Regional Biomechanical and Topographic Changes after Transepithelial vs. Epithelium-off Continuous Accelerated Corneal Cross-linking in Keratoconus: Updated Stress-Strain Index as a Superior Biomarker

Abstract

Purpose

To compare regional biomechanical and topographic outcomes between transepithelial (Epi-on) and epithelium-off (Epi-off) continuous accelerated corneal cross-linking (ACXL) in keratoconus.

Design

A retrospective cohort study.

Subjects

Seventy-seven keratoconic eyes were included: 45 with Epi-on ACXL and 32 with Epi-off ACXL.

Methods

All eyes underwent ACXL (10 mW/cm², 7.2 J/cm²). Pentacam HR and Corvis ST measurements were obtained preoperatively (pre), and at 1 month (pos1m), and 6 months (pos6m) postoperatively. The SSIv2m, M, and J₀, J₄₅ were evaluated across central (0–3 mm), paracentral (3–6 mm), and peripheral (6–9 mm) regions. Correlated measurements were analyzed using generalized estimating equations for regional and longitudinal data, and generalized linear models for intergroup comparisons, with both absolute (Δ) and relative changes (Δ%) evaluated.

Main Outcome Measures

Regional changes in SSIv2_m, M, J₀, and J₄₅ were analyzed across central (0–3 mm), paracentral (3–6 mm), and peripheral (6–9 mm) regions preoperatively and at 1 and 6 months postoperatively.

Results

At 6 months, SSIv2_m increased significantly in all regions of the Epi-off ACXL group (greatest centrally), whereas the Epi-on group showed no consistent change. ΔSSIv2_m between preoperative and pos6m was significantly greater in Epi-off ACXL than Epi-on ACXL across all regions (central: P < 0.01; paracentral and peripheral: P < 0.05). Central flattening (negative ΔM) was significant in the Epi-off ACXL group (P < 0.001) and exceeded that in Epi-on ACXL (P < 0.001). No significant changes were detected in J₀ or J₄₅. Δ%SSIv2_m exceeded Δ%M in the paracentral and peripheral (paracentral: P < 0.01; peripheral: P < 0.05) regions after Epi-off ACXL.

Conclusions

Epi-off ACXL induces stronger and more durable biomechanical stiffening and central flattening than Epi-on ACXL. Biomechanical parameters, particularly SSIv2, were more sensitive than curvature-based metrics for detecting regional treatment effects.

Financial Disclosures

Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.

Keratoconus is a progressive ectatic corneal disorder characterized by stromal thinning and anterior protrusion, leading to progressive visual impairment. Researchers in the Raine Study reported a prevalence of keratoconus as high as 1 in 84.¹ Corneal collagen cross-linking (CXL) is the primary intervention to halt the progression of keratoconus.^2,3 The accelerated CXL (ACXL) protocol, using reduced exposure time at the same total dose,⁴ is now widely used. Published evidence5, 6, 7 have established that both standard and ACXL protocols are safe and effective for stabilizing keratoconus. The conventional epithelium-off (Epi-off) technique, introduced by Wollensak et al,⁸ is highly efficacious but carries notable risks such as infection and haze.8, 9, 10 In contrast, the transepithelial (Epi-on) approach was developed to improve safety and patient comfort, though variable efficacy due to limited riboflavin penetration may necessitate retreatment.^11,12

The distribution of ultraviolet A (UVA) irradiance across the cornea is nonuniform because the central cornea receives vertical exposure (intensity I₀), while the peripheral cornea is irradiated at an incident angle α. Effective irradiance received peripherally follows the cosine law (I' = I₀·cosα) and is reduced.13, 14, 15 In vivo and ex vivo studies consistently demonstrate that the CXL demarcation line is deeper centrally than peripherally, and it has been reported that following ACXL, this line is shallower compared to standard CXL.^15,16 Regional differences in postoperative corneal curvature changes have been documented.17, 18, 19 These observations highlight the importance of considering regional effects when evaluating CXL outcomes.

While curvature-based regional responses to CXL are relatively well characterized, corresponding regional biomechanical changes have been less extensively studied. Devices like the Corvis ST, with parameters including the stress–strain index (SSI), now enables in vivo quantification of localized biomechanical properties. Stress–strain index maps²⁰ allow in vivo quantification of localized biomechanical properties and visualize regional stiffness variation. Integrated analyses of regional biomechanics and corneal morphology are poised to yield a more comprehensive understanding of the effects of CXL.

However, direct comparative analyses of how these regional biomechanical and topographic changes differ between Epi-on and Epi-off ACXL protocols remain limited. This study aimed to evaluate regional biomechanical and topographic responses following ACXL using Epi-on and Epi-off protocols in keratoconic eyes by employing a region-specific biomechanical analysis with the SSI map under identical accelerated protocols, thereby providing insight into how epithelium status influences the spatial distribution of corneal stiffening. We hypothesized that (1) regional changes would be more pronounced in the central cornea compared with paracentral and peripheral zones, (2) Epi-off ACXL would induce greater biomechanical and morphological changes than Epi-on ACXL, and (3) biomechanical indices would demonstrate greater sensitivity than topographic metrics in detecting postoperative regional changes.

Methods

Study Design and Participants

This retrospective cohort study adhered to the tenets of the Declaration of Helsinki and received ethics approval from the institutional review board of the Eye Hospital, Wenzhou Medical University. All participants provided written informed consent for the research use of their medical data. This consent was obtained prior to their CXL procedures and included the understanding that data would be anonymized for analysis. The study included 77 eyes from patients with progressive keratoconus treated at the Eye Hospital. Forty-five eyes underwent transepithelial continuous accelerated corneal cross-linking (Epi-on ACXL) and 32 eyes underwent de-epithelialization continuous accelerated corneal cross-linking (Epi-off ACXL). The surgical technique was selected based on preoperative corneal morphology (e.g., corneal curvature and pattern of irregular astigmatism), systemic health, and residual stromal thickness. Patient tolerance and counseling regarding trade-offs between invasiveness and efficacy were also considered. Consequently, baseline central corneal thickness (CCT) was potentially imbalanced between groups. A comprehensive summary of the procedural parameters is provided in Table 1.

Table 1.

Two ACXL Groups Included in Study

Epithelial Status	Intensity (mW/cm2)	Treatment Time (s)	Power (J/cm2)	Working Distance (cm)	Riboflavin Concentration	Soak Time (Min)	Group Name	No. of Eyes
Epi-on	10	720	7.2	5	0.25%	30	Epi-on ACXL	45
Epi-off	10	720	7.2	5	0.1%	30	Epi-off ACXL	32

Epi-on ACXL = transepithelial accelerated corneal cross-linking; Epi-off ACXL = de-epithelial accelerated corneal cross-linking.

Inclusion criteria in the study were a confirmed diagnosis of keratoconus with documented progression, defined as an increase of ≥1.00 D in maximum keratometry (K_max), an increase of ≥1.00 D in manifest astigmatism, or a decrease of ≥10 μm in minimum corneal thickness.^3,18,21 Exclusion criteria were corneal thickness <400 μm at the thinnest point, endothelial cell density <2000 cells/mm², severe corneal scarring or ocular surface disease, autoimmune disease, or pregnancy.²² The patient enrollment process is summarized in the CONSORT flow diagram (Fig 1).

Figure 1.

Patient selection flow diagram (based on CONSORT guidelines).

Data Collection and Imaging

Demographic data (age and sex) were recorded. Baseline corneal biomechanics were measured with the Corvis ST (Oculus Optikgeräte GmbH; version 1.3b1445), including the updated stress–strain index (SSIv2) and biomechanically corrected intraocular pressure (bIOP). A stiffness distribution map (SSI map),²⁰ which integrates collagen fiber distribution with in vivo biomechanical parameters, was used to calculate mean SSIv2 (SSIv2_m) within 3 corneal regions centered on the corneal center: central (0–3 mm), paracentral (3–6 mm), and peripheral (6–9 mm) (Fig 2). Baseline CCT and topographic keratoconus classification grade were obtained with a Pentacam HR (Oculus Optikgeräte GmbH; version 1.22r05). Epithelial thickness was measured by anterior segment OCT (RTVue XR, Optovue). Pentacam scans were repeated at 1 month and 6 months postoperatively. Only scans with a quality specification “OK” were included in analysis.

Figure 2.

The three corneal regions considered in study.

Surgical Protocols

Epi-on ACXL: After topical anesthesia (0.5% proparacaine hydrochloride), riboflavin 0.25% (Peschke M) was applied every 1 min for 30 min under controlled ambient conditions. The cornea was rinsed with saline and irradiated with continuous UVA (365 nm) using the Avedro KXL system (10 mW/cm², total fluence 7.2 J/cm², beam diameter 9 mm, distance 5 cm). A bandage contact lens (Bausch & Lomb) was placed for 1 to 3 days, and postoperative therapy included topical antibiotic–steroid, tapering fluorometholone, and artificial tears.²²

Epi-off ACXL: After topical anesthesia, the epithelium was removed in phototherapeutic keratectomy mode using the Schwind Amaris 1050 excimer laser, guided by epithelial thickness on anterior segment OCT, with ablation depth calculated as the measured central epithelial thickness plus 10 μm (9 mm treatment zone). Riboflavin 0.1% (Peschke M) was instilled for 30 min until stromal saturation was observed. Continuous UVA irradiation was identical to the Epi-on ACXL protocol. All other standard procedures have been consistently followed. A bandage contact lens was placed until full re-epithelialization (3–7 days). Postoperative 0.1% fluorometholone eye drops were applied 4 times daily for 1 month and tapered by 1 application per month thereafter.²³

Given evidence that higher irradiance can reduce biomechanical efficacy under constant total energy, we therefore applied a moderately increased total fluence of 7.2 J/cm² to support adequate cross-linking efficiency.²⁴ The efficacy of continuous UVA irradiation (10 mW/cm² for 12 min, 7.2 J/cm²) has been validated in our prior study.²⁵ The total energy dose of 7.2 J/cm² was selected for both ACXL protocols based on its established role as the standard of care, with its safety and efficacy specifically demonstrated in an accelerated continuous cross-linking protocol for patients with progressive keratoconus.26, 27, 28

Derivation of Corneal Shape Parameters

Elevation data from the Pentacam were exported at 0.1-mm intervals in both temporal–nasal and superior–inferior directions across a 9-mm central corneal zone. Zernike polynomials up to the 10th order were fit to the data.²⁹ Principal curvatures and orientations were derived from first and second derivatives. The curvature parameter M (D), and astigmatism components J₀ (with-the-rule/against-the-rule) and J₄₅ (oblique) were calculated29, 30, 31 in three concentric corneal zones as described above (Fig 2). The corneal apex, identified as the central point of the topographic map, served as the reference center for defining the concentric analysis zones. For each region, SSIv2, M, J₀, and J₄₅ were averaged at baseline (pre), 1 month (pos1m), and 6 months (pos6m). And the averaged SSIv2 in each region was denoted as SSIv2_m.

Calculation of Changes

Absolute change from baseline was calculated as:

$$\Delta \mathrm{X}={\mathrm{X}}_{\text{pos}}–{\mathrm{X}}_{\text{pre}}$$

where X_pre is baseline and X_pos is follow-up (1 or 6 months). Parameter-specific differences were denoted as ΔSSIv2_m, ΔM, ΔJ₀, ΔJ₄₅, etc.

$$\text{Relative}\text{change}\text{was}\text{calculated}\text{as}:\Delta \%\mathrm{X}=\frac{{\mathrm{X}}_{\text{pos}}–{\mathrm{X}}_{\text{pre}}}{{\mathrm{X}}_{\text{pre}}}\ast 100\%\text{,}$$

denoted as Δ%SSIv2_m, Δ%M, etc.

Statistical Analysis

Analyses were performed with SPSS v24.0 (IBM Inc.). Continuous variables are reported as mean ± standard deviation. Baseline characteristics were compared by analysis of variance (ANOVA) (continuous variables) or Fisher exact test (categorical variables). Generalized estimating equations (GEEs) with an exchangeable working correlation structure assessed longitudinal changes in SSIv2_m within and between groups, adjusting for ΔCCT and ΔbIOP. Repeated measures ANOVA was used to test changes in curvature and astigmatism. Between-group and interregional comparisons of Δ and Δ% were conducted with GEE and generalized linear models. To exclude the potential confounding effects of variations in CCT and bIOP on SSIv2_m, this study incorporated ΔCCT and ΔbIOP as covariates into the corresponding GEE analysis models. The reported adjusted ΔSSIv2_m values represent the model-estimated mean changes after accounting for these variables. A Bonferroni correction was applied to P values from the primary inferential models (GEE, generalized linear model, and repeated-measures ANOVA) to control the family-wise error rate. Statistical significance was determined based on Bonferroni-adjusted P values compared against an α level of 0.05. All other P values (e.g., from baseline comparisons) are reported without adjustment. A post hoc power analysis was performed using G∗Power software to quantify the statistical power for detecting the observed effect size in the primary outcome (intergroup difference in central ΔSSIv2_m), given the study's retrospective design and a two-tailed α of 0.05.

Results

Baseline demographic and clinical data are summarized in Table 2. Age, sex distribution, and bIOP did not differ significantly between groups. The CCT was lower in the Epi-on ACXL group (477.7 ± 40.2 μm) compared with the Epi-off ACXL group (493.7 ± 31.3 μm, standardized mean difference of Hedges’ g = –0.43, 95% confidence interval [CI] [-0.89, 0.03], P = 0.069). Topographic keratoconus classification grade distribution was similar across groups (P = 0.759).

Table 2.

Demographic Data of Study Participants

Groups	Epi-on ACXL	Epi-off ACXL	P
Age (years)	20.0 ± 4.9	20.0 ± 4.9	0.962
Sex (Male/Female)	33/12	24/8	0.869
bIOP (mmHg)	13.2 ± 2.30	13.8 ± 2.59	0.305
CCT (um)	477.7 ± 40.2	493.7 ± 31.3	0.069
TKC Grade			0.427
1	6	6	-
2	19	14	-
3	19	12	-
4	1	0	-

Epi-on ACXL = transepithelial accelerated corneal cross-linking; Epi-off ACXL = de-epithelial accelerated corneal cross-linking; bIOP = biomechanically corrected intraocular pressure; TKC = topographic keratoconus classification; CCT = central corneal thickness.

During follow-up, both groups demonstrated significant reductions in CCT from baseline to both 1 and 6 months (all P < 0.001). A significant increase in bIOP was observed exclusively in the Epi-off ACXL group at 1 month postoperatively compared with preoperative values (P < 0.05). Hence, ΔCCT and ΔbIOP were included as covariates in the subsequent analyses of biomechanical outcomes to control for their potential influence. Given the theoretical independence of topographic parameters from CCT³² and bIOP, the analyses of M, J₀, and J₄₅ were performed without covariance adjustment.

For the primary outcome (central ΔSSIv2_m), post hoc power analysis indicated > 99% power to detect the observed intergroup effect (Cohen's d = 2.42) at α = 0.05, given the sample size (N = 77), far exceeding the 80% convention. However, the repeated-measures ANOVA for J₀ and J₄₅ were underpowered at the observed effect size, precluding definitive conclusions regarding clinically meaningful intragroup differences.

Changes of SSIv2 among Different Regions

Central region (0 - 3 mm)

After adjustment for ΔCCT and ΔbIOP, both groups exhibited a significant early elevation in SSIv2_m at 1 month relative to baseline (Epi-on ACXL: adjusted ΔSSIv2_m [pos1m - pre]: 0.0264 ± 0.0109, 95% CI [0.0003, 0.0526], P < 0.05; Epi-off ACXL: adjusted ΔSSIv2_m [pos1m - pre]: 0.0352 ± 0.0140, 95% CI [0.0017, 0.0686], P < 0.05) (Fig 3A, B), with no further change observed at 6 months. Over the entire study period, only the Epi-off ACXL group maintained a significant net increase in SSIv2_m from baseline to 6 months (adjusted ΔSSIv2_m [pos6m - pre]: 0.0408 ± 0.0124, 95% CI [0.0112, 0.0704], P < 0.01).

Figure 3.

Boxplots of SSIv2_m and corneal curvature M changes before and 1-month, and 6-month after CXL; (A) changes of SSIv2_m after Epi-on ACXL, (B) changes of SSIv2_m after Epi-off ACXL, (C) changes of M after Epi-on ACXL, (D) changes of M after Epi-off ACXL; ^∗P < 0.05, ^∗∗P < 0.01, ^∗∗∗P < 0.001. ACXL = accelerated corneal cross-linking.

Paracentral region (3 - 6 mm)

SSIv2_m exhibited a significant increase at 1 month postoperatively in both group (Epi-on ACXL: adjusted ΔSSIv2_m [pos1m - pre]: 0.0265 ± 0.0097, 95% CI [0.0033, 0.0497], P < 0.05; Epi-off ACXL: adjusted ΔSSIv2_m [pos1m - pre]: 0.0340 ± 0.0110, 95% CI [0.0077, 0.0602], P < 0.01). This initial increase was followed by a significant decline between 1 and 6 months exclusively in the Epi-on ACXL group (P < 0.01). Consequently, at the 6-month mark, only the Epi-off ACXL group sustained a significant net increase from baseline (adjusted ΔSSIv2_m [pos6m - pre]: 0.0229 ± 0.0092, 95% CI [0.0009, 0.0449], P < 0.05).

Peripheral region (6 - 9 mm)

SSIv2_m exhibited a similar trend, with both groups showing a significant increase at 1 month postoperatively (Epi-on ACXL: adjusted ΔSSIv2_m [pos1m - pre]: 0.0312 ± 0.0092, 95% CI [0.0093, 0.0531], P < 0.01; Epi-off ACXL: adjusted ΔSSIv2_m [pos1m - pre]: 0.0400 ± 0.0130, 95% CI [0.0089, 0.0712], P < 0.01). The Epi-on ACXL group subsequently demonstrated a significant decline between 1 and 6 months (P < 0.01). Conversely, the Epi-off ACXL group maintained a significant net increase relative to baseline at 6 months (adjusted ΔSSIv2_m [pos6m - pre]: 0.0288 ± 0.0110, 95% CI [0.0024, 0.0552], P < 0.05).

Changes in Corneal Curvature M

Central region (0 - 3 mm)

From baseline to 1 month, corneal curvature (M) steepened significantly in the Epi-on ACXL group (ΔM [pos1m - pre]: 0.677 ± 0.119 D, 95% CI [0.379, 0.975], P < 0.001), while no significant change was observed in the Epi-off ACXL group (P > 0.05) (Fig 3C, D). Between 1 and 6 months, both groups exhibited significant flattening (both P < 0.001). Over the full 6-month period, only the Epi-off ACXL group showed a significant net central flattening compared with baseline (ΔM [pos6m - pre]: –1.008 ± 0.208 D, 95% CI [–1.539, –0.478], P < 0.001).

Paracentral region (3 - 6 mm)

In both groups, corneal flattening occurred between baseline and 1 month (Epi-on ACXL: ΔM [pos1m - pre]: –0.241 ± 0.052 D, 95% CI [–0.371, –0.111], P < 0.001; Epi-off ACXL: ΔM [pos1m - pre]: –0.270 ± 0.102 D, 95% CI [–0.530, –0.010], P < 0.05), followed by significant steepening from 1 to 6 months (Epi-on ACXL: P < 0.001; Epi-off ACXL: P < 0.05). Overall, compared with the preoperative baseline, no net change in the paracentral region was observed in the eyes of either group at 6 months postoperatively (all comparisons with P > 0.05).

Peripheral region (6 - 9 mm)

Both groups exhibited significant corneal flattening from baseline to 1 month (Epi-on ACXL: ΔM [pos1m - pre]: –0.239 ± 0.067 D, 95% CI [–0.406, –0.071], P < 0.01; Epi-off ACXL: ΔM [pos1m - pre]: –0.432 ± 0.125 D, 95% CI [–0.751, –0.114], P < 0.01). This flattening remained stable between 1 and 6 months, with no further significant changes observed in either group (all P > 0.05). Consequently, neither group showed a statistically significant net change in flattening at 6 months compared to baseline (all P > 0.05).

Changes in Corneal Astigmatism J₀ and J₄₅

Across all three regions (central, paracentral, and peripheral), no statistically significant changes in J₀ or J₄₅ were detected between baseline, 1 month, and 6 months in either the Epi-on ACXL or Epi-off ACXL groups (all P > 0.05). Similarly, no significant between-group differences were observed for astigmatism changes at any time point.

Comparisons of Absolute and Relative Changes between Groups and Regions

Absolute Changes (Δ)

ΔSSIv2_m (6 months vs. baseline)

In the Epi-on ACXL group, no significant regional differences were observed (P > 0.05). In contrast, in the Epi-off ACXL group, central ΔSSIv2_m (unadjusted ΔSSIv2_m: 0.0211 ± 0.0047, 95% CI [0.0119, 0.0302]) was significantly greater than paracentral ΔSSIv2_m (unadjusted ΔSSIv2_m: 0.0168 ± 0.0055, 95% CI [0.0060, 0.0275], P < 0.05).

ΔM (6 months vs. baseline)

No regional differences were found in the Epi-on ACXL group (P > 0.05). In the Epi-off ACXL group, central flattening (ΔM) (unadjusted ΔM: –1.0038 ± 0.2032 D, 95% CI [–1.4020, –0.6057]) was significantly greater than both paracentral (unadjusted ΔM: 0.0172 ± 0.0757 D, 95% CI [–0.1312, 0.1656], P < 0.001) and peripheral regions (unadjusted ΔM: –0.0430 ± 0.1351 D, 95% CI [–0.3078,0.2217], P < 0.001). There was no difference between paracentral and peripheral regions (P > 0.05).

ΔJ₀ and ΔJ₄₅ (6 months vs. baseline)

No significant regional differences were detected in either group (all P > 0.05).

Between-group comparisons

After adjustment for baseline topographic keratoconus classification grade, ΔSSIv2_m was significantly greater in the Epi-off ACXL group compared to the Epi-on ACXL group in the central region ([Epi-off - Epi-on]: 0.0191 ± 0.0067, 95% CI [0.0060, 0.0322], P < 0.01), the paracentral region ([Epi-off - Epi-on]: 0.0138 ± 0.0064, 95% CI [0.0013, 0.0262], P < 0.05), and the peripheral region ([Epi-off - Epi-on]: 0.0147 ± 0.0074, 95% CI [0.0003, 0.0291], P < 0.05) (Fig 4A). Central ΔM was also significantly greater in Epi-off ACXL compared with Epi-on ACXL ([Epi-off - Epi-on]: –1.0753 ± 0.2157 D, 95% CI [–1.4981, –0.6525], P < 0.001). No intergroup differences were observed for paracentral or peripheral ΔM, or for ΔJ₀ and ΔJ₄₅ (all P > 0.05) (Fig 4B).

Figure 4.

Regional difference in SSIv2_m and M between pos6m and pre (pos6m - pre), (A) SSIv2_m, (B) M; ^∗P <0.05, ^∗∗∗P < 0.001. ACXL = accelerated corneal cross-linking.

Relative Changes (Δ%)

To ensure consistency in the direction of analysis, the Δ%M values were inverted. In the Epi-on ACXL group, Δ%SSIv2_m and Δ%M did not differ significantly among 3 regions at 6 months (P > 0.05) (Fig 5A). In the Epi-off ACXL group, Δ%SSIv2_m demonstrated a relative increase over Δ%M in the paracentral ([Δ%SSIv2_m - Δ%M]: 2.4449 ± 0.7633%, 95% CI [0.9489, 3.9408], P < 0.01) and peripheral ([Δ%SSIv2_m - Δ%M]: 2.1173 ± 0.8600%, 95% CI [0.4317, 3.8030], P < 0.05) zones, but not in the central zone (P > 0.05) (Fig 5B).

Figure 5.

Regional relative difference in SSIv2_m and M between pos6m and pre (pos6m - pre). (A) Epi-on ACXL, (B) Epi-off ACXL; ^∗P < 0.05, ^∗∗P < 0.01, ^∗∗∗P < 0.001. ACXL = accelerated corneal cross-linking.

Safety and Complications

Both the Epi-on ACXL group and the Epi-off ACXL group exhibited transient early postoperative corneal haze (Fantes grade ≤ 1), which resolved within 6 months in all cases. No major complications, such as visually significant corneal haze or scarring, were observed in either group.

In our study, all Epi-off patients met the inclusion criterion of adequate residual stromal thickness: after epithelial debridement, the stromal bed thickness was > 400 μm in every case. This safety margin is consistent with current consensus guidelines for corneal cross-linking and minimizes the risk of postoperative progressive stromal thinning.

Discussion

To halt the progression of keratoconus, CXL is widely used, but regional variations in treatment effect and the influence of epithelial status remain incompletely understood. This study assessed regional biomechanical and morphological changes after ACXL using Epi-on versus Epi-off protocols. Following Epi-off ACXL, significant SSIv2_m increases were observed across all corneal regions at 6 months, with pronounced central predominance, confirming its superior stabilizing effect. The resulting spatial gradient suggests relative undertreatment in the periphery. The Δ%SSIv2_m significantly exceeded that of Δ%M in paracentral and peripheral regions, demonstrating that biomechanical metrics detect treatment responses earlier than curvature-based maps. This supports integrating biomechanical diagnostics into routine follow-up protocols. In this retrospective cohort, the results reveal that (1) Epi-off ACXL confers stronger and more enduring stiffening than Epi-on ACXL, (2) CXL effects are not spatially uniform across the treated 9-mm zone, with a central > paracentral ≈ peripheral gradient, and (3) the sensitivity of biomechanical indices in identifying postoperative regional changes exceeds that of topographic metrics.

This study introduces several methodological innovations: region-specific biomechanical analysis using concentric zones, direct comparison of Epi-on and Epi-off ACXL under identical protocols, and the application of the SSI Map to assess regional biomechanical changes after ACXL. These approaches enable spatial characterization of corneal stiffening and provide new insights into how epithelium status influences biomechanical outcomes following ACXL.

Mechanistic Interpretation and Alignment with Prior Evidence

The observed phenomenon that the treatment effect concentrated in the central region may be associated with the following 2 nonmutually exclusive mechanisms. First, intrinsic regional differences in corneal microstructure and epithelial thickness may predispose the central cornea to a stronger CXL response. The central stromal architecture may be more conducive to riboflavin permeation and UVA utilization. The epithelium over the protruding cone is typically the thinnest, which may further facilitate riboflavin penetration.^33,34 In contrast, the thicker epithelium in flatter peripheral zones likely restricts absorption and diminishes efficacy.³⁴ Second, even within the same light dose and beam diameter, effective irradiance is not uniform across the cornea: incident rays strike the central cornea nearly perpendicularly but intersect the peripheral cornea at progressively oblique angles. Under a parallel light approximation, the incident angle (α) increases with radial distance, reducing effective irradiance via the cosine law. Tomographic studies consistently show the demarcation line is deepest centrally and becomes shallower peripherally after corneal cross-linking.^14,16

The superior clinical performance of Epi-off ACXL over Epi-on ACXL may be consistent with the role of epithelial removal in enhancing stromal riboflavin penetration and oxygen availability, both of which are considered critical for an effective photochemical reaction. This mechanistic rationale is supported by clinical and translational studies indicating that an intact epithelium significantly limits riboflavin diffusion and increases photo-attenuation,35, 36, 37 thereby likely contributing to the diminished efficacy of Epi-on protocols. Despite the higher concentration used in Epi-on ACXL, stromal riboflavin uptake remains limited by the epithelial barrier, which may partly explain the reduced biomechanical changes observed in this group.³⁸ Consistent with this interpretation, a noninferiority randomized controlled trial³⁹ found transepithelial CXL to be less effective than Epi-off in preventing K_max progression over 1 year, while meta-analyses⁴⁰ have similarly concluded that Epi-off provides greater tomographic improvement and stability. Our central zone findings further align with long-term outcomes from accelerated Epi-off series, which report sustained topographic flattening and visual gains.⁴¹ Hill et al's study⁴² focused on the impact of supplemental oxygen, they found that supplemental oxygen under Epi-on conditions maintained stromal oxygen levels during high irradiation, whereas normoxic protocols led to rapid oxygen depletion. This highlights oxygen as a key limiting factor in Epi-on ACXL.

Although Epi-off ACXL achieves greater biomechanical changes, Epi-on ACXL offers important clinical advantages: reduced postoperative pain, faster healing, and lower infection risk.³⁹ These benefits make it attractive for mild disease or patients prioritizing comfort and rapid recovery. Recent advances in riboflavin delivery have also narrowed the efficacy gap.^43,44 Therefore, treatment choice should balance biomechanical efficacy against patient quality of life and recovery expectations.

Biomechanics vs. Morphology as Outcome Read-outs

A key observation is that SSIv2 detected significant regional stiffening whereas curvature-based measures (e.g., M) showed little or no change, particularly outside the central 3-mm zone. This observed dissociation between biomechanical and topographic parameters finds parallels in several independent reports. Studies evaluating dynamic corneal response parameters at the 1-month postoperative follow-up⁴⁵ have reported measurable stiffening within weeks after CXL, often before clear topographic changes are observed. Furthermore, Herber et al⁴⁶ noted consistent SSI increases across different procedures, whereas morphological outcomes appeared more variable; Flockerzi et al⁴⁷ also described detectable biomechanical improvement at 5 months that was not accompanied by significant topographic change. These collective observations would be consistent with the conceptual framework⁴⁸ that biomechanical decompensation is the primary abnormality in keratoconus, whereas curvature/elevation changes are secondary manifestations that may lag behind material changes. The SSI-map construct we used to regionalize stiffness has been formally described and validated as a surrogate of corneal material behavior, enabling spatial analyses of keratoconic weakening and treatment response.²⁰ Classic ex vivo stress–strain work⁴⁹ further corroborates that riboflavin–UVA cross-linking increases corneal elastic modulus and reduces deformability, providing a mechanistic substrate for the in vivo SSIv2 changes.

Relation to Cone-Based Literature

Many prior analyses stratified outcomes by cone location rather than fixed annular zones.⁵⁰ Some reports suggest greater flattening when the cone is central, and attenuated effects with paracentral cones,⁵¹ although findings are not uniform across studies and protocols. Prior studies demonstrate that central biomechanical parameters vary with cone location,⁵² suggesting that center-based measurements may underestimate the true treatment effect at the cone apex in eyes with eccentric keratoconus.⁵³

Our annulus-based approach avoids dependence on cone centroid definitions and permits zone-wise comparison that is directly linked to light delivery geometry. However, this method, by averaging data within concentric annuli, may fail to capture asymmetric treatment effects relative to the cone location. Future studies integrating cone-centric and annular analyses are warranted to clarify how cone position interacts with regional dose–response.

Early Morphological Response and Transient Steepening

Our study documented an early central corneal response, with significant steepening at 1 month in the Epi-on ACXL group. We propose that this represents a transient pseudoprogression phase, attributable to an initial edema followed by collagen compaction and temporary steepening, which largely resolves within 3 months as the cornea stabilizes.⁵⁴ Other researchers have reported similar early deterioration. Ghaffari et al⁵⁵ observed an increase in maximum keratometry at 6 months postoperatively compared to preoperative values. Combined with the stability of other parameters, they suggested that this finding might be attributed to transient corneal steepening associated with stromal remodeling in the early postoperative period (within several months) after CXL.

Clinical Implications

Our results support adopting Epi-off ACXL in the absence of key contraindications—such as inadequate corneal thickness (<400 μm), low endothelial cell density, or active ocular surface disease—particularly when maximal cone curvature lies within the central 3 mm, where the relative benefit in both stiffening and flattening appears greatest. Conversely, where disease is more peripheral, the reduced peripheral effect observed here and elsewhere¹⁴ suggests that protocol modifications (e.g., tailored beam profiles, compensation for obliquity, or peripheral-boost nomograms) merit investigation to homogenize the biomechanical dose across the 9-mm zone.

In the Epi-off ACXL group, central corneal biomechanics appeared to stabilize by 6 months, at which point visual rehabilitation may be considered. However, this was not consistently observed in the Epi-on group, refractive stability was not directly assessed, and clinical decisions should be individualized.

From a monitoring perspective, incorporating regional biomechanical metrics (e.g., SSIv2 maps/dynamic corneal response parameters) alongside topography may improve sensitivity to early postoperative change and help individualize retreatment thresholds.^20,45

Nevertheless, a key limitation of this and similar studies is the paucity of long-term data beyond the 6-month period. This precludes definitive clinical translation at this time, and future studies with extended follow-up are required to confirm the stability of the observed treatment effects.

Limitations

Our study's limitations include its retrospective design (risk of selection bias) and the potential influence of regression to the mean on early outcomes, despite using multicriteria progression definitions. Additionally, there is a lack of standardized cone classification, and the absence of key data on demarcation line depth. The clinical correlates of demarcation depth are not fully settled.⁵⁶ The 6-month follow-up, limited sample size, and the known variability of J₀/J₄₅ further constrain definitive conclusions. Despite adjustments for bIOP and CCT, residual confounding is possible (e.g., epithelial remodeling and hydration dynamics).⁵⁷ Baseline pachymetric differences, specifically thinner corneas in the Epi-on ACXL group, may partly explain the fewer biomechanical changes observed postoperatively, even after analysis of covariance adjustment.

Our center-based analysis ensured standardized comparisons but did not assess changes at the cone apex. Future studies evaluating apex-specific effects—particularly in eccentric cones—may better elucidate the differential efficacy of Epi-on versus Epi-off ACXL.

Conclusions

Epi-off ACXL yielded greater regional stiffening and central flattening than Epi-on ACXL at 6 months, with the strongest effects within 0 to 3 mm. Regional biomechanical metrics (SSIv2) detected treatment-related changes more sensitively than curvature-based morphology, particularly in extra-central regions. These findings suggest that the treatment efficacy differences are likely influenced by known variations in drug/light delivery between Epi-off and Epi-on protocols, coupled with peripheral reductions in effective irradiance. Furthermore, regional biomechanical assessment may provide valuable supplementary information in post-CXL monitoring. Future prospective studies featuring extended follow-up, cone-location stratification, and quantitative demarcation-line analysis are warranted to refine dose–response relationships and optimize treatment protocols.

Manuscript no. XOPS-D-26-00041.