Sectoral and Layer-Specific Corneal Thinning in Thyroid Eye Disease: A Cross-Sectional Study

Abstract

Purpose

Corneal thickness in patients with thyroid eye disease (TED) and the layer or specific region of the cornea primarily affected by corneal thickness changes remain unclear. This study aimed to examine and compare the differences in corneal thickness across various layers in patients with TED and a control group, focusing on identifying the specific corneal region where these changes are most pronounced.

Methods

This cross-sectional retrospective study included 57 eyes from 29 patients with TED and 18 eyes from 9 control participants. Basic eye examinations and corneal topography measurements were performed using a Galilei dual Scheimpflug camera and RTVue. Total corneal thickness, corneal epithelial thickness, and corneal stromal thickness (CST) were evaluated.

Results

Screening indices related to corneal irregularity, such as keratoconus prediction index and opposite sector index, were numerically higher in the TED group but remained within normal limits. In all corneal sectors, the mean total corneal thickness and CST were significantly lower in the TED group (e.g., central CST, 480.33 ± 28.79 µm in TED vs. 498.94 ± 16.04 µm in controls; P = 0.049). Notably, no significant differences in corneal epithelial thickness were observed between the two groups across all evaluated sectors.

Conclusions

Patients with TED demonstrated diffuse stromal thinning that preserved the normal sectoral distribution pattern, without focal ectatic changes or topographic abnormalities. These findings highlight the need for layer-specific corneal assessment in patients with TED, particularly when evaluating corneal health and planning anterior segment procedures.

Translational Relevance

These findings indicate that TED is linked to structural changes in the cornea, especially within the stromal layer, which may have important clinical implications for managing TED-related ocular complications.

Introduction

Thyroid eye disease (TED), characterized by autoimmune inflammation in the tissues surrounding the eyes, is a disease with unclear mechanisms. Its main features include enlarged extraocular muscles and proliferated orbital fibroblasts in the eye sockets, leading to various ocular manifestations.^1,2 These manifestations—such as proptosis, eyelid retraction, and strabismus—may alter the mechanical environment of the globe and potentially influence corneal shape or biomechanics.³ Furthermore, some studies have reported changes in the corneal microstructure with TED.^4,5 Khalil et al.⁴ observed that the permeability of the corneal epithelium is significantly higher in TED than in the control group. Villani et al.⁵ revealed that patients with TED have lower corneal nerve density and more reactive stromal keratocytes.

Consensus regarding central corneal thickness (CCT) is lacking in patients with TED. Bahçeci et al.⁶ reported that the CCT increases temporarily in patients with hypothyroidism and returns to normal after thyroxine replacement therapy. This increase has been attributed to mucopolysaccharide accumulation within the corneal stroma.⁷ In contrast, Konuk et al.⁸ discovered that hyperthyroidism and severity of orbital disease do not change the CCT in TED. These observations align with the findings of Karabulut et al.,⁹ who observed no significant difference in CCT between patients with TED and the control group. Conversely, another study reported that patients with TED exhibited a significantly thinner CCT compared with controls.³

Despite numerous studies analyzing changes in corneal thickness in patients with TED, consistent conclusions are lacking. Moreover, no study has identified the layer or specific region of the cornea that is primarily affected by changes in corneal thickness. Therefore, this study aimed to evaluate layer-specific and sector-specific differences in corneal thickness in patients with TED compared with controls, with the goal of identifying the corneal regions most vulnerable to thickness changes in TED.

Methods

Study Design

This retrospective study included 75 eyes of 38 patients, including 29 patients with TED and 9 healthy controls, who visited a tertiary medical center in Taiwan between January and December 2023. All enrolled patients were aged 20 years or older and had been diagnosed with TED based on the clinical criteria defined by the European Group on Graves’ Orbitopathy guidelines at least six months prior to enrollment. The patients had inactive disease status, defined by a clinical activity score (CAS) of 2 or less at the time of evaluation¹⁰; this selection criterion was applied to minimize the confounding effects of acute inflammation, periorbital edema, and unstable ocular surface conditions, which could interfere with imaging quality and structural corneal measurements. One eye in the TED group was excluded owing to incomplete imaging data caused by poor fixation during the examination. The exclusion criteria included a CAS of 3 or higher, a history of eye surgery, previous ocular trauma, using topical medications, or any active ocular surface disease that could influence corneal topographic findings. However, patients with TED and dry eye disease who used eye drops to manage their condition were included in the study. In addition, we excluded participants with a history of contact lens wear, systemic autoimmune or thyroid disorders other than TED, or corneal abnormalities suggestive of subclinical keratoconus (keratoconus probability of >11.6), according to previously validated thresholds.¹¹ Additionally, a control group of 18 eyes from 9 patients was enrolled, matched using age and sex with the TED group. These patients had visited the hospital for eye issues unrelated to TED and underwent standard ophthalmological examinations, corneal epithelial thickness (CET), corneal pachymetry, and corneal topography. Control patients were required to have normal corneal topography and tomography findings; no history of ocular surgery, trauma, or contact lens wear; and no active ocular surface or systemic autoimmune disease. The relatively small number of controls was due to the strict eligibility criteria, the requirement for age and sex matching with the TED group, and the need for high-quality paired corneal tomography and epithelial thickness imaging obtained during the study period. All images were reviewed by two independent examiners blinded to group allocation, and scans with segmentation errors or poor fixation were excluded to ensure data reliability. The duration of TED was recorded for each patient and defined as the time interval between the initial clinical diagnosis of TED and the date of corneal imaging.

The study followed the principles of the Declaration of Helsinki. The institutional review board of Tri-Service General Hospital, Taipei, Taiwan, approved the study (TSGHIRB No. B202105186), waived the requirement for informed consent from participants because of its retrospective nature and permitted access to follow-up clinical records.

Ophthalmological Examinations

During the initial visit, all patients provided basic personal and medical history information. Patients’ eyes were subjected to standard ophthalmological examinations, which included assessments of best-corrected visual acuity, intraocular pressure, automated refraction, biomicroscopic examination, extraocular eye movement tests, Hertel exophthalmometry, and dilated fundus examination.

Corneal Topography

Corneal topography was measured using a Galilei dual Scheimpflug camera, which improved accuracy by reducing motion errors during the examination. Several parameters, including CCT, center/surround index, differential sector index, irregular astigmatism index, inferior–superior index, keratoconus prediction index, keratoconus probability, opposite sector index, surface asymmetry index, standard deviation of corneal power, surface regularity index, average total central corneal power over a 4.0-mm central zone, the steep meridian of total corneal power, simulated keratometry, simulated keratometry astigmatism with axis, and anterior instantaneous astigmatism along the axis, were recorded for both eyes.

CET and Corneal Pachymetry

CET and total corneal thickness (TCT) were measured using the RTVue-100 spectral-domain OCT system (Optovue, Fremont, CA), which operates at a central wavelength of 830 nm with axial and lateral tissue resolutions of 5 and 15 µm, respectively. Pachymetry data were obtained using the Pachy-Wide epithelial mapping protocol. The RTVue software automatically generated epithelial and stromal thickness maps through built-in segmentation of the anterior and posterior corneal boundaries. In this system, the corneal map is displayed as a 9-mm diameter circle. The system divides it into a central 2-mm diameter circle and concentric rings at 5, 7, and 9 mm. The central 2-mm circular zone is defined as the C sector, the paracentral 2- to 5-mm as ring 1, the mid-peripheral 5- to 7-mm as ring 2, and the peripheral 7- to 9-mm as ring 3. These three outer rings are further divided into eight sectors: superior (S), superior temporal (ST), temporal (T), inferior temporal (IT), inferior (I), inferior nasal (IN), nasal (N), and superior nasal (SN). Figure 1 shows each sector further subdivided from the inside out into S1, S2, and S3 and ST1, ST2, and ST3. The corneal thickness in each sector is the average of the three regions within it. All epithelial and pachymetry maps were independently reviewed by two experienced examiners who were masked to TED vs. control group assignment. Automated segmentation was used for all measurements, and no manual boundary correction was performed. Scans with segmentation errors, motion artifacts, poor fixation, or inadequate signal strength (signal strength index <45) were excluded. Because only scans with complete agreement between the two graders were included, additional intergrader or intragrader reproducibility testing was not required. Pachymetry measurements (CCT, TCT, and CET) were obtained exclusively from the RTVue OCT system; the Galilei device was used only for anterior corneal topography.

Figure 1.

Pachymetry corneal mapping (9 mm, 25 sectors) of the right eye, mirrored for the left eye.

Statistical Analysis

Data are presented as the mean ± standard deviation and analyzed using SPSS version 23 (SPSS, Chicago, IL). The Kolmogorov–Smirnov test was applied to check the normality of data distribution. A linear mixed-effects model was used to compare each parameter between groups, with group as the fixed effect and patient ID as the random effect. The linear mixed-effects model was also applied to assess sectoral differences in corneal thickness within each group, treating sector as a fixed effect with patient ID as the random effect. This approach accounts for the nonindependence of measurements from both eyes or multiple sectors of the same individual. Sensitivity analyses additionally adjusted for age and sex as fixed effects, while retaining patient ID as a random effect. Statistical significance was set at P value of less than 0.05.

Results

Demographic Characteristics

This study included 25 women and 13 men. The TED group had 29 patients and 57 eyes, and the control group had 9 patients and 18 eyes. Demographic characteristics, including age and sex, did not significantly differ between groups (age, 54.14 ± 11.66 years vs. 51.06 ± 17.32 years; P = 0.49). Baseline demographic characteristics were comparable between the two groups and are summarized in Table. Because additional ocular biometric parameters were not available consistently in the retrospective records, age and sex represent the complete set of baseline variables collected for this cohort.

Table.

Sectoral Corneal Thickness Values (TCT, CET, CST) in TED and Control Groups

	TED Group (n = 57 Eyes), Mean ± SD	Control Group (n = 18 Eyes), Mean ± SD	LMM Δ [95% CI]	P Value
Age (years)	54.14 ± 11.66	51.06 ± 17.32		0.49
Sex
Male	11	2
Female	18	7
TCT (C)	532.72 ± 29.72	552.50 ± 17.15	Δ = 20.20 [0.50 to 39.90]	0.04*
Mean TCT (S)	621.60 ± 37.97	658.93 ± 45.70	Δ = 36.65 [9.67 to 63.63]	0.008**
Mean TCT (SN)	611.34 ± 36.54	641.96 ± 28.18	Δ = 30.13 [6.60 to 53.66]	0.012*
Mean TCT (N)	599.56 ± 33.21	625.11 ± 27.09	Δ = 25.46 [4.18 to 46.74]	0.019*
Mean TCT (IN)	594.39 ± 30.05	618.50 ± 24.03	Δ = 24.27 [5.55 to 42.99]	0.011*
Mean TCT(I)	589.72 ± 30.40	610.33 ± 24.38	Δ = 20.78 [1.02 to 40.54]	0.039*
Mean TCT (IT)	577.05 ± 32.77	597.13 ± 19.27	Δ = 20.09 [0.20 to 39.98]	0.048*
Mean TCT (T)	577.81 ± 32.44	596.94 ± 19.27	Δ = 18.89 [−1.93 to 39.71]	0.075
Mean TCT (ST)	601.86 ± 36.15	627.43 ± 26.10	Δ = 24.96 [1.23 to 48.69]	0.039*
CET (C)	52.39 ± 3.51	53.56 ± 3.52	Δ = 1.10 [−1.17 to 3.36]	0.343
Mean CET (S)	49.17 ± 4.20	48.19 ± 2.94	Δ = −0.94 [−3.42 to 1.53]	0.455
Mean CET (SN)	50.50 ± 3.38	49.56 ± 3.19	Δ = −0.94 [−3.06 to 1.18]	0.384
Mean CET (N)	51.24 ± 3.43	51.91 ± 2.54	Δ = 0.65 [−1.37 to 2.67]	0.528
Mean CET (IN)	52.17 ± 4.14	52.30 ± 2.21	Δ = 0.11 [−2.12 to 2.34]	0.921
Mean CET (I)	52.50 ± 4.57	52.04 ± 2.05	Δ = −0.47 [−2.90 to 1.95]	0.703
Mean CET (IT)	51.53 ± 3.18	51.89 ± 2.09	Δ = 0.33 [−1.48 to 2.14]	0.72
Mean CET (T)	49.35 ± 2.97	50.54 ± 2.51	Δ = 1.18 [−0.62 to 2.99]	0.199
Mean CET (ST)	49.08 ± 3.79	48.78 ± 2.64	Δ = −0.32 [−2.69 to 2.04]	0.788
CST (C)	480.33 ± 28.79	498.94 ± 16.04	Δ = 19.17 [0.02 to 38.32]	0.049*
Mean CST (S)	572.43 ± 37.79	610.74 ± 45.10	Δ = 37.59 [10.88 to 64.29]	0.006**
Mean CST (SN)	560.84 ± 36.15	592.41 ± 27.55	Δ = 31.07 [7.85 to 54.29]	0.009**
Mean CST (N)	548.32 ± 32.53	573.20 ± 25.84	Δ = 24.82 [4.08 to 45.55]	0.019*
Mean CST (IN)	542.22 ± 29.41	566.20 ± 23.38	Δ = 24.17 [5.90 to 42.44]	0.001**
Mean CST (I)	537.22 ± 30.40	558.30 ± 24.23	Δ = 21.26 [1.63 to 40.90]	0.034*
Mean CST (IT)	525.52 ± 32.05	545.24 ± 18.60	Δ = 19.77 [0.39 to 39.15]	0.046*
Mean CST (T)	528.46 ± 31.45	546.41 ± 17.78	Δ = 17.72 [−2.33 to 37.76]	0.083
Mean CST (ST)	552.78 ± 34.52	578.65 ± 25.45	Δ = 25.31 [2.74 to 47.88]	0.028*

C, central; LMM, linear mixed-effects models; SD, standard deviation.

All sectoral values (except central) represent the mean of three concentric subregional measurements within each sector.

Δ represents the difference calculated as Control – TED. Positive values indicate thinner corneas in the TED group.

^* P < 0.05.

^** P < 0.01.

Corneal Topography Parameters

The linear mixed-effects model analysis revealed no statistically significant differences between the TED and control groups across all corneal topographic parameters. Although several indices (e.g., keratoconus prediction index, keratoconus probability, and opposite sector index) were numerically higher in the TED group, none attained statistical significance. Detailed topographic values are provided in Supplementary Table S1.

Corneal Thickness Parameters

The CET and TCT of patients and the control group were averaged based on different sectors and compared (Table). Additionally, Table presents the corneal stromal thickness (CST), calculated by subtracting the epithelial thickness from the TCT. The TCT was significantly reduced in the TED group compared with that in the control group across all evaluated sectors except the T sector. Specifically, significant reductions were observed in the central (Δ = 20.20; P = 0.04), S (Δ = 36.65; P = 0.008), SN (Δ = 30.13; P = 0.012), N (Δ = 25.46; P = 0.019), IN (Δ = 24.27; P = 0.011), I (Δ = 20.78; P = 0.039), IT (Δ = 20.09; P = 0.048), and ST (Δ = 24.96; P = 0.039) sectors in the TED group compared with those in the control group. Although the difference in the T sector did not reach statistical significance (Δ = 18.89; 95% confidence interval, −1.93 to 39.71; P = 0.075), a decreasing trend in the TCT was observed in patients with TED. CST followed a similar pattern. The TED group demonstrated a significantly thinner CST in the central (Δ = 19.17; P = 0.049), S (Δ = 37.59; P = 0.006), SN (Δ = 31.07; P = 0.009), N (Δ = 24.82; P = 0.019), IN (Δ = 24.17; P = 0.001), I (Δ = 21.26; P = 0.034), IT (Δ = 19.77; P = 0.046), and ST (Δ = 25.31; P = 0.028) sectors compared with the control group. Similar to TCT, a nonsignificant reducing trend in CST was observed in the T sector (Δ = 17.72; P = 0.083). However, no significant difference was observed in the mean CET between the TED and control groups. Moreover, significant differences were observed across all sectors of the cornea within each group, regardless of mean TCT, CET, or CST (P < 0.001). Figure 2A shows the changes in TCT, CET, and CST in different sectors between the TED and control groups. Apart from the central cornea being the thinnest in the TED and control groups, the TCT showed a gradual decreasing trend in the S, SN, N, IN, I, IT, and T sectors, followed by an increase in thickness in the ST sector. A similar pattern was observed in the CST. In contrast, the CET was the thinnest in the ST sector in the TED group and the S sector in the control group. Figure 2B presents the histogram of corneal layer thickness for the TED and control groups, using the corneal sectoral map for display. This approach allowed for a clearer understanding of the thickness in different sectors of the cornea. Therefore, this study presented two cases of patients with TED and two cases of the control group with corneal epithelial and stromal thickness measurements.

Figure 2.

(A) Histogram of pachymetry map sectors comparing TED and control groups. (B) Histogram displaying corneal thickness differences between TED and control groups across different sectors of the corneal map. P values represent between-group comparisons based on linear mixed-effects models. NS, not significant. *P < 0.05; **P < 0.01.

Representative Cases

Representative examples of corneal thickness maps are shown in Figure 3. Compared with controls (Figs. 3B, 3D), which exhibited a normal pachymetric contour and greater overall corneal thickness, eyes with TED (Figs. 3A, 3C) demonstrated diffuse thinning across the corneal layers. Specifically, the TCT in the illustrated TED examples ranged from 468 to 490 µm, whereas values in the control examples were notably higher, ranging from 551 to 588 µm. This decrease was primarily attributable to stromal thinning; the CST showed a distribution pattern similar to that of TCT, although the CET remained comparable between the two groups. These qualitative observations are consistent with the group-level quantitative analysis.

Figure 3.

Representative corneal thickness maps from two TED cases and two control cases. (A and C) TED eyes demonstrate diffuse stromal thinning, particularly in the S and IN sectors, consistent with group-level quantitative findings. (B and D) Control eyes show normal pachymetric contour and greater overall stromal thickness.

Discussion

TED is a complicated autoimmune condition that causes molecular changes in the corneal stroma and ocular surface, along with significant structural changes in the orbital and periorbital tissues.¹² Nevertheless, few studies have focused on corneal topography and corneal thickness changes in patients with TED. A previous study has noted that TED can be associated with subtle changes in anterior corneal measurements compared with controls. Additionally, the TED group had significantly lower CCT than the control group.³ This study is the first to investigate the thickness of different corneal layers in patients with TED. Although the differences in most corneal surface indices did not attain statistical significance, a consistent pattern of corneal thinning—particularly involving the stromal layer—was observed in TED eyes.

Corneal topography revealed no significant differences in parameters such as Q value, surface asymmetry, and indices used for screening corneal irregularity. Although some metrics (e.g., keratoconus prediction index, keratoconus probability, and opposite sector index) were numerically higher in the TED group, these values remained within the normal range and did not attain statistical significance. The lack of significant differences in topographic parameters may reflect several factors, including the predominance of inactive TED in this cohort, variability within groups, and the limited sample size, which may reduce the ability to detect subtle anterior surface changes. In addition, the Galilei topography system may be less sensitive to early or mild corneal shape alterations, and larger studies are needed to validate these observations. Previous studies^13–16 have reported associations between thyroid dysfunction and alterations in corneal biomechanics, thickness, or ocular surface homeostasis. These relationships may reflect systemic hormonal influences on corneal extracellular matrix remodeling. However, the evidence remains inconsistent, and the present study cannot directly these mechanisms address because thyroid hormone levels were not collected. Further work is needed to explore whether systemic thyroid status contributes to corneal stromal changes observed in TED.

For pachymetry analysis, this study revealed that patients with TED had a significantly lower mean TCT in nearly all evaluated sectors except the temporal region, where a decreasing trend was noted but did not attain statistical significance. Importantly, this same pattern was observed in CST, where the TED group exhibited a significantly thinner stroma in most sectors, again sparing only the temporal side. Importantly, the observed stromal and total corneal thinning in the TED group remained present after adjusting for age and sex in a mixed-effects model, supporting that these differences are attributable to disease status rather than demographic variation. In contrast, CET remained largely comparable between the two groups in all regions. The congruence of these findings in both TCT and CST reinforces the notion that the corneal thinning in TED predominantly reflects stromal involvement. Opinions differ regarding the effect of TED on corneal thickness. Some studies report no effect,^8,9 whereas others indicate that hypothyroidism can cause temporary corneal thickening, generating inconsistent results.⁷ However, none of these studies specify the corneal layer that becomes thicker or thinner. This study specifically highlights stromal thinning in TED. Although the underlying mechanism remains unclear, autoimmune-mediated stromal matrix remodeling has been suggested and warrants further investigation.

On the contrary, this study discovered that in the TED and control groups, the TCT and CST gradually decreased from the S, SN, N, IN, I, to the IT sectors, with the thickness being relatively thinnest at the IT and T sectors. Subsequently, an increase in thickness was observed in the ST region. These findings align with previous findings obtained using different evaluation tools.^17,18 The CETs at the S and ST sectors were relatively thinner in both groups, similar to findings from previous studies.^18,19 This asymmetry could be attributed to eyelid-induced abrasion. In 1994, Reinstein et al.²⁰ suggested that the thickness profile of the corneal epithelium could be regulated by blinking and corneal friction. Doane²¹ discovered that the speed of the descent of the upper eyelid during blinking reaches its maximum when it crosses the visual axis. Thus, the eyelid could be rubbing against the corneal epithelium, exerting more pressure on the superior part of the cornea than the inferior section. This finding might explain why the superior CET is thinner than that of the inferior region. The eyelid conditions in patients with TED may differ from those of normal individuals; however, this study observed no significant difference in the thinnest CET between patients with TED and the control group. A possible reason is that patients with TED included in this study had a CAS of 2 or less, indicating relative stability in their condition.

This study had some limitations. First, the relatively small sample size, particularly in the control group (9 patients and 18 eyes), may have affected the generalizability of the results. However, the control cohort was stringently selected according to strict exclusion criteria (no contact lens wear, ocular surgery or trauma, or ocular surface or systemic autoimmune disease) and was age and sex matched to the TED group, with all eyes having high-quality paired corneal tomography and epithelial thickness maps. Second, because only patients with inactive disease (CAS ≤ 2) were included and detailed CAS stratification was unavailable in this retrospective dataset, the findings may not be fully representative of active TED. Third, although all RTVue epithelial and pachymetry maps were independently reviewed by masked graders, formal intergrader or intragrader reproducibility testing was not performed, and this remains a methodological limitation. Although CET and CST measurements relied on automated segmentation without manual correction, all scans with segmentation or fixation errors were excluded to ensure data quality; although this strategy minimizes operator-dependent bias, it may also reduce analyzable data. Fourth, many patients with TED had dry eye disease, which could influence the measurement of corneal thickness. The dry eye condition was managed with artificial tears; however, it remained a confounding factor when interpreting the results. Fifth, two different imaging platforms (RTVue for pachymetry and Galilei for topography) were used; although they assess different aspects of corneal structure and are not directly interchangeable, supplementary analyses confirmed consistent group-level trends in CCT across both modalities. Finally, because the study design was cross-sectional, confirming the relationship between corneal thickness, corneal topographic parameters, and TED was challenging. Larger studies are needed to further evaluate this correlation.

Conclusions

Patients with TED demonstrated diffuse stromal thinning in the absence of significant anterior corneal surface abnormalities. Compared with controls, TED eyes showed consistently lower total and stromal thicknesses across most corneal sectors; epithelial thickness remained comparable. The sectoral pattern of thickness distribution was similar to that of the general population. These findings suggest that TED is associated with layer-specific corneal structural changes, particularly involving the stroma, which may have clinical relevance in the evaluation of TED-related ocular manifestations.