Preoperative geometric parameters predict the outcome of lamellar keratoscleroplasty in patients with limbal dermoids

To find preoperative simple geometric parameters to predict the outcome of lamellar keratoscleroplasty in patients with corneal limbal dermoids. We retrospectively analyzed the data of 30 patients with limbal dermoids who underwent lamellar keratoscleroplasty and were followed up for more than 6 months. Seven geometric parameters were used to analyze the relationship with postoperative visual acuity, astigmatism, and scar formation and investigated for their influence on postoperative outcome. These seven parameters included the cornea-invaded area of dermoid, dermoid area, the cornea-invaded length of dermoid, the corneal limbus invaded length of dermoid, dermoid size, bed size, graft size. Furthermore, we divided patients with corneal limbal dermoids into amblyopia group and non-amblyopia group, and analyzed the clinical characteristics of the amblyopia group. The mean age of the patients at surgery was 10.6 ± 5.83 years. The 7 geometric parameters, preoperative and postoperative astigmatism of the amblyopia group were higher than those of the non-amblyopia group (all P < 0.05). Among the geometric parameters analyzed, the r coefficients between the cornea-invaded length of dermoid and postoperative visual acuity and astigmatism were the highest (r = 0.854, r = 0.714). The r coefficient between the corneal limbus invaded length of dermoid and the postoperative scar was the highest (r = 0.375). The r coefficient between age and postoperative steroid-induced high intraocular pressure was the highest (r = − 0.416). In univariate regression analysis, the cornea-invaded length of dermoid was significantly correlated with postoperative visual acuity (β = 0.270, P < 0.001) and postoperative astigmatism (β = 1.362, P < 0.001). Among the geometric parameters analyzed, the cornea-invaded length of dermoid had best stratified patients in grouping with or without amblyopia (cutoff > 2.24). The cornea-invaded length of dermoid was the most important related factor of postoperative visual acuity, astigmatism, and amblyopia. The corneal limbus invaded length of dermoid was the most important related factor of postoperative scar formation. The geometric parameters and astigmatism in patients with amblyopia were larger than those in patients without amblyopia.

amblyopia group were higher than those of the nonamblyopia group (all P < 0.05). Among the geometric parameters analyzed, the r coefficients between the cornea-invaded length of dermoid and postoperative visual acuity and astigmatism were the highest (r = 0.854, r = 0.714). The r coefficient between the corneal limbus invaded length of dermoid and the postoperative scar was the highest (r = 0.375). The r coefficient between age and postoperative steroidinduced high intraocular pressure was the highest (r = − 0.416). In univariate regression analysis, the cornea-invaded length of dermoid was significantly correlated with postoperative visual acuity (β = 0.270, P < 0.001) and postoperative astigmatism (β = 1.362, P < 0.001). Among the geometric parameters analyzed, the cornea-invaded length of dermoid had best stratified patients in grouping with or without amblyopia (cutoff > 2. 24). Conclusion The cornea-invaded length of dermoid was the most important related factor of postoperative visual acuity, astigmatism, and amblyopia. The corneal limbus invaded length of dermoid was the most important related factor of postoperative scar formation. The geometric parameters and astigmatism in patients with amblyopia were larger than those in patients without amblyopia.

Introduction
Corneal limbus dermoid, the most common tumor of corneal limbus, is a congenital, benign solid choristoma, frequently found at the inferotemporal limbus. It was reported that the incidence of corneal limbus dermoid was 0.01%-0.03% [1,2].
Patients with limbus dermoids are often accompanied by astigmatism and amblyopia [3,4]. Large and prominent dermoid can cause the foreign body sensation, dry eye, and other diseases [5][6][7]. According to the extent of invasion, limbus dermoids can be divided into 3 grades [6,8]: Grade 1, superficial lesion with a diameter less than 5 mm; Grade 2, the dermoid covering most of the cornea and extending down to the descemet's membrane; Grade 3, the dermoid invading the whole cornea and part of the anterior segment.
For the grade 1 and grade 2 limbus dermoids, corneal limbus dermoids excision combined with lamellar keratoscleroplasty is recommended, which has been proven with lower postoperative complications and recurrence rate, compared with simple excision [6,[9][10][11]. However, there is still a lack of satisfying methods to predict the postoperative visual and refractive outcome. Previously, Kim et al. [12] reported that the geometric parameters of dermoid, including orientation, angular width, and 9 ratio parameters, in the digital anterior segment photographs could be used to predict the postoperative effect of simple dermoids excision. However, few studies have been conducted to determine the parameters to predict the postoperative outcome of dermoids excision with lamellar keratoscleroplasty, which is needed for larger limbus dermoids, especially with deeper corneal invasion. Furthermore, are there any preoperative detailed values to study the outcome of lamellar keratoscleroplasty in patients with limbal dermoid? Because orientation, angular width, and ratio are inconvenient and impractical for clinicians to apply. As far as we know, there is no relevant research.
This study was conducted to explore the suitable preoperative geometric parameters to predict the postoperative outcome of dermoid excision combined with lamellar keratoscleroplasty.

Subjects
The study protocol was approved by the Ethics Committee of Joint Shantou International Eye Center (JSIEC) of Shantou University and the Chinese University of Hong Kong (Shantou city, China). The study followed the tenets of the Declaration of Helsinki. Informed consent for inclusion was waived in this retrospective study. Patients who underwent excision of pathology-proven dermoids combined with lamellar keratoscleroplasty between October 2013 and August 2021 at JSIEC in Shantou were consecutively enrolled. We regularly followed up all patients and treated amblyopia duration with all sutures removed. Treatment of amblyopia included spectacles correction and occlusion therapy. The records of the last follow-up were included in the outcome data and all sutures had been removed when the outcome data was recorded. Patients who had fewer than 6 months of postoperative follow-up were excluded.
The corneal graft was graded according to the transparency in the last follow-up after the operation, from 0 to 4 (0 = completely clear; 1 = minimal haze seen with difficulty under direct illumination; 2 = mild haze seen easily; 3 = moderately dense opacity partially obscuring iris details; 4 = dense opacity completely obscuring all details of intraocular structure) [13].
Furthermore, we divided patients into amblyopia group and non-amblyopia group. The diagnosis is established by a unilateral or bilateral reduction in best-corrected visual acuity (BCVA) not attributable to structural abnormalities of the visual pathways [14]. Amblyopia was defined as the BCVA difference of more than two logarithm of the minimum angle of resolution (logMAR) lines between the two eyes after 6 months of operation.
Surgical procedure All operations were performed by the same corneal specialist (JX) using standard lamellar keratoscleroplasty techniques. Briefly, the surgeon used scissors to perform the conjunctival incision surrounding the conjunctival side of the lesion. The smallest size trephine was used to encompass the whole lesion and the dense lipid infiltrate if present to mark the boundary needed for cutting. Then, the dermoid was dissected in a lamellar fashion by a crescent blade until a clear corneal bed was reached. Corneoscleral bed was cauterized to completely stop the bleeding. An anterior lamellar, limbal-based corneoscleral patch, 0.25-0.5 mm larger than the corneoscleral bed, was fashioned from a corneoscleral button. The donor button was placed on the excised dermoid area, and the edge was sutured with interrupted 10-0 nylon sutures. The excised lesion was sent for histopathologic examination. The conjunctiva was pulled over the exposed sclera and sutured with 10-0 nylon to the edge of the graft. Patients were treated with topical corticosteroids and antibiotics after the operation. The nylon sutures were sequentially removed 3-6 months after the surgery.

Measurement of the geometric parameters
Preoperative digital anterior segment photographs, captured under a slit lamp with a Canon EOS 1500D digital camera, were imported into ImageJ software (National Institutes of Health, Bethesda, MD), a software used to measure the geometric parameters. The maximum diameter of dermoid, measured by the operator using compasses, was obtained through the medical record. Two researchers (HW, JX) who were blind to the patient's clinical information measured the length and area in the straight-line and freehand selection modes according to the proportional relation of this size of dermoid. As shown in Fig. 1, the corneal limbus invaded length of dermoid, the cornea-invaded length of dermoid, the cornea-invaded area of dermoid, and the whole area of dermoid were measured.
The corneal limbus invaded length of dermoid is defined as the length of the straight line at 2 points where the corneal limbus and dermoid meet (Fig. 1A). The cornea-invaded length of dermoid is defined as the length of the straight line perpendicular to the straight line between the 2 points where the corneal limbus and dermoid meet, and the extension of this length passes through the center of the pupil (Fig. 1B). The manual measurement of these parameters was repeated four times by two researchers (HW and JX). The final metrics were taken as the average of the results measured by two researchers to minimize the effect of inter-rater variation. The size of the corneoscleral bed and graft has been obtained from medical records.

Statistical analysis
Data analysis was carried out using commercially available software (SPSS ver.13.0; SPSS Inc, Chicago, Illinois, USA) and MedCalc (ver.15.2.2, Mariakerke, Belgium). To assess the intraobserver and interobserver reproducibilities of the measurements of these metrics, intraobserver correlation coefficients (ICCs) with 95% confidence intervals were calculated. We used the Shapiro-Wilk test to evaluate the normality of continuous variables. Independent t test, Mann-Whitney test, and χ2 test were used to compare the differences between the No Amblyopia group and Amblyopia group. We calculated the areas under the receiver operating characteristics curves (AUCs) to determine the geometric parameters to predict postoperative amblyopia. Spearman's rank correlation analysis was used to evaluate the correlation between geometric parameters and postoperative visual acuity and postoperative astigmatism. The variable with the highest correlation coefficient tested by Spearman's rank correlation analysis was included in univariate regression analyses. Kendall's Tau-b correlation analysis was used to evaluate the correlation between geometric parameters and postoperative scar formation. The correlations between steroid-induced high intraocular pressure (IOP) and the other variables were assessed using the Point-biserial correlation. A P-value of < 0.05 was taken to be of statistical significance.

Baseline characteristics
Thirty patients were followed up for at least 6 months. The baseline characteristics are shown in Table 1. The average age of patients at the time of surgery was10.6 ± 5.83 years. The median logMAR BCVA was 0.39 ± 0.44 preoperatively, and 0.26 ± 0.33 postoperatively. The visual acuity improved (the median change was 0.13 ± 0.21, P = 0.002). The mean astigmatism before and after the surgery was 2.73 ± 1.79D and 2.46 ± 1.77D, respectively, without significant change (0.28 ± 0.79D, P = 0.068). No intraoperative complications were noted in any patient. After the first week of follow-up, complete re-epithelialization was seen in all patients. During the follow-up, one patient developed pseudopterygium secondary to graft infection, and none of the other patients developed graft infection or pseudopterygium. Steroidinduced high IOP was identified in 8 patients (26.7%) and all surgical complications were controlled medically and treated successfully. Grade 1 and Grade 2 graft opacity occurred in 4 (13.3%) and 19 (63.3%) cases. There were 7 patients (23.3%) with Grade 3 graft opacity, among which 2 were caused by suture-related complications, 1 was caused by pseudopterygium, and 2 were related to graft rejection caused by neovascularization.

Reproducibility of measurements of geometric parameters
The intraobserver and interobserver reproducibilities were high for measurements of geometric parameters of lesions. The ICCs for intraobserver reliability (HW) of the corneal limbus invaded length of dermoid, the cornea-invaded length of Clinical characteristics of amblyopia and non-amblyopia groups The 7 morphological parameters, preoperative and postoperative astigmatism in the amblyopia group were higher than those in the non-amblyopia group (all P < 0.05). The amblyopia patients underwent surgery at the age of 8.1 ± 4.5 years, significantly younger than 12.8 ± 6.1 years of the non-amblyopia patients (P < 0.05). There were significant differences in the average change of visual acuity between the two groups (0.16 ± 0.07, P = 0.036). However, there was no significant difference in the average change of astigmatism (0.01 ± 0.29D, P = 0.964). The median logMAR visual acuity and astigmatism of the non-amblyopia group had no significant change before and after the operation. In the amblyopia group, the median logMAR visual acuity was 0.72 ± 0.40 preoperatively and 0.51 ± 0.35 postoperatively. The visual acuity improved (the average change was 0.21 ± 0.21, P = 0.002). The mean astigmatism of the amblyopia group was 4.21 ± 1.23D preoperatively and 3.95 ± 1.35D postoperatively. Astigmatism had no obvious change (the average change was 0.27 ± 0.99D, P = 0.331). There was no significant difference between the two groups in the incidence of different grades of graft opacity and the incidence of steroid-induced high IOP (all P > 0.05) ( Table 2).

Correlation analysis between geometric parameters and postoperative results
Among the geometric parameters, the cornea-invaded length of dermoid had the highest r coefficient with postoperative visual acuity and astigmatism (r = 0.854, r = 0.714, all P < 0.001), followed by the cornea-invaded area of dermoid (r = 0.828, r = 0.709, all P < 0.001). The r coefficient between the corneal limbus invaded length of dermoid and the postoperative scar was the highest (r = 0.375, P = 0.012). The r coefficient between age and steroid-induced high IOP was the highest (r = − 0.416, P = 0.022) ( Table 3).

Regression analysis
In univariate regression analysis, the cornea-invaded length of dermoid could predict postoperative visual acuity (β = 0.270, P < 0.001). The cornea-invaded length of dermoid had a moderate effect on postoperative vision, and the adjusted R 2 for the model fit was 0.573, which indicated that it had a robust predictive value. The same analysis showed that the cornea-invaded length of dermoid could predict postoperative astigmatism (β = 1.362, P < 0.001). The cornea-invaded length of dermoid had a moderate influence on postoperative astigmatism, and the adjusted R 2 for the model fit was 0.514. The results of this analysis are shown in Table 3.

ROC curve analysis predicts postoperative amblyopia
ROC curve showed that among all the parameters tested, the cornea-invaded length of dermoid could best distinguish amblyopia patients from non-amblyopia patients (AUC = 0.93), followed by the cornea-invaded area of dermoid (AUC = 0.92). The cornea-invaded length of dermoid stratified the patients by the predictability of amblyopia (sensitivity = 92.9%, specificity = 100%) ( Table 4).

Discussion
Previously, there was a lack of preoperative simple geometric parameters to predict the outcome of lamellar keratoscleroplasty in patients with limbal dermoid. In this study, we found that the corneainvaded length of dermoid was the most important factor predicting the postoperative visual acuity, astigmatism and amblyopia. The corneal limbus invaded length of dermoid was the most important factor related to postoperative scar formation. Astigmatism and amblyopia caused by limbal dermoid are not only the indications of surgical intervention, but also the vision-threatening issues after the surgery [8]. We found that there was no significant change in astigmatism after surgery, regardless in the amblyopia group or the non-amblyopia group. These findings supported the previous reports that astigmatism induced by limbal dermoid might be due to alteration or molding of the intrinsic structure of the corneoscleral wall when lesions formed [13,[15][16][17]. The visual acuity was improved after the operation, in line with the results of previous studies. Fan et al. [18] followed up 66 children with limbal dermoid for more than one year after lamellar keratoscleroplasty. They found that there was no significant change in astigmatism. Yao et al. [19] followed up 43 patients with corneal limbal dermoids (40 of whom underwentdermoid excision with lamellar keratoscleroplasty), and found that the median logMAR BCVA improved significantly after the surgery. However, there was no significant change in astigmatism postoperatively. Shen et al. [13]studied10 patients with corneal limbal dermoid treated by lamellar keratoscleroplasty, and found that the postoperative astigmatism of patients with preoperative astigmatism less than 6 D did not decrease significantly, while that of patients with preoperative astigmatism greater than 6 D decreased significantly. However, the extrapolation was weak to the limited sample size, and the present study, with the maximum astigmatism 6 D, could hardly be used to verify Shen et al.'s conclusions, which needed further study.
In this study, preoperative geometric parameters were used to predict the postoperative effect. Kim et al. [12] also used preoperative anterior segment photography of patients with dermoid to study the correlation between the ratio of various geometric parameters and the results after simple dermoid resection. They reported that the invaded angular axis of dermoid to corneal diameter ratio value was the most significant factor associated with postoperative visual acuity, amblyopia development, which was similar to our result. But they also found that the invaded angular axis of dermoid to corneal diameter ratio value was the most significant factor associated with postoperative scarring, which was different from our results. This could be explained by the different surgical methods. The vascularization and pseudopterygium are more likely to occur after simple removal, while lamellar keratoplasty is considered to be a better choice because it provides the excellent appearance, tectonic stability, and low risk of vascularization and pseudopterygium [9,13,15,[20][21][22]. Furthermore, there are many other factors influencing scar formation, including the length of dermoid invading corneal limbus, interface neovascularization, graft rejection, and so on [18].
Our study used detailed values to study the relationship between preoperative simple geometric parameters and the outcome of lamellar keratoscleroplasty in patients with limbal dermoid. It is more convenient and practical for clinicians to apply. The clinicians only need to use the computer to open the preoperative digital anterior segment photographs and use the ruler tool on the computer to predict the postoperative effect of patients. To the best of our knowledge, our study was the first literature to report on the use of geometric parameters to predict the postoperative outcome of lamellar keratoscleroplasty in patients with limbal dermoid. When clinicians find that the cornea-invaded length of dermoid is too long before the operation, they should consider that patients still have amblyopia and astigmatism postoperatively, which requires regular monitoring, patient education, and amblyopia therapy after surgery. For patients with large corneal limbus invaded length by dermoid, clinicians should consider the high possibility of scar after the operation, strengthen the followup, detect the occurrence of complications in time, and take medication to avoid scar aggravation. The cornea-invaded length of dermoid was the most important factor affecting the development of postoperative visual acuity, astigmatism and amblyopia. It might be that the closer the dermoid was to the optic axis, the greater astigmatism. It provides a new insight for the influence of pathological features on postoperative refraction. The corneal limbus invaded length of dermoid was the most important related factor of postoperative scar formation, reflecting that corneal limbus with wider invasion by dermoid had more defects of limbal stem cells and was easier for neovascularization. Neovascularization would more easily lead to rejection. In our study, postoperative grade 3 graft opacities (2/7) were related to graft rejection caused by neovascularization. In a multicenter study on the causes of corneal transplantation failure, Gomez-Benlloch and associates found that the failure of deep anterior lamellar keratoplasty was mainly due to ocular surface diseases, such as limbal stem cell deficiency [23]. This is similar to our result. Another possible reason was that the longer the corneal limbus invasion length of the dermoid was, the more needles were needed to fix the graft, which was more likely to cause suture-related complications. In the current study, 2 cases of postoperative graft opacity were caused by suture-related complications. Our study found that steroid-induced high IOP was related to age, which was consistent with previous studies [24,25]. Previous studies showed that young age (< 6 years) was a risk factor for steroid-induced ocular hypertension [24].
There are several limitations in our study. First, the geometric parameters related to dermoid were obtained indirectly. Due to the eyeball curvature, the geometric parameters obtained from the two-dimensional photos might be different from the actual values. Second, the different corneal diameters, such as in children or in adults, might have an impact on length measurement. But the youngest age in our study was 4.3 years old. Anatomically, the corneal diameter of children over 3 years old is close to that of adults [26]. Therefore, the different corneal diameters in our study might have little influence on length measurement. Third, the patients in the current study were relatively older and the follow-up time was short, which might affect the prognosis of amblyopia. This study also lacked information about the preoperative and postoperative management of amblyopia. Next, we will include younger patients, extend the follow-up time, further analyze the effect of preoperative and postoperative amblyopia treatment on the long-term vision.
In a word, our study found the cornea-invaded length of dermoid was the most important factor affecting the development of postoperative visual acuity, astigmatism and amblyopia. The corneal limbus invaded length of dermoid was the most important factor affecting postoperative scar formation. The visual acuity, astigmatism and geometric parameters in patients with amblyopia were higher than those with non-amblyopia. These provide a reference for clinicians' preoperative decision-making and prognosis estimation.