Small incision lenticule extraction (SMILE) has become an alternative to laser in situ keratomileusis (LASIK) for the correction of myopia and myopic astigmatism. Due to the inherent difference in surgical techniques, it is presumed that a different pattern and extent of corneal nerve resection occurs.1 In LASIK, the stromal nerve fibers that run across the circumferential flap cut are truncated and resected, and the excimer laser ablation on the stromal bed further vaporizes deeper stromal nerves.2 On the contrary, in SMILE, the nerves near the 2.5- to 4-mm small incision and inside the refractive lenticule are interrupted, but the nerve bundles located outside the cap/lenticule area remained untouched.2 A study based on in vivo confocal microscopy (IVCM) analyses on 11 patients after LASIK showed that the number and density of subbasal nerves decreased more than 90% in the first month after LASIK, and by 6 months these nerves began to recover.3 Garcia-Gonzalez et al4 further showed that the subbasal nerve plexus did not completely recover its preoperative pattern, even 10 years after LASIK. As for SMILE, studies evaluating nerve regeneration are limited and long-term studies are lacking. A comparative study on SMILE versus LASIK in 15 patients showed that eyes treated with SMILE had a higher nerve density than eyes treated with LASIK at 1 week to 3 months but not at 6 months.5 However, the study was not randomized in design and only assessed a single nerve parameter (ie, density). Other nerve variables, such as nerve length, width, or tortuosity, were not investigated.
The consequences of nerve denervation and regeneration may be seen clinically on the ocular surface. A meta-analysis comparing the postoperative ocular surface integrity between SMILE and femtosecond laser–assisted LASIK reported that the tear break-up time, Ocular Surface Disease Index scores, and corneal sensitivity were significantly better in SMILE than LASIK at 1 and 6 months postoperatively.6 However, clinical function and nerve morphology may not always correlate.7 Clinical symptoms can be present in the absence of visible nerve pathology and vice versa.7 It is known that nerve regeneration appears to lag behind the return of corneal sensation, suggesting that the current clinical evaluation on ocular surface integrity and sensitivity may not be sensitive enough to reflect the changes in nerve status. In the current study, we aimed to investigate and compare the long-term corneal nerve status following SMILE and LASIK. A randomized controlled trial design with the data of contralateral eyes from the same patient was used to eliminate the variation between individuals and between eyes. Long-term dry eye outcomes were also reported.
Patients and Methods
Patients and Surgical Procedures
This cross-sectional study involving 24 patients was part of a registered randomized controlled trial (NCT01216475). Patients were randomized to undergo SMILE in one eye and femtosecond laser–assisted LASIK in the other eye between May 2012 and November 2016. Inclusion and exclusion criteria and surgical procedures performed were as previously described.8,9 In brief, for the SMILE procedure, a 2.8-mm circumferential incision was placed at 120° and the following parameters were used: 120 µm cap thickness, 7.5 mm cap diameter, 6.5 mm optical zone, and 145 nJ power with side-cut angles at 90°. For the LASIK procedure, a superiorly hinged 120-µm thick flap was created using the VisuMax femtosecond laser (Carl Zeiss Meditec). Excimer laser ablation was then performed using the WaveLight Allegretto WAVE Eye-Q 400 Hz excimer laser (WaveLight GmbH). All procedures were performed by the same fully qualified refractive surgeon (JM). For both procedures, the postoperative regimens were identical, consisting of topical preservative-free dexamethasone and moxifloxacin with a tapering dose. Approval for the study was granted by the institutional review board of SingHealth, Singapore (2011/109/A), and the study was conducted in accordance with the tenets of the Declaration of Helsinki.
Twelve patients whose procedures were performed from 2014 to 2016 (2.7 years group) and another 12 patients whose procedures were performed from 2012 to 2014 (5.5 years group) were randomly selected for clinical assessments and corneal nerve evaluation with IVCM. Six age-matched normal patients (12 eyes) who had no corneal pathologies and other ocular comorbidities (control group) were also included.
IVCM and Image Analysis
The IVCM scanning was performed by an experienced and masked examiner (ASJJ) (HRT3 Rostock Cornea Module; Heidelberg Engineering GmbH). The central cornea was scanned first, and then patients were asked to change the gaze to scan the superior, inferior, nasal, and temporal part of the cornea (approximately 3 mm away from the corneal apex for each). The patients were instructed to fixate on a light source at different directions with the contralateral eye to stabilize the scanning view. Image acquisition time was 0.024 s/frame, and images covered a field of view of 400 × 400 µm.
For each scanned area, five best-focused and most representative images of subbasal nerves were selected. Each nerve (main truck or branched nerve) was selected only once. The 25 selected IVCM micrographs from each eye were evaluated using both manual and automated analysis software (CCMetrics and ACCMetrics, respectively; University of Manchester).10,11 After manually tracing the nerve bundles and branches, four nerve parameters were obtained by the CCMetrics software: corneal nerve fiber density (CNFD; the number of fibers/mm2, each frame area = 0.16033 mm2), corneal nerve branch density (CNBD; the number of branch points on the main fibers/mm2), corneal nerve fiber length (CNFL; the total length of fiber mm/mm2), and nerve fiber tortuosity coefficient (TC). TC is a mathematical computation of the nerve fiber tortuosity: a straight nerve equals a tortuosity coefficient of zero, and the coefficient increases with increasing tortuosity of the nerve fibers.12 The number of nerves with nerve sprouting per frame, as well as the number of nerves with beading formation per frame, were also counted.13 Beadings were defined as hyper-reflective and well-defined areas that protrude slightly from the nerve fibers. The manual image analysis was performed by an experienced investigator (JYC) who was masked to the surgical groups.14
The images were also analyzed with the ACCMetrics software to obtain other nerve assessments in addition to CNFD, CNBD, CNFL, and TC: corneal nerve fiber total branch density (CTBD; the total number of branch points/mm2), corneal nerve fiber area (CNFA; the total nerve fiber area mm2/mm2), corneal nerve fiber width (CNFW; average nerve fiber with mm/mm2), and nerve fiber fractal dimension (CFracDim). CFracDim is a metric of corneal nerve morphology to measure the spatial loss of nerves. A high CFracDim value corresponds to an evenly distributed complex nerve fiber structure that likely belongs to a healthy individual.15
Clinical Outcome Measures
The outcome measures on dry eye comprised the Schirmer I test (without anesthesia, mm/5 minutes), ocular surface fluorescein staining (Oxford score; 0: absent, 5: severe), corneal fluorescein staining (National Eye Institute scale; 0: minimal, 15: maximal), corneal sensitivity (Cochete Bonnet aesthesiometer; Luneau Ophthalmologia; 0 to 6 for each quadrant and central cornea, 0 to 30 for whole cornea), non-invasive tear break-up time (NIBUT; Oculus Keratograph 5M; Oculus Optikgeräte GmbH), patient-reported vision-related quality of life (OSDI questionnaire), and lipid layer thickness (LLT; Lipiview, TearScience, Inc), as described previously.16 In the NIBUT and LLT assessments, three measurements with 5-minute intervals on the same visit were taken, and the average was used for analysis.
The required sample size was calculated based on the pilot data on CNFD. A sample size of 21 eyes per arm, with a power of 80% or greater and at a 5% significance, was sufficient to detect the difference between the SMILE and LASIK groups. All data were expressed as mean ± standard deviation. Intraclass correlation coefficients (ICCs) and Pearson correlation tests were applied to examine the agreement and correlation between the ACCMetrics and CCMetrics measurements in the parameters that were available in both software programs (CNFD, CNBD, and CNFL). The comparisons between the SMILE and LASIK eyes, or between the control and surgical eyes, or between the dichotomous groups were performed using an independent t test. To evaluate the nerve regeneration with time, the data of the 2.7 years group was compared with that of the 5.5 years group with an independent t test. A P value less than .05 was considered significant.
The mean age at the time of surgery was 25.3 ± 5.0 years. The mean corrected spherical equivalent (SE) was −4.87 ± 1.72 and −4.79 ± 1.56 diopters (D) for the SMILE and LASIK eyes, respectively (P = .77). Patients underwent the IVCM and clinical examinations at the average time of 4.1 ± 1.5 years after surgery (range: 2.3 to 6.5 years; 2.7 ± 0.4 years for the 2.7 years group and 5.5 ± 1.0 years for the 5.5 years group). The mean age of the normal patients was 27.2 ± 6.7 years (P = .81).
Long-Term Corneal Nerve Analysis
Nerve-sprouting patterns and nerve microneuromas, which suggested nerve regeneration, tortuous nerve bundles, and nerve beading, were found in the SMILE and LASIK eyes but not in the control group (Figure A, available in the online version of this article).
Representative in vivo confocal microscopy images on the central cornea for different groups. (A, B) The laser in situ keratomileusis (LASIK) eyes had significantly lower corneal nerve fiber density (CNFD), corneal nerve branch density (CNBD), corneal nerve fiber total branch density (CTBD), and corneal nerve fiber length (CNFL), and significantly more nerves with beading (arrows) than (C, D) the small incision lenticule extraction (SMILE) eyes. (B, D) The 5.5 years group (mean postoperative 5.5 years) had significantly higher CNFD, CNBD, and CNFL than the (A,C) 2.7 years group (mean postoperative 2.7 years), regardless of surgical types. (E) The control group had significantly higher CNFD and CNBD than the 5.5 year group in both SMILE and LASIK surgery. The postoperative eyes, irrespective of surgical types, had more tortuous nerves compared to normal patients. The (F) sprouting nerves (arrowheads) and (D) microneuromas (asterisk) were seen in surgical groups but not in the control group.
There was good agreement and correlation between the manual and automated analyses. The ICC was 0.811 (95% CI: 0.615 to 0.925), 0.773 (95% CI: 0.507 to 0.897), and 0.758 (95% CI: 0.584 to 0.842), and the Pearson correlation coefficient was 0.827, 0.851, and 0.803, for the CNFD, CNFL, and CNBD measurements, respectively (all P < .001).
The LASIK eyes had significantly less CNFD, CNBD, CNFL, and CTBD than the SMILE eyes (all P < .05), indicating that even at a mean of 4.1 years after surgery, greater negative impacts on corneal nerves were still observed for LASIK compared to SMILE. The SMILE group also had significantly fewer nerves with beading than the LASIK group (P = .045). There was no difference in the CNFA, CNFW, CFracDim, TC, and number of nerves with sprouting between the two groups (Table 1).
Long-term Comparisons of Corneal Nerve Parameters in SMILE vs LASIK
Subgroup analysis was further conducted to evaluate the postoperative nerve regeneration. The CNFD, CNBD, CNFL, and number of nerves with sprouting were significantly higher in the 5.5 years group than the 2.7 years group, in both SMILE and LASIK surgery, suggesting active nerve regeneration was occurring years after surgery (Table 2). The superiority of the SMILE group over the LASIK group was more noticeable at earlier time points (the 2.7 years group). With time, there was an increasing proportion of LASIK eyes that could reach nerve status comparable to SMILE eyes. Of note, among the 4 patients who had higher CNFD, CNBD, and CTBD in their LASIK eyes in the 5.5 years group (below the equivalence line in Figure 1), 3 of them had low to moderate myopia (SE corrected for SMILE/LASIK eyes: −4.125 D/−4.125 D, −2.25 D/−2.00 D, and −4.75 D/−5.00 D; Figure 1). When comparing to the controls, the CNFD and CNBD in the 5.5 years group, for both SMILE and LASIK groups, were significantly lower than those in the control group, indicating that the corneal nerve fiber and branch densities had not regenerated to a comparable level in normal patients even 5.5 years after surgery (Table 2).
Comparison of Corneal Nerve Parameters
Scatter plots comparing the nerve metrics (A) corneal nerve fiber density (CNFD), (B) corneal nerve branch density (CNBD), (C) corneal nerve fiber length (CNFL), and (D) corneal nerve fiber total branch density (CTBD) between the small incision lenticule extraction (SMILE) and laser in situ keratomileusis (LASIK) groups based on the time groups. The gray line indicates the equivalence. At earlier time points (2.7 years, group 1), the superiority of the SMILE group than LASIK group was more noticeable. With time, there was a higher proportion of LASIK eyes that could reach comparable nerve status as of SMILE eyes. Of note, the majority (3/4) of patients who had better CNFD, CNBD, and CTBD in their LASIK eyes at 5.5 years (group 2) (below the equivalence line) had low to moderate myopia.
Clinical Dry Eye Assessments
At 4.1 years, there was no significant difference between the SMILE and LASIK groups in the Schirmer I test results, corneal staining and ocular surface staining scores, corneal sensitivity results, and NIBUT, LLT, and OSDI scores. No significant differences were noted between the SMILE and control groups, or between the LASIK and control groups (Table A, available in the online version of this article).
Comparison of Clinical Dry Eye Parameters
Overall, we did not observe a significant correlation between the clinical dry eye outcomes and nerve parameters. When dichotomizing the data according to the corrected SE, patients with high myopia (≤−6.00 D) had significantly lower CNFD than patients with low to moderate myopia (>−6.00 D; 9.2 ± 3.3 vs 14.8 ± 3.7 fibers/mm2, P = .006) in the LASIK group, but this difference was not observed in SMILE surgery. The relationships between the corrected SE and CNFD, CNBD, CNFL, or CNTD are illustrated in Figure B (available in the online version of this article). The slope (β), which indicated the amount of change in the nerve parameters with the change of corrected SE, was steeper in the LASIK group, particularly in CNFD, suggesting that LASIK had more pronounced effects on corneal nerve plexus than SMILE, in relation to the corrected SE.
Regression plots showing the relationship between the corrected spherical equivalent (SE) and nerve parameters: (A) corneal nerve fiber density (CNFD), (B) corneal nerve branch density (CNBD), (C) corneal nerve fiber length (CNFL), and (D) corneal nerve fiber total branch density (CTBD). The blue line indicates the fitted values, and the slope (β) represents the amount of change in the nerve parameters with the change of corrected SE. The gray zone indicates the 95% CI. The slope is steeper in all the plots for the laser in situ keratomileusis (LASIK) group, suggesting that LASIK had more pronounced effects on corneal nerve plexus than small incision lenticule extraction (SMILE) in relation to the corrected SE, particularly in CNFD.
In the current study, we demonstrated the long-term effects following corneal nerve transection in SMILE and LASIK. At a mean of 4.1 years postoperatively, the LASIK eyes had significantly less corneal nerve fiber and branch densities and nerve fiber length and significantly more nerves with sprouting than the SMILE eyes, especially for high myopia. For both SMILE and LASIK, nerve regeneration activity was still observed after 2.7 years, and the nerve status had not returned to normal ranges even 5.5 years postoperatively. To our knowledge, this is the first long-term and comparative study describing the corneal nerve status following SMILE and LASIK. The randomized trial design and the use of data of paired eyes from the same patient provided a more accurate assessment by minimizing the variation between individuals and eyes, as well as selection bias.
The immediate postoperative impact on corneal nerves in SMILE compared to LASIK has been previously reported in short-term clinical studies.5 In a 6-month study, eyes treated with SMILE had a greater nerve density than eyes treated with LASIK at 1 week and 3 months postoperatively but the difference was not observed at 6 months,5 whereas our study demonstrated that a significant difference was still observed at 4.1 years. The small sample size in the former study might lead to underpowering of the statistical results, resulting in the discrepancy in the findings. We demonstrated greater CFND, CNBD, CTBD, and CNFL in the SMILE group. We hypothesized this was because the area of nerve truncation was less and the subsequent reinnervation was faster in SMILE.13 The corneal nerve perforation sites, where the stromal nerve fibers perforate the Bowman's layer and become the subbasal nerve plexus, are predominantly found in the mid-peripheral corneas.17 Hence the flapless nature in SMILE would result in less nerve truncation. Less nerve truncation not only preserves more nerves, but also preserves more Schwann cells.18 Schwann cells have been shown to provide a more favorable microenvironment for nerve reinnervation by providing better maintenance of neurite function and serving as topographic tracks for neurite regrowth.19,20
Another factor that could affect nerve recovery is the effect on corneal keratocytes. Corneal nerves release neuromediators to preserve the health and integrity of keratocytes, and keratocytes in turn release growth factors and cytokines to maintain the health and regeneration of corneal nerves.7 Keratocytes are affected more by photoablation in LASIK, as opposed to photodisruption in SMILE.21,22 This mutual support and hemostasis between corneal nerves and keratocytes are disrupted to a greater extent following LASIK. Furthermore, it has been shown that the postoperative stromal inflammatory and wound healing responses,21,23 as well as the tear inflammatory mediators, are higher following LASIK than SMILE. The difference in the extent of neurogenic inflammation might also play a causal role in the postoperative nerve regeneration through the interaction between the corneal inflammatory response and neurotransmitters.7
Although the nerve outcomes were more favorable in SMILE, the CNFD and CNBD did not return to normal ranges even 5.5 years postoperatively in either type of surgery. The CNFD was of 62.5% to 62.8% of the control values in the SMILE group (Table 2; 16.4 to 17.7/26.1 to 28.3) and 57.1% to 58.7% of the control values in the LASIK group (Table 2; 14.9 to 16.6/26.1 to 28.3). This was in accordance with previous studies in which the authors presented that the mean nerve density in LASIK was 24% less than the preoperative levels at 5 years after surgery.24 However, the rate and time course of corneal nerve recovery following SMILE have not been previously reported. In addition, the CNBD was not completely recovered, with the mean of 78.0% to 93.9% of the control levels in the SMILE eyes (Table 2; 19.5 to 35.1/25.0 to 37.4), and 64.4% to 71.9% of the control levels in the LASIK eyes (Table 2; 16.1 to 26.9/25.0 to 37.4) at 5.5 years. Increased nerve tortuosity, an indicator of nerve regeneration,25 was also observed in both the SMILE and LASIK groups compared to the control group (Figure A). These results suggest that nerve regeneration were still ongoing years after refractive surgery, even in a small incision technique.
We found that those patients who had high myopic LASIK had significantly lower CNFD than low to moderate myopic LASIK, whereas this difference was not observed in the SMILE patients. The impact on corneal nerves increased with the increase in the corrected SE, with the effects being more pronounced in LASIK than SMILE (Figure B). At 5.5 years in low to moderate myopic correction, nerves in some LASIK eyes (3 of 12) could regenerate to a level comparable to those in the SMILE eyes (Figure 1). Based on Munnerlyn's formula, the ablation depth in LASIK is proportional to the intended correction, and therefore it is anticipated that the stromal volume ablated and resultant depletion of corneal nerves were greater in high myopic correction. Moreover, the stromal tissue devoid of keratocytes is greater in high myopic correction, which may lead to a great extent of impairment of biological and nutritional cues from keratocytes for nerve regeneration.1 This difference was not found in the SMILE group, which could be explained by the faster nerve regeneration in SMILE, and the nerve density had recovered to a comparable level between high and low to moderate myopic treatment.
Corneal nerve microneuromas were seen in postoperative eyes, irrespective of surgical procedure. Microneuromas, the swellings of nerve endings, form because of mechanical trauma to corneal nerves.26 Nerves with beading or sprouting were more pronounced in the LASIK eyes, indicating greater disruption of nerves and subsequent regeneration activity in LASIK. Active neural growth after nerve injury is manifested by nerve sprouting.26 In an animal model, more eyes with sprouting nerves and a greater mean number of sprouting nerves was observed in LASIK than in SMILE, which is in agreement with the current study.13 Beading of nerves represents the natural response following nerve damage and is typically more profound in the eyes with a greater extent of loss of nerves,26 which were the LASIK eyes in our study. CFracDim is a metric indicator for the spatial loss of nerves, and the mean CFracDim value was higher but not significant in the SMILE group, suggesting more evenly distributed nerves in the SMILE group.15 The CFracDim value in the LASIK group (1.41) was close to the reported data for patients with diabetic corneal neuropathy (1.40), whereas the value in the SMILE group (1.44) was close to a normal range (1.45 to 1.50).15
There were no differences in the Schirmer I test, corneal staining and ocular surface staining scores, corneal sensitivity, NIBUT, LLT, and OSDI scores between the two groups at 4.1 years postoperatively. Because the dry eye parameters had returned to a normal level in our patients, it is to be expected that we did not observe any significant correlation between the dry eye outcomes and nerve parameters.
ACCMetrics and CCMetrics software have been used to reliably quantify the corneal nerve metrics in patients with diabetic corneal neuropathy.27,28 We showed that these two analytic tools also had good measurement agreement in patients after refractive surgery.29 One limitation of this study was the lack of preoperative confocal data, so we included age-matched normal patients as the control group because age can be a confounding factor for nerve status evaluation.30 The study was cross-sectional, because we aimed to analyze the long-term nerve profiles following the two refractive procedures. Although a longitudinal study would have been more ideal, long-term follow-up of patients after refractive surgery may be difficult due to high levels of patient dropout.31 Finally, the inherent limitation of IVCM is that it provides a small region of interest (< 500 µm2), so we scanned five non-overlapping areas for each eye for analysis. Advanced imaging modalities, such as large-area or three-dimensional imaging,32 may provide more detailed information on the corneal nerve plexus in the future.
We present the first study describing and comparing the long-term consequences of corneal denervation in SMILE versus LASIK, using a randomized paired-eye design. Even 4.1 years after surgery, the LASIK eyes still had significantly less corneal nerve fiber and branch densities, and nerve fiber length, and significantly more nerves with sprouting, than the SMILE eyes, and these differences were more pronounced in high myopic treatment. Nerve reinnervation activity was observed persistently after surgery, and the nerve status had not recovered to a normal level 5.5 years postoperatively. The data suggest that the impact on corneal nerves following both refractive surgery procedures is long-lasting, and nerve regeneration and recovery require a long time frame. From the aspect of nerve preservation and regeneration, SMILE has more favorable outcomes.