" CXL technique was used for both conventional and accelerated procedures. During the procedure, a single 3.0 ml of riboflavin 0.1%dextran 20% solution (Peschke D) was opened, and the power of the UVA illuminator (Peschke Meditrade GmbH, Huenenberg, Switzerland CXL system) was verified. Before the procedure, one drop of Isopto ® Carpine (Amman Pharma, Romania) was instilled in the eye and was prepared with topical anesthesia with oxybuprocaine hydrochloride 0.4% (Benoxi, Unimed Pharma Ltd), 1-2 drops, 10-15 minutes prior to cross-linking was instilled. Sterile operating field and lid speculum were used. A corneal de-epithelization with a blunt spatula on a 9 mm diameter optical zone was performed, followed by the instillation of riboflavin 0.1%dextran 20% solution (Peschke ® D, Huenenberg, Switzerland), every 2 minutes for 30 minutes before irradiation. Pachymetry was done after epithelial removal to ascertain that the corneal depth was more than 400 m. e optical corneal zone was additionally aligned under a UVA power of 3 mW/cm 2 for 30 minutes in conventional CXL (Dresden protocol) (total energy:7.2 J/cm 2 ) and under a power of 9 mW/cm 2 for 10 minutes in accelerated CXL (total energy: 5,4 J/cm 2 ). Riboflavin 0.1% every 2 minutes was instilled during irradiation in both groups. e cornea surface was irrigated with balanced salt solution after irradiation. At the end of the surgery, steroids and antibiotics (Tobradex) (Alcon Novartis, Dallas, Worth, USA) were instilled. A bandage soft contact lens was applied for 3 days until the corneal healing was completed. After the procedure, all patients received topical antibiotics and steroids such as Tobradex (Alcon Novartis, Dallas, Worth, USA) 5 times/day, for one month, and artificial tears 2 times/day for 2 to 3 months. Both groups were followed up postoperatively in the clinic at day 1, day 3, month 1, month 6, year 1, and yearly up to 5 years. Ocular refraction, keratometry measurements, uncorrected visual acuity (UCVA), best-corrected visual acuity (BCVA), slit-lamp examination, corneal tomography, and topography were performed at all visits, except for the first postoperative day, 4 weeks, and 6 months. Corneal topographic and aberrometer parameters were recorded from all the examinations. At one month after the procedure (in both groups), anterior segment optical coherence tomography (AS-OCT) analysis (Triton OCT, Topcon Corporation, Tokyo, Japan) was done in order to assess qualitative A-CXL-and S-CXL-induced corneal changes and treatment penetration looking after demarcation line. e image was captured when the corneal reflex was visible, and the depth of demarcation line was measured using the caliper tool provided by the manufacturer.

Statistical Analysis.
Data were presented using the number (percentage) for categorical variables and mean-± standard deviation or median (quartile 1; quartile 3) for continuous variables. Groups were compared using the independent Student t-test for continuous variables with normal distribution, the median test for continuous variables with abnormal distribution, the and chi-square test for categorical variables. e evolution of variables at different time points was compared with baseline values using the paired t-test. A p value <0.05 was considered as statistically significant.

Results
We analyzed data of 93 eyes of 93 patients with progressive KCN who were treated with the conventional (standard) "epi-off" CXL technique (S-CXL group) and 76 eyes of 76 patients treated with the "epi-off" accelerated CXL (A-CXL group). At baseline, the mean age was 26.5 ± 8.5 years (ranging between 18 and 54 years) in the S-CXL group and 29 ± 8.2 years (ranging between 18 and 48 years) in the A-CXL group (p value for the difference between groups � 0.064). ere were 52 males (55.9%) in the S-CXL group and 41 males (53.9.3%) in the A-CXL group (p � 0.184). Regarding the stage of keratoconus (Amsler-Krumeich staging), 8.6% had stage I, 53.76% of patients had stage II, and 37.63% stage III in the S-CXL group. In the A-CXL group, 7.89% had stage I, 47.37% of patients had stage II, and 44.74% stage III (Table 1).

Corneal Findings during the Treatment.
ere was no statistically significant difference between study groups in terms of preoperative flat keratometry ( Figure 3). e reduction of Kavg from baseline data was 1.06D in the S-CXL group and 0.93D in the A-CXL group.
For Kmax, the decrease was similar in both groups at all time points analyzed. In the S-CXL group, Kmax dropped from 54.05 ± 5.41 D to 52.10 ± 5.35 (p � 0.0172) at 1 year, to 51.99 ± 5.35 D (p � 0.012) at 2 years, to 51.78 ± 5.28 D (p � 0.0086) at 3 years, to 51.74 ± 5.31 D (p � 0.0075) at 4 years, and at to 51.72 ± 5.32 at 5 years (p � 0.0071). In the A-CXL group, Kmax dropped from 56.07 ± 5.39 D to 54.45 ± 5.42 D at 1 year (p � 0.0456), to 54.22 ± 5.44 D (p � 0.0368) at 2 years, to 54.02 ± 5.38 D (p � 0.0199) at 3 years, to 54.01 ± 5.38 D (p � 0.0196) at 4 years, and to 54.00 ± 5.39 D at 5 years (p � 0.0189) (Figure 4). ere was a statistically significant decrease in Kmax in both groups comparing with baseline at all time points (Table 2). e reduction of Kmax from baseline data was 2.33 in the S-CXL group and 2.07 D in the A-CXL group.
Compared to preoperative status, for conventional and accelerated CXL, Kflat, Ksteep, Kavg, and Kmax were statistically significantly lower at year 1 and were maintained statistically significantly lower at all time points (p < 0.05 for all time points as compared to preoperative values) (  e preoperative UCVA in the A-CXL group was 0.73 ± 0.19 LogMar and increased to 0.69 ± 0.24 LogMar at 1 year, to 0.67 ± 0.23 LogMar at 2 years, to 0.65 ± 0.21 LogMar at 3 years, to 0.65 ± 0.24 LogMar at 4 years, and became 0.65 ± 0.23 LogMar at 5 years. For BCVA, the improvement was not statistically significantly different between the A-CXL group and the S-CXL group at 1 year (p � 0.1142), at 2 years (p � 0.908), at 3 years (p � 0.346), at 4 years (p � 0.4575), and at 5 years (p � 0.4072). Compared to baseline, in the S-CXL group, postoperative UCVA was statistically significantly lower than preoperative values at all time points analyzed (p < 0.001 for all time points). In the A-CXL group, postoperative UCVA was statistically significantly higher compared to baseline starting from year 1 and maintained higher than preoperative levels at years 3, 4, and 5 (p < 0.001) (Figure 7).
For BCVA, the values were statistically significantly higher than the baseline values in both groups at all time points (p < 0.05) (Figure 8).    (Table 3). We did not find any significantly 50  statistically difference in the above parameters between S-CXL and A-CXL groups (p > 0.05) ( Table 3). e changes in corneal total high ocular aberration (HOA) and root mean square values (RMS) compared with the baseline decreased significantly and statistically in both groups (p < 0.05 and p < 0.005, respectively) but did not differ significantly between S-CXL and A-CXL groups ( Table 4).         Journal of Ophthalmology ere was no statistically significant difference in the corneal stromal demarcation line depth between the two groups, with a mean depth of 214 ± 12.82 μm in the S-CXL group and 203 ± 12.03 μm in the A-CXL group (p � 0.0736).
None of our patients in none of the groups lost lines of BCVA. In each case in both groups, complete epithelization was accomplished in 3 days. Infections or melting were not noticed in any of the studied group cases. Haze was present in the majority of cases in both groups for about 3 to 6 months but decreased progressively after this period.

Discussion
With recent modifications of the original Dresden protocol, accelerated CXL has become one of the interesting topics in corneal surgery. It has been shown that "epi-off" standard CXL (Dresden protocol) leads to stabilization of KC, with flattening of topographic keratometry and improvement of visual acuity in many cases [19,[40][41][42][43][44].
In our study, we aimed to evaluate the evolution of patients up to 5 years, following "epi-off" standard and accelerated CXL procedures. Results of our study showed a similar reduction in Kavg, Ksteep, and Kflat in both accelerated and conventional study groups at all time points (p < 0.05), maintained at 5 years. In a comparative study of accelerated (30 mW/cm 2 for 3 minutes at 5.4 J/cm 2 ) versus conventional CXL (Dresden protocol), Tomita et al. [30] reported for the first time a significant flattening of keratometry measurements in both groups at 1 year, following the procedure, with no statistically significant difference between the study groups (ΔK � −0.62 D in the accelerated group and ΔK � −1.77 D in the conventional group, p � 0.21). ey also showed that the difference between the mean demarcation line depth in both groups was not statistically significant (294.38 ± 60.57 μm in the accelerated group and 380.78 ± 54.99 μm in the conventional group). Furthermore, in a report which compared the results of accelerated (7 mW/cm 2 irradiation 15 minutes protocol) versus conventional CXL (Dresden protocol), Kanellopoulos [45] reported the flattening of steep keratometry and the stabilization of KCN in both groups (from 49.5 D to 46.1 D in the accelerated group and from 48.7 D to 45.8 D in the conventional group). Similarly, Shetty et al. [46] revealed that the accelerated CXL (9 mW/cm 2 for 10 minutes and 18 mW/cm 2 for 5 minutes) had comparable outcomes to standard CXL, but the accelerated CXL using 30 mW/cm 2 for 3 minutes was not as efficient. In another study, in which 77 eyes treated with accelerated CXL and 76 treated with conventional CXL were enrolled and followed for 15 months, Hashemian et al. [47] showed a similar decrease in Ksteep at 15 months in both study groups (ΔKsteep � −1.98 D in the conventional group and    Conflicting studies showed that conventional CXL improved BCVA and decreased Kmax and Kmean, but the accelerated procedure provided unchanged BCVA, Kmax, and Kmean [52]. Moreover, Choi et al. [53] and Peyman et al. [54] revealed that an accelerated KCL procedure of 30 mW/cm 2 3 minutes 40 seconds provided a lower efficiency compared with the Dresden protocol, demonstrated by the depth of the demarcation line. It was postulated that the reduced effect of 30 mW/ cm 2 in 3-4 minutes is linked with the decrease in oxygen in these high-fluence treatments [55]. e introduction of pulse treatments aimed to restore oxygen in the cornea [55]. Our findings showed a statistically significant reduction in Kmax compared to baseline in both groups (p < 0.05) which changed during the 5 years of follow-up. We used this parameter not only to establish the progression of  [32] reported in a comparative study between two procedures of accelerated CXL (18 mW/ cm 2 in 5 minutes and 9 mW/cm 2 in 10 minutes) that Kmax is a useful index of CXL evaluation results, with a statistically significant decrease in accelerated CXL in 10 minutes. We observed a decrease in both cylinder and spherical equivalents after CXL in our samples, with no statistically significant difference in the change between the study groups. Our results are similar with those reported by Woo et al. [22] who enrolled 76 patients in a prospective study and reported no difference between accelerated and standard CXL beyond the 3 months after surgery. Hashemian et al. [28] and Kanellopoulos [45] have also reported a comparable reduction in spherical equivalent and cylindrical error in both accelerated and conventional CXL.
Results on UCVA and BVCA reported in the literature are conflicting for these techniques, with either no change or an improvement reported following CXL. Wittig-Silva [56] and Coskunseven et al. [57] showed in their studies an increase in UCVA and BCVA (p < 0.01) compared to controls. Ng et al. [52] showed no significant difference in the change of BCVA at 12 months after CXL in both conventional and accelerated CXL groups. Only UCVA showed mild but significant improvement of 0.13 lines. In our study, UCVA increased in the same manner in the conventional group and the accelerated CXL group, as showed by the higher improvement at years 1 and 2. However, the improvement became comparable from the first year after the procedure. Our findings revealed that BCVA in the conventional treated group had a higher improvement at 1 year compared to the accelerated one but stabilized thereafter. us, at 3, 4, and 5 years following the procedure, patients in the accelerated group presented an improved BCVA compared to the conventional group (p < 0.05 for both time points). e results in BCVA improvement in the conventional group are similar with those we published before [19].
In our study, we evaluated the HOA and RMS parameters, and we demonstrated a statistically significant reduction in both groups compared with baseline. Similarly, Greenstein et al. [58] revealed a significant decrease in HOA, total coma, three-order coma, and vertical coma. Moreover, Caporossi et al. [14] showed a significant decline of HOA and coma aberration as early and long-term results (2 years), following CXL. Kirgiz et al. [32] demonstrated a real improvement in coma values in the accelerated CXL with 10 minutes at 1-year follow-up.
Our findings showed a statistically significant reduction in the topographical index (TP), corneal volume, ISV, IHA, IHT, BAD_D, and ART Max from baseline, but there was no significant difference between the two groups (p < 0.05 at all time points). Similarly, Kirgiz et al. [32] revealed the same results regarding the above parameters.
Demarcation line was limited to the anterior-mid stroma until there was a maximum depth of 202 ± 12.03 μm in the accelerated group when compared with conventional CXL that reached 214 ± 12.82 μm.
Doors showed similar results [38] and described the best visibility of the corneal demarcation line using AS-OCT at 1 month after CXL treatment with an average of depth of 313 μm. Wollensak et al. [11] described cellular apoptosis to a depth of 300 μm radiating with UVA at 3 mW/cm 2 . Furthermore, similar results were demonstrated by Seiler and Hafezi [59], Kymionis et al. [60], Doors et al. [38], and Yam et al. [36]. In accelerated CXL, Seiler and Hafezi [59], Moramarco et al. [37], and Kymionis et al. [60] demonstrated a demarcation line depth of 213-215 μm after 1 month following the procedure. Similar to our observation, Kimionis et al. [60] noted that a 10-minute treatment with 9 mW/cm 2 resulted in a demarcation line that was less deep (288.46 ± 42.37 μm) compared to the standard procedure (350.75 ± 49.34 μm).
ey also reported that a modified accelerated protocol of CXL (9 mw/cm 2 for 14 minutes) provides the same demarcation line depth as the conventional procedure (294.38 ± 60.57 μm in the accelerated group and 380.78 ± 54.99 μm in the conventional group). e difference was not statistically significant. Kymionis [60] and Mazzota et al. [61] demonstrated a deep demarcation line of 280 μm in the pulsed-light cross-linking procedure applied for 6 minutes. is paper has several limitations that should be acknowledged. is was a retrospective study, and there was no randomization. However, the strength of our paper comes from the relatively large sample size compared to previous reports and on the long-term follow-up of these patients, adding to previous knowledge on this subject. To our knowledge, this is the first comparative study done in Romania and is the first one with such a long period of follow-up (5 years).

Conclusion
Our study revealed the efficacy and safety of long-term follow-up (5 years) in accelerated CXL in comparison with the conventional protocol. Following both protocols, a stabilization of KCN after 1 year was obtained.
Data Availability e data used and analyzed in the present study are available from the corresponding author.

Conflicts of Interest
e authors declare that they have no conflicts of interest.

Authors' Contributions
Cristina Ariadna Nicula conceived the study and was the major contributor in surgeries, in writing the manuscript, and in collecting data. Dorin Nicula participated in surgeries and contributed in its design and coordination. Anca Rednik helped to collect the data, made the statistical analysis, and drafted the manuscript. Adriana Elena Bulboaca made the analysis and interpretation of data. All authors have read and approved the final manuscript.