Longitudinal Changes in Corneal Epithelial and Central Corneal Thickness in Type 2 Diabetes Assessed by AS-OCT: A Real-World Study

Abstract

Purpose

To evaluate longitudinal changes in central corneal epithelial thickness (CCET) and central corneal thickness (CCT) using anterior segment optical coherence tomography (AS-OCT) over 12 months in adults with type 2 diabetes receiving one of two routinely prescribed oral antidiabetic regimens, and to explore their association with glycemic changes.

Methods

In this retrospective observational cohort study, 40 adults with type 2 diabetes receiving one of two routinely prescribed treatment regimens (vildagliptin–metformin, n=19; empagliflozin–metformin, n=21) underwent AS-OCT at baseline, 3 months, and 12 months. Glycated hemoglobin and fasting plasma glucose were assessed at each visit. Precision analysis indicated detectable between-group differences of 4.7 µm for CCT and 6.1 µm for CCET.

Results

Both groups showed significant within-group glycated hemoglobin improvement (mean reduction: −2.28% and −2.88% in the vildagliptin–metformin and empagliflozin–metformin groups, respectively; p<0.05) without between-group differences. CCT and CCET showed no statistically detectable longitudinal change over 12 months, with no significant differences in longitudinal change between groups. Glycated hemoglobin change showed a modest independent association with CCT change (β = 0.81 µm per 1% glycated hemoglobin; 95% CI 0.20–1.42; p=0.011; n=32).

Conclusions

Corneal structural parameters showed no statistically detectable longitudinal change or between-group difference over 12 months. Given the modest sample size and observational design, these findings should be considered hypothesis-generating and require confirmation in larger prospective studies.

Graphical Abstract

Introduction

Type 2 diabetes (T2D) affects an estimated 589 million adults worldwide, with prevalence projected to reach 853 million by 2050, reflecting a growing burden of chronic microvascular and organ-specific complications.¹ This increasing burden underscores the importance of identifying diabetes-related structural changes and of interdisciplinary care, including integration with eye-care services.²

Ocular involvement represents a critical yet often underrecognized dimension of diabetes-related organ damage. Although diabetic retinopathy remains the most recognized ocular complication, growing evidence indicates that diabetes-related structural and functional alterations may occur well before clinically overt retinopathy develops.^3,4 These early changes are not confined to the posterior segment but also involve anterior segment structures, including the cornea, suggesting that ocular surface tissues may reflect early microvascular and metabolic stress.^3–5

The corneal epithelium is particularly vulnerable to metabolic dysregulation. Experimental and clinical studies have demonstrated that chronic hyperglycemia disrupts epithelial barrier integrity, alters cellular turnover, and promotes subclinical edema through mechanisms involving oxidative stress, advanced glycation end products, and inflammatory pathways.^5,6 Consequently, epithelial alterations are increasingly recognized as a component of diabetic keratopathy, even in patients without clinically overt retinopathy or advanced disease.^5–7

Advances in anterior segment optical coherence tomography (AS-OCT) have enabled high-resolution, noninvasive assessment of corneal layers, including central corneal thickness (CCT) and central corneal epithelial thickness (CCET). Several clinical studies have reported increased CCT and CCET values in patients with T2D compared with individuals without diabetes, suggesting that corneal thickness parameters may reflect early metabolic and diabetes-related tissue stress.^4,7 However, this evidence is predominantly cross-sectional, and longitudinal data on corneal epithelial thickness changes over time remain limited. Comparative longitudinal evidence on corneal structural changes in patients receiving different glucose-lowering regimens also remains limited. Accordingly, AS-OCT–derived corneal thickness metrics have emerged as objective structural biomarkers of diabetes-related corneal involvement and metabolic stress.^4,7,8

Given that corneal structure is influenced by microvascular and metabolic homeostasis, systemic antidiabetic therapies may influence AS-OCT–derived thickness metrics beyond glycemic control alone. Systemically, sodium–glucose cotransporter 2 inhibitors (SGLT2i) have been reported to improve endothelial function, modulate intravascular volume, and favorably influence microvascular hemodynamics;^9,10 these effects were not directly assessed in ocular tissue in the present study. In contrast, dipeptidyl peptidase-4 inhibitors (DPP-4i) primarily act through incretin-mediated pathways, and available evidence suggests that their effects on microvascular and endothelial function may be indirect and variable.^11,12

However, whether such differences translate into measurable corneal structural changes remains unclear. In particular, it is uncertain whether glycemic improvement is accompanied by epithelial or stromal remodeling, or whether corneal structural parameters instead reflect cumulative metabolic burden rather than short-term glycemic changes. Clarifying whether corneal structural parameters change over time during diabetes treatment may help define their potential role in longitudinal ocular assessment.

We hypothesized that both treatment regimens would improve glycemic control and explored whether longitudinal changes in corneal structural parameters differed between the two treatment groups. Therefore, the present study aimed to evaluate longitudinal changes in AS-OCT–derived corneal structural parameters in adults with T2D during improved metabolic control in a real-world clinical setting. We also explored whether changes in glycemic control were associated with corneal structural dynamics during follow-up.

Materials and Methods

This study was conducted and reported in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement.

Study Design and Setting

This retrospective comparative cohort study was conducted at a single tertiary care center (Istanbul Medipol University Hospital, Istanbul, Türkiye). We compared corneal structural parameters in adults with T2D receiving one of two fixed-dose oral antidiabetic combinations: empagliflozin–metformin (5/1000 mg twice daily) or vildagliptin–metformin (50/1000 mg twice daily).

Participants

Electronic clinical records from May 25, 2021, to May 25, 2023, were retrospectively reviewed. For each eligible participant, baseline (T0), 3-month (T1), and 12-month (T2) clinical, biochemical, and AS-OCT data were extracted when available. Baseline (T0) was defined as the visit at which the patient initiated the antidiabetic regimen of interest (vildagliptin–metformin or empagliflozin–metformin); all participants were treatment-naive at this visit, with no prior use of antidiabetic medication. Patients were identified retrospectively from the electronic medical records of the internal medicine outpatient clinic using International Classification of Diseases, 10^th Revision (ICD-10) diagnostic codes for type 2 diabetes mellitus within the hospital electronic health record system, restricted to patients with available AS-OCT examinations performed in the ophthalmology outpatient clinic during the study period.

Inclusion Criteria

Adults aged 20–65 years were eligible if they had T2D diagnosed according to American Diabetes Association criteria.¹³ Participants were included only if they were treated exclusively with one of the two predefined fixed-dose combination regimens (see Section 2.1), without additional oral antidiabetic agents. Eligible patients were required to have available baseline (T0) and 3-month (T1) metabolic and clinical data, as well as baseline and follow-up AS-OCT measurements of CCT and CCET with adequate image quality for analysis. For longitudinal analyses extending to 12 months (T2), patients with available AS-OCT and/or biochemical data at T2 were included in the corresponding analyses. Based on treatment regimen, participants were assigned to:

• Group A: vildagliptin–metformin

• Group B: empagliflozin–metformin

Exclusion Criteria

To minimize ocular and pharmacological confounding and ensure AS-OCT measurement reliability, patients were excluded if they had a history of ocular surgery, uveitis, glaucoma, documented keratitis, keratoconus, or corneal scarring, clinically significant dry eye disease, contact lens use, or high refractive error (spherical equivalent > ±3.00 D). Clinically significant dry eye was defined using objective tear-film testing: only patients with a Schirmer I test ≥15 mm/5 min and a non-invasive tear break-up time (NIBUT) ≥10 s were included, and patients below either threshold were excluded. Additional exclusions were pregnancy, cachexia, and systemic or ophthalmic corticosteroid or growth factor therapy within the preceding 3 months. To maintain treatment homogeneity, patients using oral antidiabetic agents other than the predefined regimens or receiving diuretic therapy were excluded. Concomitant antihypertensive and lipid-lowering therapies were permitted, reflecting routine clinical practice.

Clinical and Anthropometric Assessment

Clinical and anthropometric variables were retrospectively extracted from electronic medical records and had been obtained as part of routine outpatient care using standardized procedures. Body mass index (BMI), waist–hip ratio, neck circumference, and body fat percentage were recorded at baseline and at the 3-month follow-up when available. Waist circumference was measured midway between the lower rib margin and the iliac crest, and hip circumference at the widest point of the hips; waist–hip ratio was interpreted according to World Health Organization criteria (<0.90 for males and <0.85 for females).¹⁴ Neck circumference was measured inferior to the laryngeal prominence using a flexible tape measure, and body fat percentage was assessed by bioelectrical impedance analysis (Tanita MC-780MA, Tokyo, Japan).

Systolic blood pressure and diastolic blood pressure were measured using a manual sphygmomanometer (Erka Perfect Aneroid 201, Bad Tölz, Germany) according to standard clinical practice; hypertension was defined as systolic blood pressure ≥140 mmHg and/or diastolic blood pressure ≥90 mmHg.

Follow-Up Time Points and Data Structure

Three predefined follow-up time points were defined based on routinely documented outpatient visits and corresponding electronic medical records. Baseline (T0) included clinical, anthropometric, biochemical, and ophthalmological variables recorded at initiation of the predefined antidiabetic regimens. The 3-month follow-up (T1) comprised repeat clinical, anthropometric, biochemical, and AS-OCT assessments, consistent with routine metabolic and ocular monitoring.

At 12 months (T2), AS-OCT data were available for 17 patients in Group A and 16 in Group B; biochemical data were available for 17 in Group A and 15 in Group B. Individual reasons for missing 12-month observations could not be reliably reconstructed retrospectively; a sensitivity analysis comparing baseline characteristics of completers and non-completers (Supplementary Table S1) showed no significant differences, providing indirect support that missingness was not systematically related to baseline patient characteristics. Anthropometric measures were not routinely collected at T2; therefore, anthropometric variables were not analyzed at T2. Accordingly, T2 analyses primarily focused on ocular outcomes, with biochemical analyses conducted using available data (Figure 1).

Figure 1.

Flow chart showing patient selection, treatment groups, and data availability at baseline (T0), 3-month (T1), and 12-month (T2) follow-up. Missing values represent unavailable routine follow-up documentation; a sensitivity analysis (Supplementary Table S1) found no significant baseline differences between completers and non-completers. Anthropometric data were not routinely recorded at T2. Sample sizes for the ΔHbA1c–ΔCCT regression/correlation analyses and the baseline-adjusted ANCOVA analyses are also indicated, reflecting their differing data requirements.

Ophthalmological Examination

All ophthalmological examinations were performed as part of routine clinical follow-up and retrospectively retrieved from electronic medical records; no study-specific examinations were scheduled. Standard assessment included best-corrected visual acuity measurement using Snellen charts, slit-lamp biomicroscopy (Topcon SL-7F, Topcon Corporation, Tokyo, Japan) for anterior segment evaluation, and dilated fundus examination to assess posterior segment pathology. Fundus photography and fluorescein angiography were performed when clinically indicated as part of routine care.

Following the standard examination, AS-OCT imaging was performed to obtain quantitative measurements of CCT and CCET, which constituted the primary ocular outcome measures of the study.

AS-OCT Measurements

Anterior segment imaging was performed using the Revo FC system (software version 11.0.4; Optopol Technology, Zawiercie, Poland). The corneal imaging module generates axial, epithelial, and CCT maps using an 850-nm superluminescent diode, with a scanning speed of 80,000 A-scans/s and a scan depth of 2.4 mm in standard mode.¹⁵ Epithelial thickness maps were automatically generated over an 8-mm diameter by the built-in software.¹⁶

All AS-OCT scans were acquired according to a standardized protocol. Participants fixated on the internal target to ensure proper centration. Scan quality was assessed using the instrument’s built-in Quality Index (QI), and only scans with QI ≥ 6 were accepted. Scans below this threshold, or showing motion artifact or decentration, were excluded and repeated when necessary; when more than one acceptable acquisition was available, the acquisition with the highest QI was used for analysis. Multiple separate acquisitions were not averaged; any within-scan processing was performed automatically by the built-in software. All measurements were performed by a single experienced ophthalmologist. Both eyes were included if image quality criteria were met. To minimize diurnal variation, examinations were conducted between 09:00 and 12:00.

Rationale for Biochemical and Anthropometric Parameters

Biochemical and anthropometric parameters were evaluated to characterize the metabolic profile of patients with T2D and to provide a biological framework for interpreting corneal structural outcomes. Chronic hyperglycemia and cumulative glycemic exposure, reflected by fasting plasma glucose (FPG) and glycated hemoglobin (HbA1c), have been associated with corneal epithelial and stromal alterations through oxidative stress, low-grade inflammation, and non-enzymatic protein glycation, potentially affecting corneal thickness regulation and epithelial remodeling.^5,6 Anthropometric indices were included as surrogate markers of overall and regional adiposity linked to insulin resistance and systemic metabolic burden. In this context, ophthalmological findings may reflect broader metabolic alterations beyond localized ocular changes, supporting the role of routine eye examinations in monitoring diabetes-related systemic involvement.⁸ Accordingly, these parameters were incorporated to enable integrated interpretation of longitudinal changes in CCT and CCET during follow-up.

Laboratory Analysis

Blood and urine samples were obtained as part of routine clinical care after at least 10 h of overnight fasting. Venous blood was collected in 8.5 mL serum separator tubes (Becton Dickinson, Franklin Lakes, NJ, USA), and HbA1c samples in 2 mL EDTA tubes. All analyses were performed in the central hospital laboratory using standardized protocols. The laboratory is accredited and participates in regular internal and external quality control programs.

HbA1c was measured by boronate affinity chromatography (Quo-Lab®, EKF Diagnostics PLC, Cardiff, UK) and reported according to NGSP standards; corresponding IFCC units (mmol/mol) were provided where applicable. The same assay method and kit were used for HbA1c measurements throughout the follow-up period. FPG was measured using the hexokinase method on a Roche Cobas Integra® 400 Plus analyzer with the GLUC kit (Roche Diagnostics, Mannheim, Germany). Serum lipid parameters, including total cholesterol, triglycerides, high-density lipoprotein cholesterol, and low-density lipoprotein cholesterol, were analyzed by enzymatic colorimetric methods on a Roche Cobas Integra® 400 Plus analyzer using the CHOL2, TRIGL, HDLC4, and LDL-C3 kits (Roche Diagnostics, Mannheim, Germany). Using the same analyzer, urinary albumin was measured by an immunoturbidimetric method with the ALBT2 kit, and renal function was assessed by calculating the estimated glomerular filtration rate using the CKD-EPI equation.

Biochemical parameters were available at baseline, 3 months, and 12 months for most patients, whereas anthropometric measurements were not routinely recorded at the 12-month visit, reflecting routine clinical practice rather than study-related data omission.

Statistical Analysis

Statistical analyses were performed to evaluate longitudinal changes in corneal structural parameters and associated metabolic measures, and to compare these changes between treatment groups. Categorical variables are presented as counts and percentages, and continuous variables as mean ± standard deviation or median (interquartile range), as appropriate. Normality was assessed using the Shapiro–Wilk test.

Baseline comparisons between independent groups were conducted using Student’s t-test or the Mann–Whitney U-test for continuous variables and the chi-square or Fisher’s exact test for categorical variables. Within-group changes (baseline to 3 months and baseline to 12 months) were analyzed using paired t-tests or Wilcoxon signed-rank tests, and between-group differences in longitudinal change were evaluated by comparing absolute change values (Δ = follow-up − baseline) between groups using Student’s t-test or the Mann–Whitney U-test, as appropriate to the distribution of the change scores.

For AS-OCT outcomes, the right eye was selected a priori as the primary unit of analysis to provide one independent ocular observation per participant and thereby avoid the need to model within-participant inter-eye correlation in this modestly sized cohort; the corresponding left-eye measurements were analyzed separately as a planned sensitivity analysis rather than introducing a more complex bilateral multilevel model. Right- and left-eye AS-OCT data were available for the same participants at each time point (T0: Group A n=19, Group B n=21; T1: Group A n=19, Group B n=21; T2: Group A n=17, Group B n=16); no participant had an eligible scan for only one eye. Multivariable linear regression models were used to examine independent associations between metabolic parameters and corneal outcomes, adjusting for age and sex. These analyses were restricted to time points with complete covariate data and are summarized in Supplementary Table S2. As sensitivity analyses, linear mixed-effects models with participant-specific random intercepts were fitted for CCT and CCET, including time, treatment group, and time-by-treatment interaction as fixed effects. In addition, ANCOVA models evaluated 12-month CCT and CCET between groups after adjustment for their respective baseline values.

Missing data were handled using an available-case approach without imputation, reflecting incomplete routine follow-up documentation; a sensitivity analysis comparing baseline characteristics of completers and non-completers (Supplementary Table S1) supported the assumption that missingness was not systematically related to baseline patient characteristics (Figure 1). To control for multiple testing, Bonferroni correction was applied within predefined families of related comparisons. Specifically, corrections were applied to baseline group comparisons, three-month change analyses (ΔT1–T0), twelve-month ocular outcomes, and twelve-month biochemical outcomes. The adjusted significance threshold was defined as α = 0.05/m, where m represents the number of hypotheses tested simultaneously within each predefined comparison family. For transparency, the corresponding adjusted alpha levels are summarized in Supplementary Table S3. The focused exploratory correlations between changes in HbA1c and corneal parameters are presented with unadjusted p values in Supplementary Table S4, whereas the broader exploratory correlation analysis incorporating both HbA1c and fasting plasma glucose was adjusted for multiple testing using the Benjamini–Hochberg false discovery rate (FDR) procedure (Supplementary Table S5), reflecting the hypothesis-generating rather than confirmatory nature of these analyses. Bonferroni correction was applied to the repeated between-group comparisons, including the corresponding extended supplementary analyses (Supplementary Tables S6, S7, and S10).

The analysis addressed two co-primary objectives: first, evaluation of longitudinal changes in CCET and CCT during follow-up; and second, comparison of 12-month changes in CCET and CCT between treatment groups. Short-term metabolic responses and regression-based associations between glycemic and corneal parameters were considered secondary and exploratory analyses.

Exploratory analyses evaluated associations between changes in glycemic parameters and changes in corneal parameters using Spearman rank correlations. Focused ΔHbA1c analyses with unadjusted p values are reported in Supplementary Table S4, whereas the broader FDR-adjusted analysis incorporating both ΔHbA1c and ΔFPG is reported in Supplementary Table S5.

A two-sided p value < 0.05 (or adjusted threshold after correction, where applicable) was considered statistically significant.

To contextualize the interpretability of between-group comparisons, a post hoc precision analysis was performed for the primary outcomes. Based on the available sample size, the study allowed detection of between-group differences of approximately 4.7 µm for CCT and 6.1 µm for CCET under conventional assumptions (two-sided α = 0.05, 80% power), based on the pooled standard deviation of the 12-month change scores (ΔT2–T0) in each group. This analysis was performed to support interpretation of the findings rather than to determine sample size.

All analyses were performed using SPSS software version 22.0 (IBM Corp., Armonk, NY, USA).

Ethics Approval

This retrospective, non-interventional study was approved by the Istanbul Medipol University Non-Interventional Clinical Research Ethics Committee (Approval No: E-10840098-772.02-1718; date: March 7, 2023) and conducted in accordance with the Declaration of Helsinki. The requirement for informed consent was waived due to the retrospective design and use of anonymized routine clinical data.

Results

Study Population and Follow-Up

A total of 40 patients with T2D were included at baseline (Group A, n = 19; Group B, n = 21). Complete clinical, biochemical, anthropometric, and ophthalmological data were available at baseline (T0) and at the 3-month follow-up (T1).

At the 12-month follow-up (T2), AS-OCT data were available for 17 patients in Group A and 16 in Group B; anthropometric data were not available at T2 due to routine clinical practice. Participant flow and data availability across follow-up time points are summarized in Figure 1.

Baseline Clinical and Metabolic Characteristics

Baseline demographic, clinical, anthropometric, biochemical, and ocular characteristics are summarized in Table 1 and Supplementary Table S6. Groups were comparable for age, sex, anthropometric indices, renal function, lipid profile, and glycemic parameters. Baseline CCET was similar between groups, whereas CCT was modestly higher in Group A than in Group B.

Table 1.

Baseline Demographic, Anthropometric, Metabolic and Ophthalmological Characteristics of the Study Population

Variable	Group A (n = 19)	Group B (n = 21)	p-value
Age (years)	53.42 ± 7.16	56.52 ± 9.23	0.246
Sex (male/female) (n)	11/8	12/9	0.884
Body mass index (kg/m2)	29.32 ± 5.82	31.59 ± 6.09	0.238
Waist–hip ratio	0.95 ± 0.08	1.00 ± 0.18	0.338
Neck circumference (cm)	42.05 ± 5.65	41.57 ± 4.20	0.764
Body fat percentage (%)	25.85 ± 9.50	30.21 ± 8.37	0.134
Fasting plasma glucose (mg/dL)	200.47 ± 69.21	205.76 ± 65.87	0.806
HbA1c (%) (mmol/mol)	9.70 ± 2.42 (83 ± 26)	9.75 ± 2.13 (84 ± 23)	0.944
Fasting insulin (µIU/mL)	19.76 ± 15.38	16.82 ± 10.59	0.495
Fasting C-peptide (ng/mL)	3.32 ± 1.40	3.59 ± 1.05	0.513
HOMA-IR	8.90 ± 6.19	8.97 ± 8.12	0.974
Triglycerides (mg/dL)	123 (95.5–218.5)	184 (137–272)	0.155
eGFR (mL/min/1.73 m2)	101.84 ± 10.41	92.15 ± 12.54	0.011
Urinary albumin (mg/L)	7.8 (3.8–16.3)	7.2 (5.7–16.6)	0.787
Central corneal thickness (µm)	550.05 ± 28.19	528.67 ± 31.54	0.030
Central corneal epithelial thickness (µm)	60.21 ± 6.46	62.14 ± 5.29	0.305

Notes: Values are presented as mean ± standard deviation or median (interquartile range), as appropriate. Baseline comparisons between groups were performed using independent samples t-tests or Mann–Whitney U-tests according to data distribution. Categorical variables were compared using the chi-square test. Right-eye measurements were used for primary ophthalmological analyses. Raw two-sided p-values are presented; multiplicity-adjusted significance thresholds are detailed in Supplementary Table S3.

Abbreviations: HbA1c, glycated haemoglobin; HOMA-IR, homeostasis model assessment of insulin resistance; eGFR, estimated glomerular filtration rate. HbA1c values are presented in both NGSP (%) and IFCC (mmol/mol) units. Available sample sizes were lower for fasting insulin and HOMA-IR in Group B (n=20); all other variables used the full baseline sample.

Three-Month Changes in Clinical and Biochemical Parameters

Glycemic, anthropometric, and corneal parameter changes from baseline to the 3-month follow-up (T1) are summarized in Table 2. Both treatment groups demonstrated significant within-group improvements in glycemic control (paired t-test, both p<0.001), with comparable reductions in HbA1c between groups (independent-samples t-test, p=0.977). Anthropometric measures showed modest changes without significant between-group differences; extended clinical parameters, including blood pressure, are presented in Supplementary Table S7. Despite modest baseline differences in CCT, no significant between-group differences were observed in changes in CCT or CCET from baseline to T1; CCET showed a slight numerical decrease in the empagliflozin–metformin group and a modest increase in the vildagliptin–metformin group.

Table 2.

Baseline (T0) and 3-Month (T1) Values and Within-Group Changes

Variable	Group A (n = 19) Baseline (T0)	Group A 3-Month (T1)	Within-group p (Group A)	Group B (n = 21) Baseline (T0)	Group B 3-Month (T1)	Within-group p (Group B)	Between-group p (ΔT1–T0)
HbA1c (%)	9.70 ± 2.42	6.71 ± 0.82	<0.001	9.75 ± 2.13	6.74 ± 0.79	<0.001	0.977
Fasting plasma glucose (mg/dL)	200.47 ± 69.21	136.05 ± 28.47	<0.001	205.76 ± 65.87	128.43 ± 21.39	<0.001	0.548
Body mass index (kg/m2)	29.32 ± 5.82	28.68 ± 5.45	0.034	31.59 ± 6.09	30.20 ± 5.06	<0.001	0.081
Central corneal thickness (µm)	550.05 ± 28.19	550.26 ± 26.92	0.869	528.67 ± 31.54	529.38 ± 34.71	0.588	0.782
Central corneal epithelial thickness (µm)	60.21 ± 6.46	61.42 ± 6.75	0.349	62.14 ± 5.29	61.48 ± 7.01	0.479	0.237

Notes: Values are presented as mean ± standard deviation. Within-group comparisons (T0 vs T1) were performed using paired-samples tests and are presented for descriptive context. Between-group comparisons of the four original change-score outcomes (HbA1c, body mass index, CCT, CCET) correspond to the Bonferroni-adjusted family of four comparisons (α = 0.0125; Supplementary Table S3). Fasting plasma glucose comparisons were added for descriptive completeness in response to the Reviewer and were not included in this multiplicity-adjusted family.

Twelve-Month Corneal Outcomes

Twelve-month (T2) corneal outcomes are summarized in Table 3. No statistically significant between-group differences were observed in CCT or CCET at T2; although CCT values were numerically higher in the vildagliptin–metformin group.

Table 3.

Twelve-Month (T2) Ophthalmological Outcomes and 12-Month Within-Group Changes

Variable	Group A (n = 17) T2 value	Group B (n = 16) T2 value	Between-Group p (T2)	Group A ΔT2–T0	Group B ΔT2–T0	Between-Group p (ΔT2–T0)
Central corneal thickness (µm)	550.00 ± 27.40	527.69 ± 36.46	0.058	1.41 ± 5.12	2.75 ± 4.37	0.425
Central corneal epithelial thickness (µm)	61.53 ± 5.93	61.25 ± 8.60	0.915	1.24 ± 6.88	−0.88 ± 5.54	0.338

Notes: Values are presented as mean ± standard deviation. Analyses were restricted to patients with available anterior segment optical coherence tomography measurements at the 12-month follow-up. The between-group comparison of T2 values was performed using independent samples t-tests and corresponds to the Bonferroni-adjusted family of two comparisons (α = 0.0250; Supplementary Table S3). The ΔT2–T0 between-group comparisons are presented for descriptive context; corresponding baseline-adjusted ANCOVA results are presented in Supplementary Table S9. Right-eye measurements were used for primary analyses.

In exploratory analyses, changes in HbA1c and fasting plasma glucose (FPG) were each positively associated with changes in CCT, but not CCET, over 12 months (Supplementary Tables S4and S5). In age- and sex-adjusted linear regression analyses, 12-month changes in HbA1c were independently associated with changes in CCT (β = 0.81 µm per 1% HbA1c change; 95% CI 0.20–1.42; p = 0.011; n = 32), while baseline age and sex were not independently associated with ΔCCT (Supplementary Table S2). Left-eye sensitivity analyses, including visit-specific between-group comparisons, longitudinal within- and between-group changes, and the exploratory ΔHbA1c–ΔCCT association, were directionally and statistically concordant with the corresponding primary right-eye findings (Supplementary Table S8). Sensitivity analyses using linear mixed-effects models showed no significant time-by-treatment interactions for CCT or CCET, and ANCOVA showed no significant between-group differences in 12-month CCT or CCET after adjustment for baseline values (Supplementary Table S9).

Twelve-Month Biochemical Outcomes

Twelve-month biochemical outcomes are summarized in Table 4. At T2, no statistically significant between-group differences were observed in absolute glycemic parameters, lipid profile, renal function indices, or urinary albumin among patients with available data. Within-group changes from baseline to T2 (Supplementary Table S10) showed sustained glycemic improvement in both groups, with comparable reductions in HbA1c and FPG. Changes in renal parameters were modest and largely similar between groups. After correction for multiple testing, none of the between-group biochemical differences remained statistically significant (Supplementary Table S3).

Table 4.

Twelve-Month (T2) Biochemical Outcomes

Variable	Group A (n = 17)	Group B (n = 15)	p-value
HbA1c (%) (mmol/mol)	7.58 ± 1.80 (60 ± 20)	7.36 ± 1.47 (57 ± 16)	0.708
FPG (mg/dL)	157.41 ± 48.42	138.07 ± 29.09	0.177
T-col (mg/dL)	176.12 ± 43.31	186.33 ± 52.11	0.554
TG (mg/dL)	136.0 (90.0–257.0)	152.0 (123.5–238.0)	0.734
HDL-c (mg/dL)	44.00 ± 13.72	44.87 ± 9.90	0.838
LDL-c (mg/dL)	91.12 ± 36.81	105.07 ± 39.25	0.310
eGFR (mL/min/1.73m2)	106.47 ± 14.64	104.36 ± 8.68	0.619
Creatinine (mg/dL)	0.74 ± 0.10	0.72 ± 0.13	0.599
Urinary albumin (mg/L)	5.94 (3.97–9.05)	4.91 (3.44–7.80)	0.623

Notes: Values are presented as mean ± standard deviation, except triglycerides and urinary albumin, which are presented as median (interquartile range). Analyses were restricted to patients with available biochemical data at the 12-month follow-up. Between-group comparisons were performed using independent samples t-tests (Welch’s correction where appropriate) or the Mann–Whitney U-test for non-normally distributed variables. A p-value < 0.05 was considered statistically significant. Multiple comparison correction (Bonferroni) is detailed in Supplementary Table S3.

Abbreviations: HbA1c, glycated hemoglobin; FPG, fasting plasma glucose; T-col, total cholesterol; TG, triglycerides; HDL-c, high-density lipoprotein cholesterol; LDL-c, low-density lipoprotein cholesterol; eGFR, estimated glomerular filtration rate. HbA1c values are presented in both NGSP (%) and IFCC (mmol/mol) units.

Missing Data and Sensitivity Analyses

Missing data were mainly confined to the 12-month follow-up. Individual reasons for missing 12-month observations could not be reliably reconstructed retrospectively. Baseline characteristics did not differ significantly between participants with and without 12-month ophthalmological data (Supplementary Table S1). Anthropometric variables were unavailable at T2 and were therefore excluded from long-term analyses by design. Sensitivity analyses restricted to patients with complete ophthalmological follow-up yielded results consistent with the primary analyses, supporting the overall consistency of the findings.

Discussion

Principal Findings

In this retrospective cohort study, we evaluated longitudinal corneal thickness trajectories over 12 months in adults with T2D. Our findings provide real-world evidence on corneal structural dynamics assessed by AS-OCT during antidiabetic therapy. Both treatment regimens achieved significant improvement in glycemic control, consistent with their established efficacy.^9–12 Corneal structural parameters (CCT and CCET) showed no statistically detectable longitudinal change throughout follow-up.

Given that chronic hyperglycemia has been associated with corneal epithelial and stromal alterations,^5,8 the absence of a statistically detectable increase in CCT or CCET during follow-up is consistent with, but does not by itself establish, structural stability; failure to detect change in a modestly sized cohort does not exclude a true underlying effect below our precision threshold. Although CCET showed a numerically greater reduction in the empagliflozin–metformin group, no statistically significant between-group difference was observed. The association observed between HbA1c change and CCT was not paralleled by a corresponding association with CCET; however, a significant association for one outcome alongside a non-significant association for another does not, by itself, establish that the two associations differ from one another, and we have therefore refrained from inferring differential epithelial vs stromal sensitivity to glycemic change from this pattern. The HbA1c–CCT association itself (β = 0.81 µm per 1% HbA1c; n = 32) is modest in magnitude and should be interpreted cautiously given its exploratory nature, the modest sample size, potential residual confounding related to the non-randomized treatment allocation, and the number of comparisons performed; it should therefore be regarded as hypothesis-generating rather than confirmatory.

In further analyses, age- and sex-adjusted regression models showed that changes in HbA1c were independently associated with changes in CCT, whereas no proportional relationship was observed for CCET. This finding is consistent with evidence that corneal structural characteristics in T2D are associated with metabolic status and cumulative glycemic exposure.^5,8 Accordingly, the absence of a statistically detectable change during follow-up should be interpreted within the precision limits of the present study, while treatment-related remodeling effects, if present, may be limited within the present observation period.

To further contextualize the magnitude and precision of these estimates, the 12-month between-group differences in change from baseline (ΔT2–T0) were small (CCT: 1.34 µm (95% CI −2.04 to 4.72); CCET: −2.11 µm (95% CI −6.54 to 2.31)). Under conventional assumptions (two-sided α = 0.05, 80% power), the available sample size allowed detection of between-group differences of approximately 4.7 µm for CCT and 6.1 µm for CCET. The observed estimates are modest relative to typical clinical variation in corneal thickness measurements,¹⁷ providing a precision-based complement to conventional hypothesis testing (Supplementary Table S11).

Comparison with Previous Studies

Prior studies evaluating corneal changes in diabetes have predominantly used cross-sectional or single-treatment designs, reporting generalized corneal thickening or epithelial alterations without differentiating treatment-specific or layer-specific responses.^18–21 These findings extend this literature through a treatment-comparative, layer-specific longitudinal analysis. Methodological characteristics and findings of selected previous studies relevant to the present results are summarized in Supplementary Table S12.

Our findings align with studies reporting increased corneal thickness in T2D.^17,18

In contrast, CCT alterations appear to be more closely related to longer diabetes duration and poorer glycemic control than to short-term metabolic variation.^19,20 These findings suggest that stromal thickness may reflect cumulative glycemic exposure rather than rapidly reversible metabolic changes.

Clinical Implications

The corneal epithelium serves as the primary barrier of the ocular surface and contributes to corneal transparency, wound healing, and defense against infection. Diabetes-related epithelial alterations increase susceptibility to ocular surface complications.^5,8

Given the distinct microvascular properties of these drug classes,^9,11,12 the absence of marked between-group differences should be interpreted cautiously and does not provide evidence for differential treatment-related effects on corneal structure in this setting.

Direct microvascular or ocular perfusion parameters were not assessed. Therefore, any interpretation regarding microvascular mechanisms should be considered hypothesis-generating rather than evidence-based. Future prospective studies incorporating direct assessments of endothelial function, anterior segment perfusion imaging, and systemic vascular biomarkers are needed to clarify potential mechanistic pathways.

From a clinical perspective, incorporation of anterior segment assessment into routine diabetes follow-up has been increasingly emphasized within integrated care models,^2,8 and structured collaboration between metabolic care providers and eye-care specialists may support individualized risk stratification and longitudinal monitoring of ocular complications. Within this framework, AS-OCT–derived parameters may serve as adjunctive structural markers for integrated follow-up. The present study did not evaluate the prognostic value, diagnostic accuracy, effect on clinical management, or associations with subsequent diabetic complications of AS-OCT–derived corneal parameters; these applications remain to be established. Future longitudinal studies are needed to determine whether early corneal epithelial changes have prognostic value for broader microvascular outcomes in T2D.

Limitations and Strengths

Several limitations of this study should be acknowledged. First, the retrospective, single-center design limits causal inference and generalizability, and may be subject to residual confounding despite careful patient selection and standardized assessment procedures. The exact number of patients screened and excluded prior to the final cohort could not be reliably reconstructed from the retrospective records, precluding a formal patient-screening flow diagram. Because imaging data were retrospectively obtained from routine clinical records, a prospectively assessed between-scan repeatability coefficient for AS-OCT measurements was not available. Second, the sample size was modest, particularly at the 12-month follow-up, which may have limited statistical power to detect small between-group differences. Diabetes duration, a recognized determinant of corneal structural alterations in cross-sectional studies, could not be reliably ascertained for all participants and was therefore not included in the analyses; age at diagnosis was similarly unavailable, precluding its use as a surrogate. Diabetic retinopathy severity and other ocular-surface factors (eg, tear film status beyond the exclusion of clinically significant dry eye) were not systematically graded and could not be entered as covariates. As treatment allocation was based on clinical judgment rather than randomization, confounding by indication cannot be excluded, and the exploratory HbA1c–CCT and FPG–CCT associations, which remained significant in the FDR-adjusted broader correlation analysis (Supplementary Table S5), remain susceptible to residual confounding and to inflated type I error given the number of comparisons performed. Other potentially relevant ocular and diabetes-related factors, including diabetic neuropathy, retinopathy stage, endothelial cell count, corneal sensitivity, and ocular surface disease severity, were not systematically available in the retrospective records and could not be incorporated into the analyses. Medication adherence could not be systematically assessed from the retrospective records, which may have introduced additional variability in treatment exposure. In addition, the absence of a healthy control group and corneal biomechanical assessment should be considered when interpreting the corneal structural findings.

However, the longitudinal within-subject design partially mitigates this concern, as diabetes duration more strongly affects cross-sectional than within-subject corneal measurements (see Strengths, below). The absence of detectable change over this interval remains a precision-limited finding rather than confirmed structural stability.

Missing data were largely confined to the 12-month follow-up. Individual reasons for missing 12-month observations could not be reliably reconstructed retrospectively; therefore, outcome-dependent loss cannot be excluded. However, baseline characteristics did not differ significantly between participants with and without 12-month ophthalmological data (Supplementary Table S1), providing indirect reassurance against systematic baseline differences between completers and non-completers. Anthropometric measurements were not routinely collected at the 12-month visit and were therefore unavailable by design, whereas biochemical data at T2 were available only for a subset of patients, consistent with real-world follow-up patterns. Nevertheless, incomplete long-term data may have introduced selection or attrition bias, particularly for anthropometric and biochemical parameters. This may have limited the ability to fully characterize associations between metabolic changes and corneal outcomes at T2.

Accordingly, longer-term analyses should be interpreted with caution. Future prospective studies with predefined and standardized follow-up schedules, including systematic collection of anthropometric, metabolic, and ophthalmological data at all time points, would help minimize missing data and provide a more comprehensive assessment of longitudinal effects.

Despite these limitations, this study has several strengths. All ophthalmological assessments were performed using standardized AS-OCT protocols with consistent image quality criteria, and right-eye measurements were used for primary analyses. The longitudinal within-subject design, in which each participant served as their own control, partially mitigated inter-individual variability in baseline corneal characteristics. The availability of both short- and longer-term follow-up allowed evaluation of temporal corneal changes during antidiabetic therapy in a real-world clinical setting, supporting the external validity and clinical relevance of the findings.

Although modest, the sample size yielded relatively narrow confidence intervals around the between-group estimates, and findings were consistent across change-score, contralateral-eye, and precision-based analyses, supporting the robustness of the interpretation.

Conclusion

In this retrospective cohort study, empagliflozin–metformin and vildagliptin–metformin achieved comparable and sustained glycemic improvement over 12 months.

Corneal epithelial and central thickness parameters showed no statistically detectable longitudinal changes during follow-up and no significant between-group differences, within the limits of the study’s sample size, observational design, and statistical precision. The results should therefore be interpreted as hypothesis-generating. Adequately powered prospective studies with longer follow-up are needed to determine whether corneal structural changes emerge over longer periods, differ across glucose-lowering therapies, or are associated with the development or progression of diabetic microvascular complications.

Funding Statement

The authors received no specific funding for this work.

Data Sharing Statement

The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Ethics Approval and Consent to Participate

This retrospective, non-interventional study was approved by the Istanbul Medipol University Non-Interventional Clinical Research Ethics Committee (Approval No: E-10840098-772.02-1718; March 7, 2023) and conducted in accordance with the Declaration of Helsinki. The requirement for informed consent was waived due to the retrospective nature of the study and the use of anonymized clinical data.

Author Contributions

İhsan Boyacı conceived and designed the study, collected and analyzed the data, interpreted the results, and drafted the manuscript. Göktuğ Demirci performed the ophthalmological assessments, contributed to data interpretation, and critically revised the manuscript. All authors made a significant contribution to the work reported, whether that is in the conception, study design, execution, acquisition of data, analysis and interpretation, or in all these areas; took part in drafting, revising, or critically reviewing the article; gave final approval of the version to be published; have agreed on the journal to which the article has been submitted; and agree to be accountable for all aspects of the work.

Disclosure

The authors declare that they have no competing interests.