Changes in the morphology and function of the sub-basal corneal nerve plexus after accelerated corneal cross-linking: A one-year follow-up study

Abstract

Purposes

Observe the recovery of nerve morphology and function in the sub-basal corneal plexus before and one year after accelerated corneal cross-linking surgery.

Methods

This study included 34 patients (47 eyes) with progressive keratoconus who underwent accelerated corneal cross-linking. Evaluations were performed preoperatively and at 1, 3, 6, and 12 months postoperatively using Pentacam topography, Cochet-Bonnet esthesiometer, and in vivo confocal microscopy. At least 3 in vivo confocal microscopy images from one eye of each keratoconus patient were selected for analysis. The nerve fiber density, nerve branch density, nerve fiber length, and nerve fiber width of the central cornea were quantified using ACCMetrics software. The sub-basal corneal plexus tortuosity was quantified by measuring the amplitude of the curves and the area under the curve formed by the main nerve fiber.

Results

Central corneal sensitivity significantly decreased at 1 and 3 months post-accelerated corneal cross-linking (P < 0.001) but returned to preoperative levels by 6 months. No significant sensitivity decreased was noted in superior, nasal, and temporal regions at all follow-up time points. The nerve branch density, nerve fiber length and nerve fiber width were significantly reduced at 1, 3 and 6 months (P ≤ 0.022) but returned to normal levels at 1 year (P = 0.348, P = 0.087). The nerve fiber density, amplitudes and area under the curves remained significantly reduced at 1 year (P ≤ 0.041).

Conclusions

Corneal sensitivity in the central region fully recovered by 6 months post-accelerated corneal cross-linking, while sub-basal corneal plexus recovery was slightly delayed. By 1 year, the nerve fiber density, and nerve tortuosity remained reduced. This indicates the need for ongoing monitoring of sub-basal corneal plexus status to ensure long-term safety.

1. Introduction

Keratoconus (KC) is a progressive corneal ectatic disease characterized by localized thinning and protrusion of the central or paracentral cornea. Without timely intervention, it may lead to progressive visual impairment, increased irregular astigmatism, corneal scarring, and even vision loss.¹ Corneal cross-linking (CXL) is an effective therapeutic approach for halting or decelerating the progression of keratoconus. Standard/Conventional corneal cross-linking (S-CXL) using the Dresden protocol has been a safe and effective treatment for progressive corneal disease.² This method involves removing the central 8–9 mm corneal epithelium, applying a 0.1% riboflavin solution for 30 min, and irradiating with 370 nm ultraviolet light at 3 mW/cm² for 30 min with a total energy dose of 5.4 J/cm². However, the lengthy treatment process restricts clinic workflow and causes discomfort for patients. The accelerated corneal cross-linking (A-CXL) protocol^3,4 reduces irradiation time by increasing irradiance, allowing for the same or slightly higher total energy output in a shorter period. While previous studies have confirmed the efficacy of A-CXL,^3,4 safety data for the 7.2 J/cm² protocol remain limited. This study selected a protocol that balanced treatment effectiveness and time efficiency, which employed an irradiance of 30 mW/cm² for 4 min to deliver a total energy of 7.2 J/cm².

In Standard/Conventional corneal cross-linking, the central 8–9 mm of corneal epithelium is typically scraped, resulting in the near-complete loss of the corneal sub-basal nerve plexus (SNP).^5,6 In addition, studies have shown significant structural and morphological changes in the anterior ∼300 μm of corneal tissue, including corneal nerves, following CXL surgery.5, 6, 7, 8 Damage to corneal innervation from surgery is a major cause of decreased corneal sensitivity and may lead to complications such as postoperative neurotrophic keratitis and ocular surface damage.^9,10 Postoperative SNP regeneration is crucial for maintaining epithelial health, supporting the protective blink reflex, and providing corneal trophic support.¹¹

In vivo confocal microscopy (IVCM) is a non-invasive method that allows for examining the living cornea at the microstructural level. To quantify SNP in IVCM images, analysis software has advanced from the fully manual CCMetrics software (University of Manchester, Manchester, UK) and the semi-automated Neuron J plug-in (inserted into ImageJ National Institutes of Health, Bethesda, Maryland, USA) to the Automatic CCMetrics software (ACCMetrics, University of Manchester, Manchester, UK), allowing for SNP nerve quantification. Previous studies have demonstrated good agreement between automated measurements with ACCMetrics and manual measurements with CCMetrics.^12,13 Compared to manual quantification, automated measurement systems eliminate inconsistencies caused by differences in assessor experience and greatly optimize the interpretation of results.¹⁴

Previous studies^15,16 evaluating the SNP after A-CXL have primarily focused on characterizing the nerve recovery process or quantifying total nerve fiber length per mm² and nerve fiber density per mm². However, the sub-basal nerve guidance is closely linked to the migration of corneal epithelial cells and can reflect the state of the epithelial surface.^17,18 Furthermore, alterations in the stromal layer, such as a decrease in keratocyte density, are also associated with nerve injury.¹⁹ These pathophysiological changes are captured in various quantitative nerve parameters, including nerve fiber length, nerve density, branching density and tortuosity. Critically, nerve alterations can occur prior to the manifestation of clinical symptoms in many diseases.²⁰ In this study, we have expanded the analysis by quantifying nerve branching, tortuosity, and width after A-CXL, along with a subjective evaluation combined with corneal sensitivity. To deepen our understanding of the regenerative capacity of the SNP.

2. Methods

2.1. Subjects

Patients diagnosed with progressive keratoconus who underwent bilateral or unilateral A-CXL treatment from September 2021 to January 2023 at Tianjin Medical University Eye Hospital were included in this study. All patients were followed up regularly at our hospital for at least 1 year postoperatively, as required. This study was approved by the Ethical Review Committee of Tianjin Medical University Eye Hospital (2020KY-25). All procedures were conducted in accordance with the Declaration of Helsinki and relevant laws regarding human subjects in research.

Inclusion criteria were as follows: patients older than 14 years, with progressive keratoconus, pachymetry of 375 μm or more at the thinnest point. Progressive keratoconus is defined by^21,22 documented changes in corneal tomographic parameters from baseline measurement exceed the contemporary 95% measurement repeatability limits for the Pentacam Scheimpflug corneal tomographer^23,24 (Table 1), or an increase in cylindrical lens power over 1.0 D or spherical equivalent power over 0.5 D within 6–12 months.

Table 1.

Criteria for early and intermediate-late cone corneal progression.

Kmax <55 D group	Kmax ≧55 D group
Kmax growth ≧1 D	Kmax growth ≧2.5 D
K1 or K2 growth on the anterior corneal surface ≧1 D	K1 or K2 growth on the anterior corneal surface ≧2.5 D
K2 growth on the posterior corneal surface ≧0.5 D	Reduced corneal thickness at the thinnest point ≧22 μm
The reduction of thinnest corneal thickness ≧16 μm

K1, Flat keratometric meridian; K2, Steep keratometric meridian; Kmax, Steepest keratometric.

The exclusion criteria were as follows: acute corneal edema, severe ocular atopic reactions, clinically significant dry eye, corneal endothelial cell density less than 2000 cells/mm², corneal scarring, ocular or systemic diseases that may affect the cornea, a history of surgery or trauma, and pregnancy or breastfeeding. Additionally, patients with other corneal ectatic disease, such as pellucid marginal degeneration (PMD), should also be excluded.

2.2. Examinations and imaging

Follow-up evaluations were conducted preoperatively and at 1 month, 3 months, 6 months, and 12 months postoperatively. Demographic data, including age and gender, were collected during the initial visit. At each visit, assessments included uncorrected distance visual acuity, best-corrected visual acuity, corneal endothelia (SP-3000P; Topcon Canada, Inc., Waterloo, Ontario, Canada), corneal topography using the Pentacam Scheimpflug corneal tomographer (Oculus Instruments, Wetzlar, Germany), SNP imaging with IVCM (Heidelberg Retina Tomography III Rostock Cornea Module, HRT-III; Heidelberg Engineering GmbH, Heidelberg Germany)and corneal sensitivity measured with a Cochet-Bonnet esthesiometer (Luneau Ophthalmlogie, Chartres, France). All contact lens wearers were instructed to discontinue lens use for at least 1 week prior to each follow-up visit.

Corneal sensitivity was assessed at the central, superior, inferior, nasal, and temporal regions located approximately 3.5 mm from the center of the cornea, using a Cochet-Bonnet esthesiometer. During the measurement, the end of the nylon wire was gently brought into vertical contact with the cornea, applying slight pressure to bend the wire. Subjects were asked to indicate whenever they felt the stimulus. Corneal sensitivity was tested three times at a single filament length. If stimuli elicited a sensation at least twice, that filament length was recorded as the corneal sensitivity. If there was no response to at least two measurements, the filament length was reduced by 5.0 mm, and the process was repeated. All measurements were conducted by the same experienced ophthalmologist (Z. M) to ensure consistency.

The Rostock Cornea Module of the Heidelberg Retina Tomography III (HRT-III; Heidelberg Engineering GmbH, Heidelberg, Germany) was used for SNP imaging. The device was set to "section mode" to scan for SNP in the central area. Clear images of the SNP on Bowman's layer were obtained, each covering a range of 400 μm × 400 μm. Clear images of SNP based on uniform criteria: a clear focus on the sub-basal plexus with visible, well-contrasted nerve fibers and even distribution. Images with nerve vortices, white spots, uneven quality, blurriness, overexposure, distorted nerves, and corneal stromal compression folds were excluded. Three to five images were randomly selected at each detection point for nerve tracking analysis.

Automated analysis of SNP was conducted using ACCMetrics software (University of Manchester, Manchester, UK). The study evaluated four parameters: nerve fiber density (CNFD), the number of main fibers per mm² (fibers/mm²); nerve branch density (CNBD), the number of branch points on the main fibers per mm² (branches/mm²); nerve fiber length (CNFL), the total length of all nerve fibers per mm² (mm/mm²); and nerve fiber width (CNFW), the average width of all nerve fibers per mm² (mm/mm²) (Fig. 1A and B).

Fig. 1.

Representative IVCM images of SNP nerves before and after morphometric analysis. (A) IVCM images of SNP in keratoconus. (B) IVCM images of SNP automatically labeled by ACCMetrics software, where the red lines indicate the main fiber, the blue lines indicate the nerve branches, and the green dots indicate branch points on the main fiber. (C) ImageJ semi-automatically labeled IVCM images of SNP, with red indicating the main nerve trunk, yellow lines outlining the AUC formed by the curved nerve fibers and the ideal straight nerve path, and orange lines representing the amplitude beneath the nerve curve.

Quantitative analysis of trunk nerve tortuosity was quantified by measuring the amplitude of the curve and the area under the curve (AUC) using ImageJ software (US National Institutes of Health, Bethesda, MD, http://imagej.nih.gov/ij/).²⁵ We first simulated the ideal straight path of the nerve fiber using ImageJ software, then measured the amplitude formed by the actual nerve fibers relative to this straight path. The AUC was calculated by manually tracing the curved nerve. The software then automatically connected the starting and ending points of the tracing to form a closed region, and calculated the area within it (Fig. 1C). For nerves with a serpentine appearance, only the segment of the nerve curve within the boundaries of the imaging was measured. To assess the inter-/intra-observer reproducibility of the tortuosity analysis using ImageJ, measurements were repeated within two observer (with one repeating measurements) in thirty randomly chosen eyes, and the intraclass correlation coefficient (ICC) were calculated.

2.3. Surgical technique

A-CXL surgery was performed on all patients by one experienced ophthalmologist (R. W). After applying 0.4% Oxybuprocaine hydrochloride eye drops, the corneal epithelium within the central 9.0 mm of the cornea was mechanically removed using a smooth spatula. A sufficient amount of riboflavin eye drops (Peschke GmbH, Switzerland) was applied to completely cover the exposed corneal stroma, with drops administered at least once every 2 min for 20 min. The ocular surface was then rinsed with a balanced salt solution. The cornea was irradiated using the KXL I Corneal Crosslinking Therapy System (Avedro Inc., Waltham, MA, USA) giving 365 nm wavelength ultraviolet light A (UVA), with an irradiation range of 9 mm in diameter, for a duration of 4 min at an irradiance of 30 mW/cm² (total energy 7.2 J/cm²). Finally, a bandage contact lens was placed after the procedure was completed. Usually on the third postoperative day, the bandage contact lens was removed once complete epithelial repair was confirmed. Patients were instructed to use 1.0% flumethasone eye drops (Santen Inc., Osaka, Japan) four times a day for one week, gradually tapering the dose of the drops over the next four weeks. Gatifloxacin eye drops (Otsuka Pharmaceutical Co., Ltd., China) were used three times a day for one week. Adjust the dosage as appropriate and promptly based on the postoperative follow-up of the operated eyes.

2.4. Statistical analysis

All statistical analyses were performed using SPSS (version 21.0). Data normality was assessed with the Kolmogorov-Smirnov test. Repeated-measures analyses were conducted using generalized estimating equations (GEE) to evaluate corneal sensitivity and SNP characterization across multiple preoperative and postoperative measurements. The GEE model included time as a categorical fixed factor to assess its effect on the outcome measures. Baseline age and Kmax were included as covariates. Bonferroni test was used for Post-hoc pairwise comparison. The ICC was calculated by using a two-way mixed effect model. A P-value less than 0.05 was considered statistically significant.

3. Results

This study involved 34 keratoconus patients (47 eyes) with a mean age of 21.35 ± 5.78 years. A total of 29 patients with 35 eyes returned for follow-up at 1 month, 23 patients with 33 eyes at 3 months, 29 patients with 42 eyes at 6 months, and 31 patients with 43 eyes at 1 year. We observed no instances of haze, endothelial changes, or delayed epithelial healing in our cohort during the follow-up period. The preoperative and postoperative values for the flat keratometric meridian (K1), Steep keratometric meridian (K2), Steepest keratometric (Kmax) and Thinnest corneal thickness were presented in Table 2. Importantly, there were no postoperative adverse events reported, such as delayed epithelial healing or infection.

Table 2.

Corneal characteristics before and after A-CXL.

	Pre-op	1 year Post-op	P value
K1 (D)	46.69 ± 5.04	46.23 ± 4.89	0.656
K2 (D)	50.23 ± 5.78	49.89 ± 5.40	0.763
Kmax (D)	58.04 ± 10.19	56.73 ± 9.61	0.482
TCT (μm)	472.74 ± 33.64	471.21 ± 31.11	0.136

K1, Flat keratometric meridian; K2, Steep keratometric meridian; Kmax, Steepest keratometric; TCT, Thinnest corneal thickness.

3.1. Corneal sensitivity changes over time

Corneal sensitivity only in the central region significantly decreased at 1 and 3 months postoperatively, but returned to preoperative levels by 6 months postoperatively (P < 0.001). Sensitivity in the inferior region was significantly higher at 6 months postoperatively compared to preoperative levels (P = 0.022) (Fig. 2).

Fig. 2.

Corneal sensitivity at different regions before and after A-CXL at 1, 3, 6, and 12 months.

∗Corneal sensitivity between these time points were statistically significant, P < 0.05.

The regional separate effects analysis revealed that corneal sensitivity was in the superior, inferior, nasal, and temporal regions compared to the central cornea at 1 and 3 months postoperatively (P < 0.001). At 6 months and 1 year, there were no statistically significant differences in corneal sensitivity between the central and peripheral regions. Details on the mean and standard deviation of sensitivity in these regions are provided in Table 3.

Table 3.

Corneal sensitivity in different regions pre and post A-CXL (mm).

	preoperative	1 month	3 months	6 months	12 months
central	57.77 ± 0.87	37.15 ± 3.07	45.17 ± 2.76	57.73 ± 0.65	57.87 ± 0.79
superior	55.05 ± 1.06	57.69 ± 0.71a	57.24 ± 1.06b	57.55 ± 0.64	58.20 ± 0.66
inferior	55.26 ± 1.27	55.90 ± 1.02a	57.35 ± 0.77b	58.59 ± 0.39	57.96 ± 0.56
nasal	58.36 ± 0.71	58.30 ± 0.60a	57.95 ± 0.61b	58.39 ± 0.68	59.22 ± 0.38
temporal	58.65 ± 0.73	58.45 ± 0.53a	58.70 ± 0.48b	58.91 ± 0.41	59.23 ± 0.38

^aThe difference in corneal sensitivity between the central region of the cornea and other peripheral regions at 1 month postoperatively was statistically significant, P < 0.05.

^bStatistically significant difference in corneal sensitivity between the central and other peripheral regions of the cornea at 3 months postoperatively, P < 0.05.

3.2. Nerve changes over time

At 1 month postoperatively, 82.86% of eyes showed sub-basal nerve absence, which decreased to 18.75% at 3 months, 9.52% at 6 months, and 6.98% at 1 year. IVCM images revealed SNP absence at 1 month postoperatively. By 3 months, subepithelial nerve flocks resembling the morphology of Langerhans cells were observed, forming a "pseudo-dendritic regeneration pattern" (Fig. 3, red arrows). At 6 months, nerve regeneration continued, with disconnected fibers. By 1 year, the nerve fibers gradually elongated and formed an interconnected nerve network.

Fig. 3.

IVCM images of SNP nerves in the same keratoconus patient

(A-E): pre- and post-operative SNP images at 1, 3, 6, and 12 months after automatic labeling by ACCMetrics software. (F–J): Original SNP images before and 1, 3, 6, and 12 months after surgery.

The results of quantitative analysis of SNP nerves at various time points were presented in Table 4. The CNBD was significantly lower in 1, 3, and 6 months compared to preoperative levels (P < 0.001), but no statistically significant difference was observed at 1 year (P = 0.349). Similarly, CNFL and CNFW was significantly reduced at 1, 3, and 6 months (P < 0.001, P ≤ 0.022), with no significant difference detected at 1 year relative to preoperative levels (P = 0.107, P = 0.081). The CNFD was significantly reduced compared to preoperative levels at 1, 3, 6 months, and 1 year (P < 0.001). Both the amplitude and AUC of nerve tortuosity at 1 year remained significantly lower than preoperative values (all P < 0.001). Additionally, both intra- and inter-observer assessments demonstrated good reliability for AUC and amplitude measurements. The intra-observer ICC values were 0.755 and 0.795, and the inter-observer ICC values were 0.752 and 0.790, respectively (all P < 0.001).

Table 4.

Changes in the sub-basal corneal plexus before and at 1, 3, 6, and 12 months after A-CXL.

	preoperative	1 month	3 months	6 months	12 months
CNFD	21.68 ± 7.15	5.1 ± 2.82∗	3.8 ± 3.18∗	8.49 ± 4.24∗	15.2 ± 6.06∗
(fibers/mm2)		<0.001	<0.001	<0.001	<0.001
CNBD	34.0 ± 17.75	1.1 ± 1.56∗	3.3 ± 5.49∗	10.2 ± 8.94∗	26.9 ± 18.85
(branches/mm2)		<0.001	<0.001	<0.001	0.349
CNFL	13.16 ± 3.48	3.9 ± 1.78∗	3.1 ± 1.70∗	6.66 ± 3.45∗	11.76 ± 4.19
(mm/mm2)		<0.001	<0.001	<0.001	0.107
CNFW	21.44 ± 1.48	2.61 ± 7.08∗	14.32 ± 10.34∗	20.34 ± 1.52∗	20.6 ± 0.94
(mm/mm2)		<0.001	0.021	0.022	0.081
AUC	762.80 ± 227.96	–	–	491.51 ± 142.90	498.42 ± 63.31
(μm2)				<0.001	<0.001
Amplitude	15.37 ± 2.40	–	–	13.16 ± 1.99	13.20 ± 1.41
(μm)				<0.001	<0.001

CNBD, Central Nerve Branch Density; CNFD, Central Nerve Fiber Density; CNFL Central Nerve Fiber Length; CNFW, Central Nerve Fiber Width; Parameters with statistically significant differences from preoperative are indicated by ∗.

4. Discussion

This study conducted a comprehensive assessment of the recovery of SNP structure and function both preoperatively and longitudinally up to 1 year postoperatively.

Corneal sensitivity is a clinical indicator used to assess the morphology and function of corneal nerve fiber damage and regeneration after CXL. Previous studies have examined corneal sensitivity after CXL in rabbit and KC patients using the Cochet-Bonnet esthesiometer,^26,27 and reported decreased corneal sensation and nerve damage. In our study, corneal sensitivity in the central region decreased at 1 and 3 months after A-CXL, and returned to the preoperative level by the 6 months. This is consistent with the recovery process of central corneal sensitivity after A-CXL (30 mW/cm²,4 min, total energy 7.2 J) reported in Ozgurhan EB et al.¹⁵ In contrast, Wasilewski et al.²⁷ showed that corneal sensitivity had not yet reached preoperative levels at 6 months after S-CXL. Similarly, the percentage of eyes with nerve deficits by 6 months after A-CXL was lower than that reported for S-CXL in previous studies. Jordan C et al.'s study of postoperative S-CXL patients, SNP deficits was observed in 71%, 55%, 18%, and 7% of eyes at 1, 3, 6, and 12 months postoperatively, respectively.²⁸ In our study, 82.86%, 18.75%, 9.52%, and 6.98% of eyes had SNP deficits at 1, 3, 6, and 12 months. A comparison of outcomes between the S-CXL and A-CXL protocols reports is presented in Table 5. These phenomena are possibly due to differences in the healing processes of S-CXL and A-CXL. A previous study reported a more pronounced decrease in nerve density and anterior stromal keratocyte density at 1, 3, and 6 months after S-CXL compared to A-CXL.²⁹ Wen et al.³⁰ also highlighted differences in outcomes between S-CXL and A-CXL, with their findings suggesting the decreases in CCT and endothelial cell density were less with A-CXL than with S-CXL. The S-CXL procedure, which involves a longer duration of UV irradiation, likely has a more pronounced effect on stroma and SNP nerves. However, specific studies investigating the corneal sensitivity recovery are very limited.

Table 5.

Comparative summary of studies on sub-basal nerve plexus recovery following corneal cross-linking.

Study (Year)	CXL Protocol	Radiant Energy	Irradiation Time	Key Findings
Mazzotta et al. (2008)31	S-CXL	3 mW/cm2	30 min	● Nerve growing from the surrounding nonirradiated area from 1 month.
				● Restoration of fiber density and interconnections to the preoperative architecture was achieved by 24 months.
Croxatto et al. (2010)32	S-CXL	3 mW/cm2	30 min	● Regeneration initiated by 1 month. Near-preoperative appearance by 24-36 months.
Jordan et al. (2014)28	S-CXL	3 mW/cm2	30 min	● Sub-basal Nerve Deficit Rate: 71% eyes at 1 month, 55% eyes at 3 months, 18% eyes at 6 months, 7% eyes at 12 months.
				● The mean sub-basal nerve density reduced at 1 (1.10 ± 2.5 mm Nerve/mm2), 3 (4.22 ± 6.23 mm Nerve/mm2), and 6 (7.08 ± 5.4 mm Nerve/mm2) months postoperatively.
				● Returned to preoperative values at 12 months (13.32 ± 5.3 mm Nerve/mm2)
Kontadakis et al. (2013)33	S-CXL	3 mW/cm2	30 min	● Total nerve length per image reduced at 1 month (163 ± 371), 3 months (663 ± 788) and 6 months (919 ± 549) postoperatively.
				● Returned to preoperative values at 9 months (1422 ± 920).
Hashemian et al. (2014)29	S-CXL	3 mW/cm2	30 min	● Postop-Preop Sub-basal nerve density (fiber/mm2): 4.2 ± 0.87 at 1 month, −3 ± 0.81 at 3 months, −2.15 ± 0.7 at 6 months, −1.25 ± 1.1 at 12 months, −0.1 ± 0.3 at 15 months
	A-CXL	30 mW/cm2	3 min	● Postop–Preop Sub-basal nerve density (fiber/mm2): 4.2 ± 0.87 at 1 month, −3 ± 0.81 at 3 months, −2.15 ± 0.7 at 6 months, −0.36 ± 0.5 at 12 months, +0.12 ± 0.2 at 15 months.
				● Nerve density decrease was more pronounced after S-CXL than A-CXL at 1, 3, and 6 months.
Ozgurhan et al. (2015)15	A-CXL	30 mW/cm2	4 min	● Nerve Fiber Density (fiber/mm2): Significantly decreased in the 1, 3, 6 (8.33 ± 5.30) months postoperative and recovered by 1 year (21.66 ± 11.09).
Parissi et al. (2016)34	S-CXL	30 mW/cm2	30 min	● Nerve tortuosity showed no significant changes from preoperative levels at 1-5 years postoperatively. At 7 to 12 months, nerve density did not differ from preoperative levels.
				● Nerve density continued to increase up to 4–5 years post-treatment.
Current Study	A-CXL	30 mW/cm2	4 min	● CNFL, CNFW & CNBD: Recovered to preoperative levels by 1 year.
				● CNFD, & Tortuosity: Remained significantly reduced at 1 year.
				● Sub-basal Nerve Deficit Rate: 82.86% eyes at 1 month, 18.75% eyes at 3 months, 9.52% eyes at 6 months, 6.98% eyes at 12 months.

Previous quantitative analyses of post-CXL nerves have shown that nerve fiber density (fiber/mm²)^15,29 and total nerve length (mm Nerve/mm²)^28,33 decrease significantly in the early postoperative period and largely return to preoperative levels at 1 year. In our study, CNFL largely returned to preoperative levels at 1 year postoperatively, consistent with previous studies. The first quantitative analysis of postoperative CNFW after CXL showed that it remained reduced at 1 year postoperatively, although the difference was not statistically significant. Similarly, previous studies of trans-epithelial refractive keratectomy (Trans PRK),³⁵ Laser-Assisted in situ keratomileusis (LASIK) and penetrating keratomileusis (PK)³⁶ have found that SNP nerves regenerated around 1 year postoperatively are usually still thin. In addition, the CNFD remained low at 1 year postoperatively in this study. This could be due to the fact that the nerve fibers were still thin or discontinuous at that time making it difficult for the ACCMetrics software to automatically trace them. These results demonstrate that the restoration of nerve architecture remained incomplete at the 1-year postoperative mark. These consistent with previous studies reporting that the SNP had regenerated to form a network similar to the preoperative state at 12–24 months.^31,32

Considering the morphology of nerve fibers after CXL, many parameters exist to describe and quantify their length, density, but almost no measurements of nerve tortuosity per se. A novel finding of our study is the persistently reduced nerve tortuosity observed even at 1 year after A-CXL. In previous study, until 1–5 years after S-CXL the tortuosity of SNP nerves in KC patients found no significant differences compared to preoperative periods.³⁴ This pattern stands in clear contrast to the recovery of nerve tortuosity following refractive surgical procedures. For instance, nerve tortuosity was not significantly reduced after LASIK and, after LASEK, it decreased at 1 month but returned to preoperative levels by 3 months.³⁷ The distinct baseline nerve morphology and pathological microenvironment in keratoconic corneas, compared to healthy subjects, might contributes to the incomplete regeneration observed at 1 year. SNP nerve migration occurs when epithelial cells move from the corneal peripheral toward the center,¹⁷ so we hypothesized that it may take longer for the nerve to migrate and take on a curved state after regeneration of the corneal epithelium and SNP nerves.

The Cochet-Bonnet esthesiometer detects subjective corneal sensation in SNP damage. The IVCM examination provides an objective assessment of these damaged nerves. Consistent with previous studies,¹⁵ corneal sensitivity returned to preoperative levels more rapidly than nerve morphology parameters. Moreover, in our study no correlation was found between corneal sensitivity and SNP morphological parameters either preoperatively or throughout the recovery period. The healing process of the nerve fibers did not align perfectly with recovery of corneal sensitivity. Corneal sensitivity can return to preoperative levels before full recovery of the corneal nerve fibers. This may be due to the fact that corneal sensitivity reaches normal levels when approximately 1/3 of the corneal nerve reserve is present.³⁸ And we hypothesize that Postoperative corneal nerve hypersensitivity may compensate for the incomplete nerve structural. This heightened sensitivity may stem from mechanisms: first, that newly regenerating nerve fibers exhibit spontaneous discharges and abnormal responses³⁹; and secondly, that inflammatory mediators released during epithelial wound healing can locally increase the sensitivity of remaining and regenerating nerves.⁴⁰ Thus, albeit structurally incomplete, a partially regenerated nerve plexus may sustain nearly normal sensory function. In the present study, The transient elevation of inferior corneal sensitivity observed at 6 months postoperatively may also be linked to the release of inflammatory mediators and abnormal responses from regenerating nerves.^39,40 Considering methodological limitations, ACCMetrics is excellent for tracking established nerves but may lack the sensitivity to detect the finest, early regenerating neurites. Consequently, the functional contribution of these early, thin fibers might be "invisible", creating a lag in the objective record of recovery.

Nerve recovery is associated with multiple factors on the ocular surface.^41,42 In the tear fluid, the CNFW, CNFL, and other neuromorphisms correlate significantly with inflammatory mediators (e.g., interleukin 6) and neuropeptides (e.g., substance P).⁴² The earlier sensory recovery after A-CXL is potentially beneficial for maintaining ocular surface homeostasis, tear film stability, and reducing the risk of dry eye symptoms. However, in this study sub-basal nerve regeneration remains incomplete even at 12 months postoperatively. CXL treatment is commonly used in young patients. Long-term monitoring is particularly relevant for young keratoconus patients, as prolonged neural deficits may predispose them to neurotrophic complications over their lifetime. In addition, the axons of the sub-basal corneal nerve originate outside the corneal stroma and are not affected by CXL. Changes in the SNP may be involved in the pathogenesis and progression of the disease.⁴³ Therefore, nerve alterations could continue to progress after CXL and may reflect the underlying disease process. Long-term observation of subtle SNP nerve changes detected by IVCM after CXL may aid in assessing keratoconus progression, especially when topographic measurements are unreliable in advanced cases.

This study has several limitations that should be considered. First, the relatively small sample size necessitated the inclusion of data from both eyes of some patients — a potential bias that was addressed using a GEE model to correct for inter-eye correlations. Second, IVCM scans is that it provides a small region of interest (400 μm²) and an operator-dependent image acquisition process, meaning the analyzed areas may not be to standardise scanning location each time. Furthermore, The Cochet-Bonnet esthesiometer, although widely used in clinical practice, its stimulus range is relatively coarse, potentially limiting its ability to detect subtle changes. Also, it only measures mechanical sensitivity and does not assess chemical or thermal sensitivity. Finally, the follow-up period of only one year is not long enough to observe whether corneal nerve morphology returns to preoperative levels.

Study approval

The authors confirm that any aspect of the work covered in this manuscript that involved human patients or animals was conducted with the ethical approval of all relevant bodies and the study was performed in accordance with the Declaration of Helsinki，and the protocol was approved by the Ethics Committee of Tianjin Medical University Eye Hospital (2020KY-25).

Author contributions

The authors confirm contribution to the paper as follows: ZM designed the study and was a major contributor to writing the manuscript and measuring corneal sensitivity. HC performed the tortuosity measurements and analyzed the data. XW interpreted the date. NG and WC collected the data. CL and DL critically revised the manuscript. RW contributed to the conception of the study and performed the surgical procedures. All authors reviewed the results and approved the final version of the manuscript.

Funding

This work was supported by the National Natural Science Foundation of China [Grant number 820709229]; Tianjin Key Medical Discipline (Specialty) Construction Project [Grant number TJYXZDXK-037A]; InnoHK initiative, the Hong Kong Special Administrative Region Government [Grant number P0043871] and Research Centre for SHARP Vision 350 [Grant number P0039545].

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.