Long term results of accelerated 9 mW corneal crosslinking for early progressive keratoconus: the Siena Eye-Cross Study 2

Purpose To assess clinical results of the 9 mW/5.4 J/cm2 accelerated crosslinking (ACXL) in the treatment of progressive keratoconus (KC) over a span of 5 years. Methods The prospective open non-randomized interventional study (Siena Eye-Cross Study 2) included 156 eyes of 112 patients with early progressive KC undergoing the Epi-Off 9 mW/5.4 J/cm2 ACXL at the Siena Crosslinking Centre, Italy. The mean age was 18.05 ± 5.6 years. The 20-min treatments were performed using the New KXL I (Avedro, Waltham, USA), 10 min of 0.1% HPMC Riboflavin soaking (VibeX Rapid, Avedro, Waltham, USA) and 10 min of continuous-light UV-A irradiation. Uncorrected distance visual acuity (UDVA), corrected distance visual acuity (CDVA), Kmax, coma, minimum corneal thickness (MCT), surface asymmetry index (SAI), endothelial cell count (ECC) were measured, and corneal OCT performed. Results UDVA and CDVA improved significantly at the 3rd (P = 0.028), Δ + 0.17 Snellen lines and 6th postoperative month, respectively (P < 0.001), Δ + 0.23 Snellen lines. Kmax improved at the 6th postoperative month (P = 0.03), Δ − 1.49 diopters from the baseline value. Also, coma aberration value improved significantly (P = 0.004). A mild temporary haze was recorded in 14.77% of patients without affecting visual acuity and without persistent complications. Corneal OCT revealed a mean demarcation line depth at 332.6 ± 33.6 μm. Conclusion The 5-year results of Epi-Off 9 mW/5.4 J/cm2 ACXL demonstrated statistically significant improvements in UCVA and CDVA, corneal curvature and corneal higher-order aberrations which confers a long-term stability for progressive ectasia. Based on the results of the Siena Eye-Cross Study 2, the 9 mW/5.4 J/cm2 ACXL is a candidate to be the natural evolution of Epi-Off CXL treatment for the management of early progressive corneal ectasia, and thus optimize clinic workflow.

We report the long-term (5-years) clinical results of the 9 mW/5.4 J/cm 2 ACXL protocol also named "Dresden Accelerated Protocol" in a large cohort of patients performed in Italy at the Siena Crosslinking Centre named "Siena Eye-Cross Study 2".

Dataset and study design
The prospective long-term open non-randomized, noncomparative interventional Siena Eye-Cross Study 2 was approved by the institutional review board (IRB) of the Siena Crosslinking Center following the tenets of the Declaration of Helsinki and included 156 eyes of 112 patients who underwent an Epi-Off 9 mW/5.4 J/cm 2 ACXL procedure for progressive KC and completed the 5-year follow-up. All patients were affected by progressive (stage I and II) KC

Inclusion criteria
Progressive stage II KC was identified according to Krumeich's staging system [38]. Progression of KC was defined as an increase in apical keratometry (AK) ≥ 1 diopter (D) on the anterior corneal topography using a Scheimpflug-Placido corneal tomography system Sirius, Costruzione Strumenti Oftalmici (C.S.O.), Florence, Italy; minimum corneal thickness (MCT) reduction ≥ 10 μm; worsening of uncorrected distance visual acuity (UDVA) and corrected distance visual acuity (CDVA) ≥ 0.1 decimal equivalent or change of ≥ 0.5 in mean refractive spherical equivalent (MRSE) in the last 6 months of clinical and instrumental observation [39]. MCT was at least 400 μm (epithelium included), measured with no-contact optical pachymetry. Corneas were clear with no sub-apical opacities or scars and no markedly visible Vogt's striae, no previous infectious keratitis or autoimmune diseases and no severe dry eye. All patients included in the study gave their specific written informed consent.

Surgical procedure
All treatments were performed at the Siena Crosslinking Centre, Italy, by the same surgeon (CM). The 9 mW/5.4 J/cm 2 ACXL treatment was performed using the KXL I system (Avedro, Waltham, MS, USA) under topical anesthesia (4% Oxibuprocaine chlorydrate 1.6 mg/0.4 mL drops) that was applied 10 min before the treatment after a premedication with 2% pilocarpine instilled 30 min before the operation. After applying a closed valves eyelid speculum, a 9-mm diameter epithelium was removed with a blunt metal spatula. After epithelial removal, a dextran-free plus hydroxyl-propyl methylcellulose (HPMC) disposable 0.1% riboflavin isotonic solution (VibeX Rapid, Avedro, Waltham, MA, USA) was instilled for one drop every minute for 10 min of corneal soaking, before starting continuous light UV-A irradiation. During UV-A-irradiation, 2 drops of riboflavin solution were administered every 2.5 min for a total of 10 min of UV-A exposure at 9 mW/cm 2 of UV-A power and standard Fluence of 5.4 J/cm 2 . At the end of the UV-A irradiation, the cornea was washed with balanced saline solution (BSS) and medicated with preservativefree netilmicin plus dexamethasone, cyclopentolate eyedrops and dressed with a therapeutic bandage soft contact lens for 4 days. After therapeutic contact lens removal, fluorometholone 0.2% eye-drops (tapered 3 times/day) and sodium hyaluronate 0.2% lacrimal substitutes were administered for 6 to 8 weeks.

Measurements and devices
Ophthalmic evaluations were performed before CXL and at all follow-up visits (1, 3, 6, 12, 24, 36, 48 and 60 months). The evaluation included UDVA, CDVA and a slit-lamp clinical examination. Scheimpflug based corneal tomography (Sirius, CSO, Florence, Italy) was used to measure maximum curvature simulated k reading (Kmax), coma high-order aberration, MCT, tomography derived SAI and topographic cylinder (CYL). Anterior segment optical coherence tomography (AS-OCT) with the I-Vue (Optovue, Freemont CA, USA) was performed to assess the demarcation line depth at the first post-operative month. Endothelial cell count was measured by the I-Konan Non-Co Robot (Konan Inc., Hyogo, Japan) preoperatively and every year until the end of the study.

Statistical analysis
A two-tailed paired samples t test was used to compare each baseline measurement with the respective followup measurements. Differences with P < 0.05 were considered statistically significant. Data were collected and analyzed with PRISM 6.0 GraphPad Software (La Jolla, California, USA).

Baseline data
Visual acuity was represented in Snellen lines and decimal equivalents (d.e.  Table 1. UDVA improved from baseline, becoming statistically significant at the 3rd postoperative month (P = 0.028) and remained significant until the end of follow-up (Fig. 1).
CDVA showed a statistically significant improvement at the 3rd (P < 0.001) postoperative month, remaining significant until the end of follow-up, (Fig. 2).
The baseline and follow-up measurements of Kmax documented that the value improved significantly at the 6th postoperative month, from 55.7 D to 54.36 D (P = 0.03) and this improvement remained statistically significant up to the 5th year follow-up (P = 0.001; delta − 1.49 D, Fig. 3).
The coma high order aberration mean values improved during the follow-up, becoming statistically  significant at the 1st postoperative month (P = 0.004, Fig. 4). Tomographic mean MCT value decreased significantly at the 1st (P = 0.0001) and 3rd postoperative month (P = 0.03), returning to baseline at the 6th month and maintaining a substantial stability up to 5-years followup (Fig. 5).
The topographic SAI values showed a significant change at the 1st month (P = 0.04), (Fig. 6).
The CYL average values remained stable along the entire follow-up in the whole study cohort, despite a positive trend. In the pediatric group, the CYL values showed a temporary significant improvement at the 1st month (P = 0.034) but became insignificant after the 3rd postoperative month (Fig. 7).
ECC showed no significant postoperative variations during any follow-up point (P > 0.05).
AS-OCT evaluation of the demarcation line depth at the 1st postoperative month showed a depth of 332.6 ± 23.6 μm in the overall study cohorts (Fig. 8).
The overall follow-up results are reported in Table 2.

Adverse events
No postoperative infections, persistent haze or endothelial cell failure were encountered during the entire follow-up period in the whole study cohort. In 16 eyes (10.25%) of 13 patients, a mild temporary haze at the 1st postoperative month at slit lamp examination was observed. However, this did not affect visual acuity and disappeared between the 3rd and the 6th month visit once topical steroids therapy was started (preservative-free 1 mg/mL dexamethasone eye-drops, tapered 4 times a day for 4 to 8 weeks). Thirteen eyes (8.33%) of 8 patients had a Kmax progression of 1 D within the 2nd-3rd year follow-up visits, returning to baseline value after 30 ± 6 months. All patients with such progression were affected with severe allergic papillary conjunctivitis. No retreatment was performed in the entire 5-year follow-up period. Data for adverse events are reported in Table 3.

Discussion
The 5-years follow-up results after accelerated epitheliumoff CXL (9 mW/cm 2 for 10 min) in the broadest case series of 156 KC eyes of 88 consecutive patients and with the longest follow-up currently reported internationally, showed clinically significant improvements in UDVA and CDVA, corneal curvature reduction with typical central flattening of the apical region of the KC with paracentral compensatory steepening of the flattest areas, inducing an improved corneal symmetry and corneal higher-order aberrations reduction (Fig. 9). In addition, the procedure was well tolerated in all patients, conferring a longterm (over 5-years) stability for progressive KC. No significant adverse events such as postoperative infections, persistent haze or endothelial cell failure were encountered during the entire follow-up period. A mild temporary haze that did not affect visual acuity was observed at the 1st postoperative month in around 15% of patients, disappearing after the 3rd month of clinical observation after topical steroid therapy. Maximum corneal curvature progression of 1 D, which returned to baseline, was recorded between the 24th and 36th month follow-up in 13 eyes of 8 patients (8.33%). Patients that returned to baseline after initial functional improvement were all affected with allergies and were eye-rubbers. Interestingly, despite no retreatments being performed in the 5-years follow-up, none of these patients showed worsening in either UDVA or CDVA. Severe allergic conjunctivitis, eye-rubbing, and blepharitisassociated ocular surface chronic inflammation, however, were recognized as possible causes of poorer response and corneal topography instability [40].
The 5-years progression, intended as Kmax deterioration of 1 D or more occurred in 8.33% of patients, however, the improvement in CDVA did remain significant.
AS-OCT evaluation of the demarcation line depth at the 1st postoperative month showed a depth of 332.6 ± 33.6 μm in the overall study cohort. To ensure a longlasting stability of keratoconus and secondary ectasia, UV-A power settings and exposure time were targeted to allow a treatment penetration at least at 250 μm. Such depth of treatment penetration would allow a large portion of the anterior stroma to be cross-linked, which would be desirable especially since the anterior 40% of the central corneal stroma is the stiffest region of the cornea. Moreover, according to Kohlhaas et al. [41], the CXL treatment significantly stiffens the cornea only in the anterior 200 μm. The depth-dependent stiffening effect can be explained by the absorption behavior for UV-A on a riboflavin-soaked cornea, where about 70% of UVA irradiation is absorbed within the anterior 200 μm and only 20% in the next 200 μm [42]. This aspect may affect longterm ectasia stabilization according to different crosslinking penetration rate [43]. While maintaining a standard 5.4 J/cm 2 dose delivery, it was demonstrated by Mazzotta et al. in the "M nomogram" for the standardized treatment of all thickness ectatic corneas including the thin and ultrathin KC maintaining the standardized fluence of the original Dresden protocol (5.4J/cm 2 ) in all cases, that UV-A power can be calibrated between 9 and 15 mW/cm 2 Fig. 6 Topographic SAI showed a significant change at the 1st postoperative month (P = 0.04) with continuous or pulsed-light mode of exposure [44]. Indeed, the stress-strain testing data from an experimental study performed by Krueger et al. [29] demonstrated a substantial biomechanical equivalence between 3 mW/5.4 J/cm 2 CXL for 30 min of continuous UVA light exposure, 9 mW/5.4 J/cm 2 ACXL for 10 min of continuous UVA light exposure and 15 mW/5.4 J/cm 2 ACXL with continuous or pulsed-light exposure [25].
Despite laboratory data from Richoz et al. [32] showing lower stiffening effect due to oxygen availability in ACXL procedures, the 9 mW/5.4 J/cm 2 was estimated to have the best oxygen diffusion profile in the ACXL procedures panorama. Several clinical studies also demonstrated that the 9 mW/5.4/cm 2 Epi-Off ACXL obtained comparable visual outcomes and KC stabilization with conventional 3 mW/5.4 J/cm 2 CXL (Dresden protocol). Moreover, the AS OCT and in vivo confocal microscopy (IVCM) studies by Mazzotta et al. [45] showed that the mean depth of the demarcation line after the 9 mW/5.4 J/cm 2 ACXL was at 332 ± 20 μm, the nearest to the conventional 3 mW/5.4 J/ cm 2 CXL Dresden protocol 320 μm (range 270-350 μm) [1,4].  The 9 mW/5.4 J/cm 2 ACXL for 10 min of continuous light UVA exposure was first demonstrated to be effective in stabilizing topographic parameters after 12-months of follow-up in mild-moderate KC affected corneas by Elbaz et al [17]. An improvement in the UDVA and stabilization of all tested corneal parameters were noted after the treatment. Moreover, the 9 mW/5.4 J/cm 2 ACXL was safe for corneal endothelium, stabilizing the progression of iatrogenic ectasia with a significant reduction in topographic keratometric values and a significant increase in CDVA, comparable with conventional 3 mW/cm 2 CXL in a midterm (two-years) follow-up as recently documented by Turhan et al [37]. A recent study performed by Sadoughi et al. [33] comparing the conventional 3 mW/5.4 J/cm 2 CXL to 9 mW/5.4 J/cm 2 ACXL in patients with bilateral progressive KC where the fellow eyes were randomly assigned to conventional 3 mW/5.4 J/cm 2 CXL or to 9 mW/5.4 J/cm 2 ACXL, revealed similar refractive, visual, keratometric and aberrometric outcomes after 12 months of follow-up. The study confirmed that the one and twoyears functional outcomes of the 9 mW/5.4 J/cm 2 ACXL protocol were similar and comparable with the conventional 3 mW/5.4 J/cm 2 CXL.
In a comparative study of 29 eyes treated with 9 mW/cm 2 for 10 min, Kirgiz A et al. [35] reported that ACXL using 10 min of UVA irradiance at 9 mW/cm 2 showed better topographic and coma values improvements vs. 5 min of UVA at 18 mW/cm 2 irradiance, independent of keratoconus severity. Furthermore, Lang et al. [34] in a comparative study using standard CXL protocol and accelerated protocols in patients with progressive keratoconus reported that Epi-Off 3 mW/5.4 J/cm 2 CXL for 30 min, 5.4 J/ cm 2 and 9 mW/5.4 J/cm 2 ACXL for 10 min showed similar improvements in Kmax and CDVA.
Kobashi et al. [36] in a recent meta-analysis compared the clinical results of ACXL and standard corneal collagen cross-linking (SCXL) in progressive keratoconus of randomized controlled trials, which showed a comparable efficacy and safety profile at the 1-year follow-up between the two procedures.
Besides performing stress-strain experiments, CXL efficacy has also been shown to drop from SCXL to short ACXL protocols using inflation tests or air-puff tonometry. SCXL and ACXL have shown similar results [46,47].
Although 9 mW/5.4 J/cm 2 ACXL for 10 min provides less biomechanical increase under laboratory conditions [32], it is not known exactly how much stiffening each keratoconic cornea would need for the progressive character of the disease to be stabilized.
Our study data suggest that 9 mW/cm 2 for 10 min may be sufficient to efficiently prevent progression. Nevertheless, we are currently evaluating in the  laboratory whether a high-fluence ACXL Epi-Off setting using 9 mW/7.2 J/cm 2 for 13 min and 20 s would further increase the biomechanical stability. If this were the case, then an increased fluence might provide an additional level of efficacy to the treatment.

Conclusion
The 5-years results of the 9 mW/cm 2 Epi-Off ACXL with 5.4 J/cm 2 energy, evaluated in the Siena Eye-Cross Study 2, demonstrated statistically significant improvements of UCVA, CDVA, corneal curvature and corneal higherorder aberrations, and thus confers a long-term stability for progressive KC. In our series, bilateral treatments were performed in 56.4% of patients and the second eye was operated after a minimum time interval of 30 days to a maximum of 60 days (mean 40 days). If keratoconus progression is documented in both eyes, the treatment of the second eye should be performed in a timely fashion, especially in pediatric patients to avoid further risk of progression and higher economic burden [48]. The 9 mW/5.4 J/ cm 2 for 10 min ACXL protocolsupported by photochemistry, microstructural and long-term clinical databecame our Epi-Off CXL treatment of choice, which reduces CXL treatment time from 1 h to 20 min allowing better patient comfort while maintaining overall CXL safety and efficacy. According to the Siena Eye-Cross Study 2, this accelerated crosslinking protocol is a candidate to be the natural evolution of the original Dresden Epi-Off CXL treatment for the management of early progressive corneal ectasia, thus optimizing clinic workflow, and patient compliance being really efficiacious, less time-consuming and more cost-effective.