Corneal Irregular Astigmatism And Visual Function On Anterior Segment Optical Coherence Tomography In TGFBI Corneal Dystrophy

Yuito Abe University of Tokyo Graduate School of Medicine Takashi Omoto (  oomoto-tky@umin.ac.jp ) University of Tokyo Graduate School of Medicine Kohdai Kitamoto University of Tokyo Graduate School of Medicine Tetsuya Toyono University of Tokyo Graduate School of Medicine Junko Yoshida International University of Health and Welfare Ryo Asaoka Seirei Hamamatsu General Hospital Satoru Yamagami Nihon University School of Medicine Takashi Miyai University of Tokyo Graduate School of Medicine Tomohiko Usui International University of Health and Welfare


Introduction
Transforming growth factor beta-induced (TGFBI) corneal dystrophy is a bilateral corneal disease characterized by an abnormal deposition of extracellular matrix. [1][2][3][4][5][6][7][8][9] Its clinical characteristics differ per disease, and lesions can appear in each corneal layer. 3 Patients with TGFBI dystrophy present with reduced best-corrected visual acuity (BCVA) with disease progression. However, the severity of opacity and visual function might not always be correlated. 10,11 With recent technological advancements, images and accurate information about the cornea and anterior segment of the eyes can be obtained. Corneal irregularities are evaluated by analyzing wavefront aberrations, which can explain the decrease in BCVA and contrast sensitivity in normal and diseased eyes. [10][11][12][13][14][15][16][17][18] Recently, with the advent of anterior segment optical coherence tomography (AS-OCT) using infrared light with high depth of eld, irregular astigmatisms in the anterior and posterior surfaces, even in cloudy corneas, can now be quanti ed. Fourier analysis is performed to quantify the shape analysis of the anterior segment, which can explain low BCVA and contrast sensitivity. 19 Stromal opacity and a higher-order aberration (HOA) of the cornea can affect BCVA in corneal dystrophy. 10,20 The current study focused on granular corneal dystrophy (GCD) and lattice corneal dystrophy (LCD), which account for most TGFBI dystrophy cases.
GCD and LCD are the same in that both are related to allelic mutations of the TGFBI gene, [21][22][23] whereas they are known to show quite different ndings in cornea; different types of deposits are observed in various layers. 3 However, few have investigated the anterior and posterior surfaces of the cornea in TGFBI dystrophy and its different subtypes in detail. 10 Therefore, in the current study, we aimed to evaluate anterior and posterior irregular astigmatism in GCD and LCD via a Fourier harmonic analysis of AS-OCT data.

Results
In total, 70 GCD eyes in 37, 35 LCD eyes in 21, and 32 control eyes in 23 participants were included in the current study. The GCD and LCD groups had signi cantly worse BCVA than the control group. Table 1 and Figure 1 show the characteristics of the participants and representative photos of the groups, respectively. In the analysis according to phenotypes, when comparing the GCD and control eyes, there was no signi cant difference in terms of the anterior and posterior components between the two groups ( Figure 2). Furthermore, these components were not associated with BCVA (Table 2). In the comparison between the LCD and normal groups, the anterior (asymmetry) (P = 0.030) and all of the posterior components (P < 0.001) signi cantly differed ( Figure 2). The Fourier components of the LCD eyes were signi cantly associated with BCVA in the anterior (regular astigmatism, asymmetry, and higher-order irregularity; P = 0.012, <0.001, and <0.001) and posterior (higher-order irregularity; P = 0.029) components (Table 2). In the analysis according to subtypes of the GCD group, the eyes of patients with GCD1 and controls were similar ( Figure 3). There was no association between BCVA and each of the components (Table 2). However, the posterior components (asymmetry, higherorder irregularity) of GCD2 were signi cantly higher than those of the normal eyes (P = 0.0012, 0.0071) ( Figure 3). Nevertheless, the posterior components were not signi cantly correlated with BCVA (Table 2).
Only the posterior components of the eyes in patients with LCD3A signi cantly differed (spherical; P = 0.045) ( Figure 3). The anterior components (asymmetry, higher-order irregularity; P = 0.047, 0.036) were positively associated with BCVA. Meanwhile, there was a negative correlation between the posterior component (regular astigmatism; P = 0.0048) and BCVA (Table 2).

Discussion
The speci c pro le of the layer-by-layer Fourier components in GCD and LCD eyes, in addition to the association between them and BCVA, was discussed in this research. Moreover, not only each phenotype but also each subtype was analyzed. GCD and LCD are known to show quite different ndings in cornea; both amyloid and hyaline deposits in GCD eyes, whereas amyloid deposits in LCD eyes and the depth of the deposits vary by the subtypes. 3 These ndings were analyzed using AS-OCT in detail and these results should be bene cial for clinicians when evaluating visual function of the patients.
In GCD eyes, interstitial opacities between granular deposits were often con ned to the anterior subepithelial layer (Figure 1). Structural disturbances in the anterior and posterior surfaces were apparently smooth. However, there was a slight but signi cant increase in posterior irregular astigmatism (asymmetry, higher-order irregularity) in GCD2 eyes ( Figure 3). Interestingly, the posterior, but not the anterior, components increased despite the relatively super cial existence of opacity. In a previous report by Yagi-Yaguchi et al., visual acuity was positively correlated with opacity grade, age, astigmatism in GCD2 eyes, but not with HOAs. 10 The HOA value in the correlation analysis in a previous study was not calculated layer by layer. 10 The current study analyzed the anterior and posterior components separately. However, the results were similar. That is, the Fourier components of the GCD eyes were not correlated with BCVA, which did not change when analyzed according to each subtype. Nevertheless, the GCD eyes had signi cantly lower BCVA than the control eyes. The factors causing a decrease in BCVA are challenging to analyze. However, the relatively low contributions of astigmatism factors reinforce the possibility that scattering caused by opacity may have a stronger effect on visual acuities.
In LCD eyes, there was an increase in Fourier components, and BCVA in the anterior and posterior surfaces was positively associated with these components. In the analysis according to each LCD subtype, the results differed between the subtypes and were characteristic to each subtype. As shown in Figure 1, the cornea of one patient with LCD1 had stromal opacities with various re ectivity and sizes near the anterior surface. Previously, Yagi-Yaguchi et al. showed that visual acuity was correlated with HOAs, but not with grade, age, and astigmatism, in patients with LCD1, unlike GCD. 10 The study results were similar. That is, there was an increase in anterior components (regular astigmatism, asymmetry, and higher-order irregularity), which were associated with BCVA.
As shown in Figure 1, the cornea of patients with LCD4 commonly presented with opacities in the posterior surface, which could be described as protrusions in the anterior chamber. In LCD4 eyes, all components of the posterior surface increased, and almost all were positively associated with BCVA. The results of the current research were consistent with those of previous studies. 27,28 In LCD1, PTK is sometimes performed because of a signi cant anterior surface lesion. 29 Nevertheless, in LCD4, corneal transplantation is still considered as the primary treatment as it can improve BCVA since the posterior irregularity cannot be corrected using rigid gas permeable contact lenses and the posterior opacity cannot be managed with PTK. Each subtype of LCD had different characteristics. In each subtype, the results of the corneal shape analysis were similar to the typical AS-OCT imaging ndings and were consistent with what has been noted clinically.
The current study had several limitations. The degree of opacity was not evaluated. As mentioned above, the effect of scattering caused by opacity on visual function cannot be ignored. Thus, the quanti ed data about opacity and its association with BCVA must be investigated. Subtype classi cation was based on clinical, not genetic, diagnosis. Furthermore, some patients who underwent cataract surgery were included. The effect of cataract surgery on corneal shapes cannot been completely ruled out, 30-32 even though these changes are temporary based on several reports. [33][34][35] In conclusion, in GCD eyes, except for a slight increase in the posterior components of GCD2 based on a subtype analysis, none of the components differed from the control, and BCVA was not signi cantly associated with these components. In LCD eyes, the anterior and posterior components increased, and BCVA was signi cantly associated with the anterior and posterior components. In the analysis according to each LCD subtype, the results differed between the subtypes; both anterior and posterior components differed and anterior components positively associated with BCVA in LCD1, only the posterior components differed and the associations with BCVA were positive in anterior and negative in posterior components in LCD3A and posterior components increased and there was a positive association between posterior components and BCVA in LCD4.

Methods
This retrospective observational study was approved by the institutional review board of University of Tokyo Hospital (20200006NI). Informed consent was obtained in the form of opt-out in the website, following the above-mentioned approval. This research was performed according to the tenets of the Declaration of Helsinki.

Participants
Patient data were obtained from the University of Tokyo database. GCD and LCD patients who underwent AS-OCT imaging, in addition to routine examinations, such as slit-lamp microscopy, fundus examinations, and measurements of BCVA, were included. Similarly, patients who had no corneal disease, visual impairment, and other ocular diseases (BCVA of 20/20 or higher), and those were age-matched were included as normal controls. Patients with inaccurate diagnosis, corneal epithelial defects, corneal scarring, and a history of corneal surgery, such as phototherapeutic keratectomy (PTK) and corneal transplantation, which can affect the corneal surface, were excluded.
Two corneal specialists (YA and TM) identi ed dystrophy phenotypes and subtypes according to the International Classi cation of Corneal Dystrophies (IC3D) using the clinical information in the medical records. 3 Then, 4 eyes of 2 patients with GCD type 1 (GCD1), 66 eyes of 35 patients with GCD type 2 (GCD2), 11 eyes of 7 patients with LCD type 1 (LCD1), 11 eyes of 6 patients with LCD type IIIA (LCD3A), and 13 eyes of 8 patients with LCD type IV (LCD4) were included in the nal analysis. In total, 32 eyes of 23 controls were also included. Patients with LCD2 were excluded because the condition was different from TGFBI dystrophy. 3 Patients with other vision-affecting diseases were not included in the analysis of the association between BCVA and Fourier components (n = 9 eyes with GCD [cataract, glaucoma, and diabetic retinopathy] and n = 6 eyes with LCD [cataract, age-related macular degeneration, and central nervous system disorder]).

Oct Assessment
All AS-OCT images were obtained using SS-1000 CASIA or SS-2000 CASIA2 (Tomey Corporation, Inc., Aichi, Japan). Corneal shape analysis was performed using two different layered assessments: axial power map of the anterior and posterior surfaces. Values within 3-mm diameter were analyzed. Via a Fourier harmonic analysis, corneal dioptric data were expanded into four components: spherical, regular astigmatism, asymmetry, and higher-order irregularity.

Statistical Analysis
First, we compared the Fourier components between the phenotypes and control eyes. Second, the association between BCVA and the components was analyzed. Next, similar analyses were performed according to each subtype per phenotype.
TGFBI dystrophy, as is well known, develops bilaterally. 1-3 Therefore, we used the multivariate linear mixed-effects model in which the random effect was the subjects, with adjusting for age. The model adjusts for the hierarchical structure of the data, modeling in a way in which measurements are grouped within subjects to reduce the possible bias of including both eyes of one patient. 24,25 This was followed by Dunnett's test for multiple comparisons when comparing the values between each sub-group in the phenotypes and the control. 26

Additional Information
The authors declare no competing interests.