Abstract
Diabetic corneal neuropathy (DCN) and diabetic retinopathy (DR) are microvascular complications and share common pathophysiological mechanisms. However, the relationship between them remains poorly defined. In this cross-sectional study, we aimed to investigate the association among DCN, DR, and tear mediators in 1,654 eyes from 822 participants, comprising 634 patients with type 2 diabetes and 188 healthy participants. Our data demonstrated that compared with control participants, all patients with diabetes had significantly impaired corneal nerve metrics, increased dendritic cell length and density, and larger corneal microneuromas, even in the absence of DR. Patients with nonproliferative DR (NPDR) and proliferative DR (PDR) showed significantly reduced corneal nerve parameters compared with those with no DR. Furthermore, patients with PDR presented significantly worse ocular surface clinical manifestations than patients with no DR, patients with NPDR, and control participants. Cumulative link mixed models demonstrated that corneal sensitivity and corneal nerve parameters were significantly associated with the severity of DR. Tear substance P concentrations were significantly lower across all stages of DR compared with control participants. Tear MMP-9, substance P, and IGFBP-3 levels were significantly associated with corneal nerve and ocular surface parameters. This study demonstrates that DCN precedes the onset of DR and worsens with the severity of DR. Corneal nerve status could be an early indicator and predictor of DR.
Article Highlights
Both diabetic corneal neuropathy (DCN) and diabetic retinopathy (DR) are microvascular complications and share common pathophysiological mechanisms. The links between them are poorly understood.
We investigated the relationship between DCN and DR by studying imaging features, clinical manifestations, and tear biomarkers.
Our study demonstrated that DCN precedes the onset of DR, and DCN worsens with the severity of DR stage. Corneal sensitivity and corneal nerve parameters are significantly associated with DR severity. Tear MMP-9, substance P, and IGFBP-3 were significantly altered in DR.
Corneal nerve status could be an early indicator and predictor of DR.
Introduction
Diabetes is a chronic metabolic disease characterized by sustained and uncontrolled hyperglycemia, leading to macrovascular and microvascular damage. Common diabetic microvascular complications include diabetic retinopathy (DR), diabetic nephropathy, and diabetic neuropathy. Diabetic neuropathy impairs nerves throughout the body, including corneal nerves, resulting in a condition known as diabetic corneal neuropathy (DCN) (1). The prevalence of DCN is 47–64% in diabetes (2). DCN can result in progressive neurotrophic keratopathy, persistent epithelial defects, corneal ulceration, scarring, and potential vision loss. DR is one of the leading causes of blindness, and the global prevalence of DR is ∼22.7% among populations with diabetes (3). Taken together, ocular surface complications secondary to diabetic keratopathy and vision loss resulting from DR have imposed substantial medical and economic burdens.
It has been established that neuronal and microangiographic alterations in microvascular dysfunction are interlinked in diabetes (4). More importantly, studies have demonstrated that retinal neurodegeneration occurs prior to the onset of DR (5–7). Similarly, neural changes in diabetic peripheral neuropathy are detectable before vascular changes are identifiable, suggesting that small nerve fiber alterations, such as DCN, may be the earliest detectable signs of diabetes (8). In vivo confocal microscopy (IVCM) allows for direct visualization of corneal nerves, and the assessment of corneal nerves has been considered a reliable and sensitive surrogate marker for diabetic peripheral neuropathy and neurodegenerative diseases in U.S. Food and Drug Administration clinical trials (1).
Several studies have investigated the corneal nerve alterations in patients with various severities of DR, yielding inconsistent findings. Corneal nerve fiber loss has been observed in patients with DR compared with healthy control participants (9), with greater reductions in corneal nerve fiber length (CNFL) and corneal nerve fiber density (CNFD) corresponding to more advanced DR stages (10), suggesting the potential relationship between DCN and DR. Conversely, some studies found no significant differences in corneal nerve parameters or corneal sensitivity between patients with and without DR (11). Additionally, investigations in both type 1 and type 2 diabetes revealed no correlation between corneal nerve fiber morphology and retinal nerve fiber layer thickness or vascular metrics, demonstrating independent neurodegenerative changes in the cornea and retina during early diabetes (12,13). In addition to corneal nerve alterations, several animal and clinical studies showed an increased density of corneal dendritic cells (DCs), markers of inflammatory reaction, in individuals with diabetes, while a few observed a decrease (14). Furthermore, corneal microneuromas, indicative of corneal nerve injury and an unsuccessful regenerative attempt of damaged axons (15), have been reported in a single study to be increased in diabetes and correlated with poorer glucose control (16). However, studies on corneal microneuromas in diabetes are still very limited.
Analysis of tear profiles offers valuable insights into the underlying pathophysiology of DCN and DR. Understanding these mechanisms would enhance our knowledge of diabetic microvascular complications and potentially contribute to the development of targeted therapeutic strategies. Furthermore, acquiring tear samples is noninvasive and more accessible compared with the collection of vitreous or aqueous fluid. Compared with healthy participants, patients with DR have higher concentrations of interleukins, tumor necrosis factor, and matrix metalloproteinases (MMPs) in tears (17). Nevertheless, this study did not investigate the morphology of the corneal nerve. Markoulli et al. (18) reported a significant correlation between corneal nerve parameters, such as corneal nerve fiber width (CNFW) and corneal nerve fractal dimension (CFracDim), and the tear concentrations of tissue inhibitor of metalloproteinase 1 (TIMP-1) and substance P (SP). However, the authors did not evaluate the severity of DR in the study populations.
The relationship between DCN and the severity of DR, as well as potential tear biomarkers, has not been well established in a large population. In the current study, we aimed to comprehensively investigate the interlink between DCN and DR by evaluating corneal imaging features, tear profiles, and clinical manifestations in patients with type 2 diabetes. Understanding the relationship between these two diabetic microvascular complications may enable earlier detection of DR using DCN. This could facilitate timely interventions to prevent or slow disease progression, thereby reducing the disease burden.
Research Design and Methods
Study Design and Participants
This cross-sectional study included 1,654 eyes of 822 participants. Of the participants, 634 were diagnosed with type 2 diabetes based on the following criteria: a physician’s diagnosis of type 2 diabetes, random blood glucose ≥11.1 mmol/L and/or HbA1c ≥6.5%, or patient-reported use of diabetic medications (19,20). Exclusion criteria (Supplementary Table 1) were corneal neuropathies due to a nondiabetic cause, presence of active ocular surface disorders or corneal pathology, presence of other corneal and retinal diseases not associated with diabetes, history of ocular surgeries, history of laser photocoagulation and intravitreal injections of anti–vascular endothelial growth factor or corticosteroid agents in the previous 3 months, concurrent use of topical or systemic immunosuppressants or steroids that might affect the DC analysis, and history of contact lens use of >1 year. Fundus examinations were performed by retinal specialists, and the stage of DR was determined based on the Early Treatment Diabetic Retinopathy Study (ETDRS) classification. Each eye was assigned to one of three groups: no DR, nonproliferative DR (NPDR), and proliferative DR (PDR) (21). A total of 188 age-matched healthy control participants with no history of diabetes, systemic inflammatory diseases, ocular diseases, or prior ocular surgeries were recruited. The study was approved by the Institutional Review Board of SingHealth (no. 2022/2046), Singapore, and the study was conducted in accordance with the Declaration of Helsinki. Informed consent was obtained from all participants.
IVCM Scans and Image Analysis for Corneal Nerves, DCs, and Microneuromas
IVCM (Heidelberg Retina Tomography III with the Rostock Cornea Module; Heidelberg Engineering GmbH) was performed to examine corneal epithelial cells, subbasal nerve plexus, DCs, and corneal microneuromas using the standardized protocols published previously (22,23). In brief, the scanning depth ranged from the superficial corneal epithelium to the anterior stroma. The subbasal nerve plexus of the central cornea and four quadrants of peripheral areas located 3 mm away from the corneal apex were scanned. From each area of each eye, the five best-focused and nonoverlapping images were selected (a total of 25 images per eye). All scans and image selections were performed by an independent and masked ophthalmologist (C.L.).
Each microimage of subbasal layer was then analyzed using ACCMetrics image analysis software (University of Manchester) for the following nerve parameters (24): CNFD (the number of fibers per mm2, each frame area = 0.16033 mm2), corneal nerve branch density (CNBD) (the number of branch points on the main fibers per mm2), CNFL (the total length of fiber in mm per mm2), corneal total branch density (CTBD) (the total number of branch points per mm2), corneal nerve fiber area (CNFA) (the total nerve fiber area in mm2 per mm2), CNFW (the average nerve fiber width in mm per mm2), and CFracDim (Supplementary Fig. 1). The mean value obtained from the analysis of the 25 selected images was used for further statistical analysis.
DCs were defined as small, white cells either isolated from the nerve or connected to the nerve branch. Five to 10 representative images with DCs were selected for each eye. The AIconfocal Rapid Image Evaluation System (ADCIS, Saint-Contest, France) was used for the analysis of DCs, and the following parameters were obtained: cell count (the number of cells per image), cell density (the cell count divided by the sum of the detected area [cells/μm2]), average length (the mean width of DCs [μm]), average area (μm2), and elongation (the absolute value of the difference between the major and minor axis divided by the sum of the major and minor axis) (25). The mean value derived from the analysis of all the selected images was used for subsequent statistical analysis.
Corneal microneuromas are microscopic, irregularly shaped enlargements of terminal subbasal nerve endings that form at sites of nerve injury or damage (15). Images showing microneuromas were selected and quantified manually using ImageJ software (National Institutes of Health, Bethesda, MD). The parameters measured included microneuroma area (μm2), average length (µm), and perimeter (μm) (26). A single, experienced, and masked investigator (I.X.Y.L.) conducted the image analysis.
Ocular Surface Assessments for DCN
Ocular surface assessments encompassed the Schirmer I test (without anesthesia), with results recorded as the length of strip wetting (mm) over 5 min (27). Tear breakup time (TBUT) was recorded as the interval (seconds) between the last blink and the appearance of the first dry spot (28). Corneal sensitivity was evaluated by a Cochet-Bonnet esthesiometer (Luneau Ophthalmologia, Chartres, France) at five locations: the central cornea and each of the four quadrants. The length (0–6 cm) of filament that elicited a blink reflex or escape response was recorded for each location. The total corneal sensitivity score ranged from 0 to 30 cm (29,30). Three measurements in the same visit were taken for the abovementioned tests, and the average value of each eye was used for analysis. Ocular surface fluorescein staining (Oxford score: 0 = absent, 5 = severe) and corneal fluorescein staining (National Eye Institute [NEI] scale: 0 = minimal, 15 = maximal) were used to assess the ocular surface integrity (31), with higher scores indicating worse ocular surface integrity. All assessments were conducted by an independent and masked ophthalmologist (C.L.). The subjective symptoms were assessed using the Ocular Surface Disease Index (OSDI), a 12-item questionnaire that assesses the severity of ocular surface symptoms and their impact on vision-related functioning. The total OSDI score was 0–100 following the calculation with the following equation: total score = (sum of scores for all questions answered × 100)/(total number of questions answered × 4) (32).
Tear Cytokine and Mediator Analysis
Tear samples were taken from the wet Schirmer strips as described previously (33). Tear strips were stored at −80°C until analysis. On the day of analysis, the strips were cut into small pieces and submerged into 200 µL ice-cold tear elution buffer containing 0.55 mol/L NaCl (S3014-500G; Sigma Life Sciences), 0.33% Tween-20 (P1379-500ML; Sigma Life Sciences), 0.55% BSA (9998S; Cell Signaling Technology), and 1× protease inhibitor (78425; Thermo Fisher Scientific) and subjected to sonication and homogenization at 20% amplitude for 20 s, followed by 17 h of incubation at 4°C with gentle agitation at 450 rpm (22). The homogenization step was repeated once the next day before centrifugation at 11,000 rpm at 4°C for 20 min. Each eluted tear protein sample contained clear supernatants and was subjected to analysis using ELISA for the following analytes: SP (6× dilution), insulin-like growth factor binding protein 3 (IGFBP-3) (6× dilution), MMP-9 (10× dilution), and TIMP-1 (15× dilution) (R&D Systems, Minneapolis, MN). The optical density measurement was read at 450 nm using an Infinite M200 (Tecan, Männedorf, Switzerland), and an optical density reading at 540 nm was set as a reference (22). The concentrations of each analyte were interpolated from the standard curve using GraphPad Prism 8 software (GraphPad Software, San Diego, CA).
Statistical Analysis
The normality of data distribution was evaluated using the Shapiro-Wilk test. Normally distributed data are expressed as mean ± SD, while nonnormally distributed data are presented as median (interquartile range). The mean values of both eyes were used for the comparisons across four different groups. For patients with asymmetric DR severity between eyes, DR classification was based on the eye with the more advanced stage. ANOVA and Bonferroni post hoc tests were applied to normally distributed variables, and Kruskal-Wallis test and Dunn multiple comparisons test were used for nonnormally distributed continuous variables. Multivariate linear mixed models (LMMs) were used to determine the associations among IVCM imaging features, ocular surface parameters, and tear mediators to account for the correlation between the eyes and to adjust for potential confounders, including age, sex, duration of diabetes, BMI, and HbA1c. Statistical analyses were performed using SPSS (version 28) software (IBM Corp., Armonk, NY). Cumulative LMMs (CLMMs) were performed to evaluate the association among corneal sensitivity, corneal nerve metrics, and the stages of DR, with adjustment for the same set of covariates, using the ordinal package in R (version 4.2.1) (R Foundation for Statistical Computing) (34). P < 0.05 was considered statistically significant.
Data and Resource Availability
The data sets generated and/or analyzed during the current study are available under a research agreement with the corresponding author upon reasonable request.
Results
Participants Characteristics
Among the 634 patients with type 2 diabetes, 277 (43.7%) were female, and 382 (60.3%) were Chinese, 175 (27.6%) Indian, 45 (7.1%) Malay, and 32 (5.0%) other ethnicities. The mean age was 59.9 ± 9.8 years. Of these patients, 427 (67.4%) had no DR, 156 (24.6%) had NPDR, and 51 (8.0%) had PDR. Among the 188 control participants, 87 (46.3%) were female, and 119 (63.3%) were Chinese, 52 (27.7%) Indian, 13 (6.9%) Malay, and 4 (2.1%) other ethnicities. The mean age was 59.3 ± 10.2 years, which was comparable with that in the cohort with diabetes (P = 0.280) (Table 1).
Table 1.
Demographic characteristics of control participants and patients with no DR, NPDR, and PDR
| Demographic | Control group (n = 188) | No-DR group (n = 427) | NPDR group (n = 156) | PDR group (n = 51) | P |
|---|---|---|---|---|---|
| Age, years | 59.3 ± 10.2 | 60.3 ± 11.5 | 58.0 ± 12.2 | 58.9 ± 13.4 | 0.280 |
| Sex, female/male | 87/101 | 192/235 | 65/91 | 20/31 | |
| Ethnicity, Chinese/Indian/Malay/other | 119/52/13/4 | 263/122/22/20 | 87/38/19/12 | 32/15/4/0 | |
| Duration of diabetes, years | — | 14.3 ± 8.3 | 18.9 ± 9.3 | 22.0 ± 10.8 | <0.001 |
| HbA1c, % (mmol/mol) | 5.8 (40) ± 0.5 | 7.9 (63) ± 1.4 | 8.3 (67) ± 1.6 | 8.3 (67) ± 1.3 | <0.001 |
| BMI, kg/m2 | 24.9 ± 9.6 | 27.2 ± 5.0 | 27.9 ± 5.2 | 27.2 ± 4.3 | <0.001 |
| Fasting blood glucose, mmol/L | 5.4 ± 1.4 | 8.8 ± 2.7 | 9.0 ± 3.6 | 8.8 ± 5.0 | <0.001 |
Data are mean ± SD unless otherwise indicated. Boldface indicates significance at P < 0.05 for overall comparisons among the data of the four groups.
Corneal Nerve Changes in Various Stages of DR
Corneal nerve images and fundus photos showed that in patients with diabetes, even in the absence of DR, corneal nerve density and length had been markedly reduced compared with control participants. The corneal nerve status, including nerve density and length, as well as tortuosity, deteriorated with the severity of DR stages (Fig. 1).
Figure 1.
Representative IVCM images, fundus photos, and slit-lamp photographs of control participants and patients with different severities of DR. Corneal nerve images of healthy control participants (A) and patients with no DR (B), NPDR (C), and PDR (D). Fundus photos of healthy control participants (E) and patients with no DR (F), NPDR (G), and PDR (H). Slit-lamp microscopic images of healthy control participants (I) and patients with no DR (J), NPDR (K), and PDR (L). Corneal nerves were significantly impaired before DR manifested clinically, and corneal nerve status and ocular surface integrity deteriorated with the severity of DR stages.
On quantitative analysis, compared with control participants, all patients with diabetes, irrespective of their DR status, had significantly lower CNFL, CNFD, CTBD, CNBD, CNFA, and CFracDim and significantly higher CNFW (all P < 0.001). Compared with patients with no DR, patients with NPDR and patients with PDR showed significantly reduced CNFD (P = 0.003 and P = 0.009, respectively), CNFL (P = 0.003 and P = 0.002, respectively), and CFracDim (both P = 0.001). Additionally, patients with PDR demonstrated significantly lower CNFD than those with no DR (P = 0.009). Patients with NPDR showed significantly reduced CNFA (P = 0.030) and increased CNFW (P = 0.009) compared with patients with no DR. Patients with PDR had worse corneal nerve metrics than those with NPDR, although the difference was not statistically significant (Table 2 and Fig. 2).
Table 2.
Comparison of IVCM imaging features and ocular surface assessments in control participants and patients with no DR, NPDR, and PDR
| Parameter | Control group (n = 188) | No DR group (n = 427) | NPDR group (n = 156) | PDR group (n = 51) | P (overall) | P (control vs. no-DR group) | P (control vs. NPDR group) | P (control vs. PDR group) | P (no DR vs. NPDR group) | P (no DR vs. PDR group) | P (NPDR vs. PDR group) |
|---|---|---|---|---|---|---|---|---|---|---|---|
| Corneal nerve | |||||||||||
| CNFD (n/mm2) | 16.2 ± 7.7 | 12.0 ± 5.0 | 10.5 ± 5.3 | 9.6 ± 4.9 | <0.001 | <0.001 | <0.001 | <0.001 | 0.003 | 0.009 | 0.725 |
| CNBD (n/mm2) | 16.9 ± 13.6 | 10.9 ± 8.2 | 9.6 ± 8.0 | 8.4 ± 7.4 | <0.001 | <0.001 | <0.001 | <0.001 | 0.273 | 0.195 | 0.828 |
| CNFL (mm/mm2) | 10.6 ± 3.8 | 8.2 ± 2.6 | 7.4 ± 2.6 | 6.8 ± 2.5 | <0.001 | <0.001 | <0.001 | <0.001 | 0.003 | 0.002 | 0.527 |
| CTBD (n/mm2) | 27.0 ± 16.8 | 20.0 ± 12.9 | 17.5 ± 11.1 | 16.2 ± 10.4 | <0.001 | <0.001 | <0.001 | <0.001 | 0.067 | 0.129 | 0.905 |
| CNFA (µm2/mm2) | 4,834.3 ± 1,664.0 | 3,998.4 ± 1,654.2 | 3,684.0 ± 1,251.5 | 3,576.5 ± 1,192.3 | <0.001 | <0.001 | <0.001 | <0.001 | 0.030 | 0.260 | 0.750 |
| CNFW (µm/mm2) | 21.5 ± 1.2 | 21.9 ± 1.1 | 22.2 ± 1.3 | 22.2 ± 1.4 | <0.001 | <0.001 | <0.001 | <0.001 | 0.009 | 0.180 | 0.820 |
| CFracDim | 1.43 ± 0.05 | 1.39 ± 0.05 | 1.37 ± 0.05 | 1.36 ± 0.05 | <0.001 | <0.001 | <0.001 | <0.001 | 0.001 | 0.001 | 0.485 |
| DC | |||||||||||
| Count | 9.0 (5.2–12.5) | 10.5 (6.7–16.0) | 11.1 (7.6–15.4) | 10.5 (7.4–17.2) | 0.010 | 0.009 | 0.082 | 0.058 | 0.973 | 0.905 | 0.820 |
| Density (μm−2) | 0.0202 ± 0.0044 | 0.0231 ± 0.0034 | 0.0227 ± 0.0028 | 0.0230 ± 0.0031 | <0.001 | <0.001 | <0.001 | <0.001 | 0.462 | 0.956 | 0.960 |
| Area (μm2) | 45.1 ± 6.0 | 45.0 ± 6.1 | 45.6 ± 6.2 | 45.3 ± 5.8 | 0.782 | 0.999 | 0.935 | 0.999 | 0.733 | 0.992 | 0.991 |
| Elongation | 0.58 ± 0.16 | 0.62 ± 0.06 | 0.61 ± 0.06 | 0.60 ± 0.06 | <0.001 | <0.001 | <0.001 | 0.122 | 0.976 | 0.796 | 0.702 |
| Average length (μm) | 4.8 ± 5.9 | 10.8 ± 1.4 | 10.9 ± 1.6 | 10.8 ± 1.4 | <0.001 | <0.001 | <0.001 | <0.001 | 0.975 | 0.993 | 0.963 |
| Microneuromas | |||||||||||
| Total area (μm2) | 283.3 (139.2–363.5) | 317.2 (209.4–537.3) | 372.1 (225.8–536.1) | 421.4 (238.5–608.8) | 0.057 | 0.063 | 0.046 | 0.453 | 0.920 | 0.997 | 0.998 |
| Average size (μm) | 34.7 (24.8–53.3) | 73.4 (37.6–172.0) | 68.4 (37.0–132.1) | 82.6 (19.6–205.5) | <0.001 | <0.001 | 0.004 | 0.062 | 0.846 | 0.999 | 0.988 |
| Perimeter (μm) | 21.7 (17.0–61.9) | 39.3 (22.8–106.8) | 30.6 (21.9–48.2) | 35.2 (15.6–70.4) | 0.054 | 0.078 | 0.737 | 0.998 | 0.396 | 0.494 | 0.947 |
| Ocular surface assessments | |||||||||||
| Corneal sensitivity (cm) | 28.6 ± 2.0 | 28.6 ± 2.4 | 28.4 ± 2.5 | 27.0 ± 4.3 | <0.001 | 0.951 | 0.930 | <0.001 | 0.668 | <0.001 | <0.001 |
| TBUT (s) | 5.0 (3.0–6.0) | 5.0 (3.0–6.0) | 4.0 (3.0–5.0) | 4.0 (2.0–5.0) | <0.001 | 0.998 | 0.010 | 0.024 | 0.004 | 0.018 | 0.938 |
| Schirmer test (mm) | 10.0 (6.0–14.5) | 8.0 (4.0–15.0) | 6.0 (3.0–11.0) | 6.0 (2.0–10.0) | <0.001 | 0.374 | 0.001 | 0.002 | 0.008 | 0.020 | 0.831 |
| Oxford score | 0.0 (0.0-0.0) | 0.0 (0.0-0.0) | 0.0 (0.0-0.0) | 0.0 (0.0–1.0) | 0.001 | 0.965 | 0.981 | <0.001 | 0.998 | <0.001 | 0.003 |
| NEI score | 0.0 (0.0–1.0) | 0.0 (0.0-0.0) | 0.0 (0.0-0.0) | 0.0 (0.0–2.0) | 0.001 | 0.367 | 0.411 | <0.001 | 0.993 | 0.005 | 0.022 |
| OSDI score | 7.4 ± 9.7 | 6.9 ± 10.6 | 9.4 ± 14.4 | 14.2 ± 13.6 | 0.008 | 0.968 | 0.525 | 0.019 | 0.284 | 0.009 | 0.207 |
| Tear mediators | |||||||||||
| SP (pg/mL) | 1,131.0 ± 569.6 | 636.1 ± 290.6 | 569.3 ± 150.6 | 603.7 ± 304.8 | <0.001 | <0.001 | <0.001 | <0.001 | 0.843 | 0.978 | 0.974 |
| MMP-9 (ng/mL) | 3.20 ± 2.21 | 6.71 ± 6.57 | 6.97 ± 5.24 | 10.77 ± 8.92 | <0.001 | 0.060 | 0.037 | <0.001 | 0.998 | 0.021 | 0.035 |
| IGFBP-3 (ng/mL) | 0.049 ± 0.053 | 0.060 ± 0.087 | 0.042 ± 0.042 | 0.108 ± 0.133 | 0.004 | 0.939 | 0.985 | 0.015 | 0.788 | 0.070 | 0.005 |
| TIMP-1 (ng/mL) | 0.57 ± 0.71 | 0.55 ± 0.69 | 0.44 ± 0.94 | 0.84 ± 1.45 | 0.328 | 0.999 | 0.935 | 0.630 | 0.964 | 0.556 | 0.283 |
Data are mean ± SD or median (interquartile range) unless otherwise indicated. Boldface indicates significance at P < 0.05.
Figure 2.
Bar charts showing the corneal nerve parameters, ocular surface clinical parameters, and tear mediators in healthy control participants and patients with no DR, NPDR, and PDR. CNFD (A), CNBD (B), CNFL (C), CTBD (D), CNFA (E), CNFW (F), CFracDim (G), corneal sensitivity (H), Schirmer test (I), TBUT (J), NEI score (K), Oxford score (L), OSDI score (M), tear MMP-9 (N), tear SP (O), and tear IGFPB-3 (P). *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.
Corneal Microneuromas and DCs in Different Stages of DR
Patients with no DR, NPDR, and PDR showed significantly increased DC length and density compared with control participants (all P < 0.001). Patients with no DR also presented significantly higher DC counts than control participants (P = 0.009), while both the NPDR and no-DR groups demonstrated significantly increased elongation compared with the control group (both P < 0.001) (Table 2 and Fig. 3).
Figure 3.
Representative IVCM images of DCs and bar charts showing the DC parameters in healthy control participants and patients with different severities of DR. IVCM images of healthy control participants (A) and patients with no DR (B), NPDR (C), and PDR (D). Red arrows indicate DCs. Bar charts of DC count (E), DC density (F), DC length (G), and DC elongation (H). **P < 0.01, ***P < 0.001, ****P < 0.0001.
For the evaluation of microneuromas, significantly larger size and total area of microneuromas were observed in the NPDR group compared with the control group (P < 0.001 and P = 0.047, respectively). Patients with no DR also exhibited significantly larger microneuroma size than control participants (P = 0.005) (Table 2 and Fig. 4).
Figure 4.
Representative IVCM images of microneuromas of control participants and patients with different severities of DR. IVCM images of healthy control participants (A) and patients with no DR (B), NPDR (C), and PDR (D). Red arrows indicate corneal microneuromas. Bar charts of microneuromas area (E) and microneuromas size (F). *P < 0.05, **P < 0.01, ****P < 0.001.
Comparisons of Ocular Surface Assessments in Different Stages of DR
Compared with patients with no DR, patients with NPDR, and control participants, patients with PDR had significantly worse corneal sensitivity (all P < 0.001), Oxford scores (P < 0.001, P = 0.003, and P < 0.001, respectively), and NEI scores (P = 0.005, P = 0.022, and P < 0.001, respectively). Patients with NPDR demonstrated significantly reduced TBUT and Schirmer test values compared with control participants (P = 0.010 and P = 0.001, respectively) and patients with no DR (P = 0.004 and P = 0.008, respectively). Similarly, patients with PDR showed significantly lower TBUT and Schirmer values compared with control participants (P = 0.024 and P = 0.002, respectively) and patients with no DR (P = 0.018 and P = 0.020, respectively). Furthermore, the subjective symptoms of the PDR group were significantly more severe than in the no-DR group and control group in terms of the total OSDI score (P = 0.009 and P = 0.019, respectively) (Table 2 and Fig. 2).
Changes in Tear Cytokines and Neuromediators in Different Stages of DR
Tear SP concentrations were significantly lower in patients with any stage of DR, including no DR, NPDR, and PDR, compared with control participants (all P < 0.001). Tear MMP-9 concentrations were significantly increased in the PDR group compared with the control (P < 0.001), no-DR (P = 0.021), and NPDR (P = 0.035) groups. Patients with NPDR also exhibited significantly higher tear MMP-9 concentrations than control participants (P = 0.037). For tear IGFBP-3, significantly higher concentrations were noted in patients with PDR compared with those with NPDR (P = 0.005) and control participants (P = 0.015). No significant differences were observed in tear TIMP-1 concentrations across the groups (Table 2 and Fig. 2).
Relationship Among Corneal Imaging Parameters, Ocular Surface Assessments, and Tear Mediators in Patients With Type 2 Diabetes
Among the patients with diabetes, multivariate LMM demonstrated that corneal sensitivity was significantly positively associated with CNFD (β = 0.051; P = 0.007), CNBD (β = 0.027; P = 0.013), CNFL (β = 0.120; P = 0.002), CTBD (β = 0.016; P = 0.033), CNFA (β = 139.548; P = 0.044), CFracDim (β = 5.449; P = 0.003), area of DC (β = 0.242; P = 0.017), and elongation of DCs (β = 0.057; P = 0.029) and negatively associated with CNFW (β = −157.599; P = 0.040) and density of DCs (β = −0.0002; P < 0.001). Oxford scores were significantly positively associated with CNFW (β = 39.734; P = 0.047) and density of DCs (β = 0.0001; P < 0.001) and negatively associated with the area of DCs (β = −1.015; P = 0.018). A higher NEI score was also significantly associated with higher density (β = 0.0003; P < 0.001) and smaller area (β = −0.320; P = 0.022) of DCs. A higher OSDI score was significantly associated with a larger area of DCs (β = 0.115; P = 0.002) (Table 3).
Table 3.
Multivariate LMMs for the association between ocular surface assessments and IVCM imaging parameters
| Corneal sensitivity | TBUT | Schirmer test | Oxford score | NEI score | OSDI score | |||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Parameter | β (95% CI) | P | β (95% CI) | P | β (95% CI) | P | β (95% CI) | P | β (95% CI) | P | β (95% CI) | P |
| Corneal nerve | ||||||||||||
| CNFD | 0.051 (0.014 to 0.088) | 0.007 | 0.012 (−0.019 to 0.044) | 0.436 | −0.074 (−0.202 to 0.053) | 0.254 | −0.001 (−0.011 to 0.008) | 0.788 | 0.016 (−0.015 to 0.048) | 0.309 | 0.003 (−0.005 to 0.011) | 0.443 |
| CNBD | 0.027 (0.006 to 0.049) | 0.013 | −0.011 (−0.028 to 0.007) | 0.251 | −0.030 (−0.103 to 0.044) | 0.427 | −0.001 (−0.006 to 0.005) | 0.859 | 0.003 (−0.015 to 0.021) | 0.776 | −0.002 (−0.007 to 0.003) | 0.475 |
| CNFL | 0.120 (0.045 to 0.195) | 0.002 | −0.021 (−0.087 to 0.044) | 0.522 | −0.050 (−0.314 to 0.214) | 0.709 | 0.000 (−0.020 to 0.020) | 0.998 | 0.039 (−0.026 to 0.104) | 0.238 | 0.001 (−0.017 to 0,019) | 0.911 |
| CTBD | 0.016 (0.001 to 0.031) | 0.033 | −0.011 (−0.063 to 0.040) | 0.093 | −0.011 (−0.063 to 0.040) | 0.667 | 0.000 (−0.004 to 0.004) | 0.994 | 0.001 (−0.011 to 0.014) | 0.845 | −0.002 (−0.006 to 0.001) | 0.186 |
| CNFA | 139.548 (3.806 to 275.290) | 0.044 | −113.246 (−227.632 to 1.140) | 0.053 | −94.497 (−579.528 to 390.534) | 0.703 | −2.632 (−38.629 to 33.364) | 0.886 | 50.227 (−65.696 to 166.151) | 0.396 | −27.461 (−61.352 to 6.431) | 0.114 |
| CNFW | −157.599 (−308.045 to −7.153) | 0.040 | −85.759 (−208.268 to 36.751) | 0.171 | 116.672 (−408.604 to 641.948) | 0.663 | 39.734 (0.650 to 78.823) | 0.047 | 57.891 (−70.248 to 186.031) | 0.376 | −21.435 (−49.897 to 7.026) | 0.141 |
| CFracDim | 5.449 (1.845 to 9.054) | 0.003 | −0.592 (−3.696 to 2.512) | 0.709 | −0.268 (−12.892 to 12.357) | 0.967 | 0.107 (−0.852 to 1.065) | 0.827 | 2.051 (−1.036 to 5.139) | 0.193 | −0.229 (−1.106 to 0.649) | 0.610 |
| DC | ||||||||||||
| Cell count | 0.022 (−0.290 to 0.333) | 0.892 | −0.182 (−0.551 to 0.187) | 0.334 | 0.033 (−0.053 to 0.119) | 0.448 | −1.771 (−3.124 to −0.418) | 0.011 | −0.132 (−0.564 to 0.300) | 0.550 | −0.062 (−0.186 to 0.062) | 0.330 |
| Density | −0.0002 (−0.0003 to −0.0001) | <0.001 | 0.0000 (−0.0001 to 0.0000) | 0.226 | −0.000 (−0.000 to −1.454) | 0.050 | 0.0001 (0.0000 to 0.0010) | <0.001 | 0.0003 (0.0001 to 0.0004) | <0.001 | −0.0000 (−0.0001 to 0.0000) | 0.192 |
| Area | 0.242 (0.044 to 0.440) | 0.017 | 0.137 (−0.088 to 0.361) | 0.235 | 0.019 (−0.035 to 0.712) | 0.493 | −1.015 (−1.850 to −0.180) | 0.018 | −0.320 (−0.592 to −0.048) | 0.022 | 0.115 (0.045 to 0.185) | 0.002 |
| Elongation | 0.057 (0.006 to 0.108) | 0.029 | −0.007 (−0.103 to 0.089) | 0.881 | 0.011 (−0.012 to 0.035) | 0.357 | −0.071 (−0.381 to 0.240) | 0.657 | −0.054 (−0.145 to 0.038) | 0.251 | 0.012 (−0.060 to 0.082) | 0.751 |
| Average length | 0.049 (−0.022 to 0.120) | 0.176 | 0.054 (−0.062 to 0.170) | 0.364 | 0.019 (−0.006 to 0.044) | 0.133 | −0.312 (−0.694 to 0.069) | 0.109 | −0.037 (−0.150 to 0.077) | 0.527 | 0.006 (−0.064 to 0.075) | 0.876 |
| Microneuroma | ||||||||||||
| Total area | −7.766 (−22.008 to 6.475) | 0.287 | −1.143 (−16.227 to 13.940) | 0.882 | 4.785 (0.653 to 8.918) | 0.024 | 22.922 (−25.404 to 71.249) | 0.354 | −3.236 (−19.540 to 13.069) | 0.698 | −11.945 (−30.310 to 6.420) | 0.206 |
| Average length | 1.647 (−4.924 to 8.219) | 0.624 | −1.167 (−9.681 to 7.348) | 0.789 | 3.463 (1.382 to 5.545) | 0.001 | 2.335 (−25.404 to 71.249) | 0.869 | −1.347 (−10.423 to 7.729) | 0.771 | −3.714 (−11.006 to 3.579) | 0.321 |
| Perimeter | −0.609 (−4.502 to 3.284) | 0.760 | −0.375 (−6.271 to 5.522) | 0.901 | 1.304 (−2.486 to 5.095) | 0.501 | 12.611 (−0.927 to 26.150) | 0.071 | −2.549 (−5.940 to 0.843) | 0.144 | −24.565 (−57.220 to 8.090) | 0.145 |
Boldface indicates significance at P < 0.05 and adjusted for age, sex, duration of diabetes, BMI, and HbA1c.
Tear SP was significantly associated with the size of microneuromas (β = −0.161; P = 0.041). Tear MMP-9 was significantly associated with the elongation (β = −0.100; P = 0.010) and average length (β = −0.002; P = 0.030) of DCs, suggesting a link between higher concentrations of tear MMP-9 and immature DCs. Elevated tear IGFBP-3 was significantly associated with higher NEI score (β = 0.301; P = 0.032) (Table 4). After adjusting for age, sex, duration of diabetes, BMI, and HbA1c, CLMMs showed that lower corneal sensitivity (odds ratio [OR] 1.346), CNFD (OR 1.152), CNBD (OR 1.319), and CNFL (OR 1.238) were significantly associated with more severe stages of DR (all P < 0.001).
Table 4.
Multivariate LMM analysis for the association between tear mediators with ocular surface assessments and IVCM imaging parameters
| Tear SP | Tear MMP-9 | Tear IGFBP-3 | Tear TIMP-1 | |||||
|---|---|---|---|---|---|---|---|---|
| Parameter | β (95% CI) | P | β (95% CI) | P | β (95% CI) | P | β (95% CI) | P |
| Ocular surface assessment | ||||||||
| Schirmer test | 0.002 (−0.007 to 0.011) | 0.619 | −0.088 (−0.313 to 0.136) | 0.442 | 0.516 (−0.608 to 1.639) | 0.371 | 0.296 (0.175 to 0.416) | 0.001 |
| TBUT | 0.000 (−0.002 to 0.002) | 0.935 | −0.024 (−0.071 to 0.024) | 0.335 | −0.053 (−0.270 to 0.165) | 0.637 | 0.019 (−0.007 to 0.044) | 0.159 |
| Oxford score | −0.000 (−0.001 to 0.001) | 0.542 | −0.007 (−0.027 to 0.013) | 0.500 | 0.020 (−0.070 to 0.111) | 0.660 | −0.004 (−0.015 to 0.007) | 0.523 |
| NEI score | 0.001 (−0.002 to 0.003) | 0.628 | −0.015 (−0.085 to 0.055) | 0.672 | 0.301 (0.030 to 0.572) | 0.032 | −0.036 (−0.071 to −0.001) | 0.049 |
| Corneal sensitivity | 0.001 (−0.004 to 0.006) | 0.722 | −0.014 (−0.180 to 0.152) | 0.869 | 0.080 (−0.571 to 0.731) | 0.810 | 0.025 (−0.054 to 0.104) | 0.534 |
| Corneal nerve | ||||||||
| CNFD | −0.003 (−0.007 to 0.002) | 0.279 | −0.067 (−0.207 to 0.073) | 0.350 | 0.279 (−0.344 to 0.903) | 0.383 | 0.029 (−0.043 to 0.101) | 0.434 |
| CNBD | −0.001 (−0.008 to 0.006) | 0.776 | 0.043 (−0.169 to 0.255) | 0.693 | 0.667 (−0.247 to 1.581) | 0.157 | 0.018 (−0.090 to 0.126) | 0.741 |
| CNFL | −0.001 (−0.003 to 0.002) | 0.594 | −0.022 (−0.089 to 0.044) | 0.512 | 0.202 (−0.096 to 0.499) | 0.188 | 0.010 (−0.025 to 0.045) | 0.572 |
| CTBD | 0.001 (−0.010 to 0.011) | 0.896 | 0.114 (−0.210 to 0.438) | 0.492 | 1.123 (−0.251 to 2.497) | 0.114 | 0.001 (−0.163 to 0.164) | 0.998 |
| CNFA | 0.071 (−1.027 to 1.169) | 0.900 | 2.519 (−31.193 to 36.231) | 0.884 | 70.287 (−75.271 to 215.845) | 0.347 | 4.05 (13.00 to 21.20) | 0.643 |
| CNFW | 0.001 (0.000 to 0.002) | 0.043 | −0.009 (−0.031 to 0.036) | 0.628 | 0.097 (−3.674 to 6.744) | 0.227 | 0.003 (−0.015 to 0.022) | 0.753 |
| CFracDim | −0.00001 (−0.0001 to 0.00002) | 0.509 | −0.001 (−0.002 to 0.001) | 0.385 | 0.002 (−0.004 to 0.007) | 0.565 | 0.0001 (−0.0005 to 0.0007) | 0.727 |
| DC | ||||||||
| Count | 0.002 (−0.004 to 0.009) | 0.491 | −0.220 (−0.453 to 0.014) | 0.072 | 0.263 (−0.607 to 1.132) | 0.557 | 0.044 (−0.059 to 0.147) | 0.408 |
| Density | −0.00001 (−0.00004 to 0.00003) | 0.846 | 0.00007 (−0.00004 to 0.0002) | 0.229 | 0.00002 (−0.0005 to 0.0005) | 0.942 | 0.000(−0.004 to 0.006) | 0.741 |
| Area | −0.0002 (−0.0096 to 0.0090) | 0.952 | −0.278 (−0.590 to 0.034) | 0.087 | −0.218 (−1.444 to 1.008) | 0.730 | 0.019 (−0.126 to 0.163) | 0.801 |
| Elongation | −0.001 (−0.004 to 0.002) | 0.472 | −0.100 (−0.173 to −0.027) | 0.010 | 0.041 (−0.269 to 0.352) | 0.795 | 0.013 (−0.024 to 0.050) | 0.488 |
| Average length | −0.00002 (−0.0001 to 0.00005) | 0.517 | −0.002 (−0.004 to −0.0003) | 0.030 | −0.001 (−0.010 to 0.008) | 0.854 | 0.0003 (−0.0001 to 0.0012) | 0.591 |
| Microneuroma | ||||||||
| Total area | −0.175 (−0.523 to 0.173) | 0.341 | −6.777 (−20.355 to 6.802) | 0.361 | −34.968 (−96.282 to 26.346) | 0.282 | 1.478 (−3.030 to 5.986) | 0.533 |
| Average size | −0.161 (−0.305 to −0.018) | 0.041 | −2.286 (−7.689 to 3.117) | 0.417 | −20.251 (−38.533 to −1.969) | 0.042 | −0.997 (−2.432 to 0.438) | 0.188 |
| Perimeter | −0.205 (−0.633 to 0.222) | 0.362 | 0.340 (−14.122 to 14.802) | 0.964 | −28.067 (−99.673 to 43.578) | 0.456 | −0.174 (−5.709 to 5.361) | 0.952 |
Boldface indicates significance at P < 0.05 and adjusted for age, sex, duration of diabetes, BMI, and HbA1c.
Discussion
In the current study, we comprehensively investigated corneal nerve metrics, ocular surface integrity, and tear mediators in patients with different stages of DR in a large population. We demonstrated that corneal nerve status was impaired even in the absence of clinically manifested DR. As the DR status deteriorated, the impairment of the corneal nerves, corneal sensitivity, tear stability, and ocular surface integrity became more prominent. Tear neuroinflammatory and inflammatory mediators were significantly associated with corneal nerves, DC variables, and ocular surface integrity. These findings suggest that corneal nerve fiber damage precedes DR and worsens with the severity of DR. Thus, corneal nerve status may serve as an early indicator and predictor of DR.
In our cohort of patients with diabetes, the prevalence of NPDR and PDR was 24.6% and 8.1%, respectively. These prevalences are consistent with a meta-analysis reporting the prevalence of NPDR and PDR among individuals with type 2 diabetes of 27% and 6%, respectively (35). Of note, we found that all the corneal nerve metrics were impaired compared with control participants, even when no DR signs were present, suggesting that corneal nerve damage occurs prior to the clinical onset of DR. As the degree of DR advanced, corneal nerve status gradually worsened. These findings support the parallel progression of diabetic microvascular complications, including DR and DCN, which share common pathogenetic mechanisms. Chronic hyperglycemia results in the simultaneous course of multiple metabolic pathways, such as advanced glycation end products formation, polyol pathway, hexosamine pathway flux, protein kinase C activation, oxidative stress, and inflammatory responses (2). These pathways result in neuronal degeneration and microvascular dysfunction, the key factors in the development of DR and DCN. Our CLMM analysis revealed that corneal sensitivity and corneal nerve parameters, including CNFD, CNBD, and CNFL, were significantly independently associated with DR severity. The results reinforce the relationship among DCN, ocular surface, and DR, and corneal nerve degeneration parallels retinopathy in patients with diabetes. The observed associations suggest that patients with diabetes who have impaired corneal sensitivity and decreased CNFD, CNBD, and CNFL are susceptible to developing DR, although this remains an extrapolation that requires confirmation through longitudinal studies. More importantly, recent studies have shown that neurodegeneration occurs early in diabetes, even preceding clinically evident microangiopathy. Chronic hyperglycemia reduces perfusion to neurons and slows nerve conduction velocity, ultimately resulting in neural damage and denervation (15). The small nerve fibers, such as corneal nerves, which lack myelin protection, are susceptible to injury (15). The loss of neuronal innervation may contribute to the onset of microvascular dysfunction, capillary nonperfusion, and subsequent proliferative changes (36). Meanwhile, the integrity of the blood-retinal barrier offers some initial protection against the metabolic disturbances and oxidative stress associated with diabetes (37). Consequently, corneal nerve fiber damage occurs prior to the clinical onset of both retinal neurodegeneration and DR (38). These highlight the importance of identifying small fiber neuropathies, such as DCN, for the early detection and prediction of microvascular complications, including DR (6). IVCM enables the visualization of small fiber nerves directly, making it easier to use in clinical settings. In some patients, the fundus examination could be hindered due to cloudy refractive interstitials, such as corneal haze and lens opacity, or the dilated examination is infeasible. All these affect accurate monitoring and delay the early diagnosis of and intervention for DR. The corneal nerve status could reflect the neuronal and vascular disturbances of DR.
There was a significant increase in the density, length, and elongation of DCs in all patients with diabetes compared with control participants, indicating that chronic inflammation of the ocular surface in diabetes occurs in the absence of observable signs of DR. Inflammation is also an important part of the pathogenesis of DR (39). Our results agree with previous studies, in which researchers reported significantly higher DC density in patients with diabetes compared with control participants (40,41). In contrast, some studies reported that DC density decreased with the progression of diabetes (42,43). It is not clear whether DCs are neuroprotective or neurotoxic during neurogenic inflammation (14). An increased density of corneal DCs has been shown to correlate with a decrease in CNFD, implicating the interrelationship between neuroinflammation and corneal nerve fiber damage in diabetes (44). We also showed the increased area and size of microneuromas in eyes of patients with no DR and NPDR compared with control participants. Corneal microneuromas have been regarded as pathological indicators of corneal neuropathy and ocular surface dysfunction (15), representing unsuccessful axon attempts to regenerate and restore neural continuity following injury. Our findings of microneuromas further corroborate that more severe corneal nerve damage occurs in more severe DR stages.
In the current study, we found that worse ocular surface health in the PDR group compared with the other groups, as evidenced by significantly decreased corneal sensitivity, Schirmer values, and TBUT, increased corneal staining scores and OSDI scores. The impaired nerve innervation in PDR decreases corneal sensitivity and diminishes blink frequency and trophic support, thereby resulting in worse homeostasis of the ocular surface and more severe ocular surface discomfort. The loss of corneal sensation also reduces lacrimal tear production since the corneal receptors are the afferent limb of the lacrimal reflex arc (45). Our findings are consistent with previous studies reporting that the more severe corneal denervation in patients with PDR is linked to abnormal tear function, as well as more frequent and severe dry eye symptoms in patients with diabetes (45). Singer et al. (36) also reported that corneal sensitivity was impaired in PDR eyes compared with NPDR eyes and proposed that corneal sensitivity testing may identify patients with DR. Of note, our results further highlight that structural and morphological alterations of the corneal nerves precede or exist in the absence of clinically detectable functional changes: impaired corneal nerve metrics had been observed in the no-DR group, while clinical ocular surface alterations or decreased corneal sensitivity manifested only in the PDR stage. These collectively suggest that using IVCM to monitor the anatomical corneal nerve plexus, rather than using clinical functional tests such as corneal sensitivity, would be more sensitive and reliable for the early detection of DR.
It is no surprise that corneal sensitivity and ocular integrity were associated with corneal nerve parameters. Notably, we found that ocular surface and corneal staining were closely associated with more DC parameters than corneal nerve metrics, suggesting that ocular surface integrity may reflect underlying immune-mediated and inflammatory activities. Our findings also align with previous studies on dry eye, which have reported a significant association between conjunctival and corneal staining, the presence of DCs, and the expression of inflammatory cytokines (46,47).
Tear sample collection is noninvasive, presenting a superior source of biofluid for identifying biomarkers compared with aqueous and vitreous fluids. We observed elevated tear MMP-9 concentrations in eyes with DR, and the concentrations were significantly associated with the maturity of DCs. The presence of immature DCs is associated with an increased ability to initiate an inflammatory response. Tear MMP-9 has been found to be linked with ocular surface inflammation (18), and increased MMP-9 has been reported in patients with type 2 diabetes in whom it is associated with corneal sensitivity (48). Additionally, elevated MMP-9 levels were previously detected in the vitreous and retina of both patients and animal models with DR (17,49), implying its role in DR pathogenesis. Specifically, increased MMP-9 in the diabetic retina contributes to vascular permeability through proteolytic degradation of the tight junction complexes (49). These results highlight the interconnection between tear MMP-9 and the pathogenesis of DR and DCN. Tear SP concentrations were reduced in groups with diabetes compared with the control group, and the concentrations were significantly associated with the size of microneuromas in a negative direction, suggesting the relationship between tear SP levels and corneal nerve degeneration in diabetes. SP is a key neuropeptide involved in neurogenic inflammation in diabetes (50), and our results align with previous findings showing a decreased concentration of tear SP in both patients with type 1 and type 2 diabetes, as well as a correlation between SP levels and corneal nerves (50,51). We also found an increased tear IGFBP-3 concentration in the PDR group compared with the NPDR and control groups, with an increased level of tear IGFBP-3 significantly associated with impaired ocular surface integrity. IGFBP-3 has been known to act as a stress response protein secreted by the corneal epithelium in response to hyperglycemia, and it is significantly associated with corneal nerve loss (52,53). Consistent with previous studies reporting 3- to 3.5-fold higher tear IGFBP-3 levels in individuals with diabetes than in counterparts without diabetes (54,55), our findings demonstrate that the concentration of tear IGFBP-3 increases with the severity of DR.
Due to the nature of the cross-sectional study, we could not determine the longitudinal changes in corneal nerves and ocular surface manifestations with the progression of DR. However, conducting such a study would require a time-intensive longitudinal design, as the development of mild NPDR has been reported to take ∼8 years (56), and the median duration for progression from moderate NPDR to PDR is ∼2 years (57). Additionally, future work is required to decipher the shared pathophysiological mechanisms of DCN and DR, particularly on the molecular level. This will also contribute to the identification of potential biomarkers or therapeutic targets in common pathways. Further studies that include a larger cohort with substaging of NPDR and incorporation of vascular imaging data, such as optical coherence tomography angiography and fluorescein angiography, will further elucidate the interrelationship between DCN and DR.
In conclusion, our study demonstrates that DCN presenting before DR is clinically manifested, and the corneal nerve and ocular surface status significantly deteriorate with the severity of DR stages. The evaluation of corneal nerves may serve as a surrogate marker to predict the occurrence and progression of DR. It is also suggested to assess the corneal nerves and ocular surface status in patients with DR, even in those with no DR. By examining the connection between DCN and DR, our work provides insights that could improve early detection, monitoring, and management of diabetes-related eye diseases.
This article contains supplementary material online at https://doi.org/10.2337/figshare.30999436.
Article Information
Duality of Interest. No potential conflicts of interest relevant to this article were reported.
Author Contributions. C.L. contributed to the manuscript writing, investigation, formal analysis, data curation, and visualization. I.X.Y.L. and A.A. performed the data curation and visualization. C.C.X. contributed to the methodology and reviewed the manuscript. M.Y. and R.K.T.W. contributed to the acquisition of data. C.-Y.C. reviewed the manuscript. Y.-C.L. contributed to the conceptualization, methodology, investigation, formal analysis, manuscript review, supervision, and funding acquisition. Y.-C.L. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Funding Statement
This study is supported by the clinician scientist award MOH-CSAINV24jul-0005 and clinician scientist individual research grant CSIRG24jul-0010 from the Singapore National Medical Research Council.
Supporting information
References
- 1. So WZ, Qi Wong NS, Tan HC, et al. Diabetic corneal neuropathy as a surrogate marker for diabetic peripheral neuropathy. Neural Regen Res 2022;17:2172–2178 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Mansoor H, Tan HC, Lin MT-Y, Mehta JS, Liu Y-C. Diabetic corneal neuropathy. J Clin Med 2020;9:3956. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Teo ZL, Tham Y-C, Yu M, et al. Global prevalence of diabetic retinopathy and projection of burden through 2045: systematic review and meta-analysis. Ophthalmology 2021;128:1580–1591 [DOI] [PubMed] [Google Scholar]
- 4. Simó R, Stitt AW, Gardner TW. Neurodegeneration in diabetic retinopathy: does it really matter? Diabetologia 2018;61:1902–1912 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Cabral-Pacheco GA, Garza-Veloz I, Castruita-De la Rosa C, et al. The roles of matrix metalloproteinases and their inhibitors in human diseases. Int J Mol Sci 2020;21:9739. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Toprak I, Fenkci SM, Fidan Yaylali G, Martin C, Yaylali V. Early retinal neurodegeneration in preclinical diabetic retinopathy: a multifactorial investigation. Eye (Lond) 2020;34:1100–1107 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Tavares Ferreira J, Alves M, Dias-Santos A, et al. Retinal neurodegeneration in diabetic patients without diabetic retinopathy. Invest Ophthalmol Vis Sci 2016;57:6455–6460 [DOI] [PubMed] [Google Scholar]
- 8. Sharma S, Vas P, Rayman G. Small fiber neuropathy in diabetes polyneuropathy: is it time to change? J Diabetes Sci Technol 2022;16:321–331 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Messmer EM, Schmid-Tannwald C, Zapp D, Kampik A. In vivo confocal microscopy of corneal small fiber damage in diabetes mellitus. Graefes Arch Clin Exp Ophthalmol 2010;248:1307–1312 [DOI] [PubMed] [Google Scholar]
- 10. Hafner J, Zadrazil M, Grisold A, et al. Retinal and corneal neurodegeneration and their association with systemic signs of peripheral neuropathy in type 2 diabetes. Am J Ophthalmol 2020;209:197–205 [DOI] [PubMed] [Google Scholar]
- 11. Zhivov A, Winter K, Hovakimyan M, et al. Imaging and quantification of subbasal nerve plexus in healthy volunteers and diabetic patients with or without retinopathy. PLoS One 2013;8:e52157. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Carmichael J, Fadavi H, Tavakoli M. Neurodegeneration of the cornea and retina in patients with type 1 diabetes without clinical evidence of diabetic retinopathy. Front Endocrinol (Lausanne) 2022;13:790255. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Hafner J, Karst S, Sacu S, Scholda C, Pablik E, Schmidt-Erfurth U. Correlation between corneal and retinal neurodegenerative changes and their association with microvascular perfusion in type II diabetes. Acta Ophthalmol 2019;97:e545–e550 [DOI] [PubMed] [Google Scholar]
- 14. Liu F, Liu C, Lee IXY, Lin MTY, Liu Y-C. Corneal dendritic cells in diabetes mellitus: a narrative review. Front Endocrinol (Lausanne) 2023;14:1078660. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Toh CJL, Liu C, Lee IXY, Yu Lin MT, Tong L, Liu Y-C. Clinical associations of corneal neuromas with ocular surface diseases. Neural Regen Res 2024;19:140–147 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Issar T, Tummanapalli SS, Kwai NCG, et al. Associations between acute glucose control and peripheral nerve structure and function in type 1 diabetes. Diabet Med 2020;37:1553–1560 [DOI] [PubMed] [Google Scholar]
- 17. Amorim M, Martins B, Caramelo F, et al. Putative biomarkers in tears for diabetic retinopathy diagnosis. Front Med (Lausanne) 2022;9:873483. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Markoulli M, Colorado LH, Edwards K. The relationship between corneal nerve morphology and inflammatory mediators and neuropeptides in healthy individuals. Optom Vis Sci 2020;97:145–153 [DOI] [PubMed] [Google Scholar]
- 19. American Diabetes Associations. Standards of medical care in diabetes–2010. Diabetes Care 2010;33(Suppl. 1):S11–S61 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. Sabanayagam C, He F, Nusinovici S, et al. Prediction of diabetic kidney disease risk using machine learning models: a population-based cohort study of Asian adults. Elife 2023;12:e81878. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Early Treatment Diabetic Retinopathy Study Research Group . Early photocoagulation for diabetic retinopathy. ETDRS report number 9. Ophthalmology 1991;98:766–785 [PubMed] [Google Scholar]
- 22. Chin JY, Lin MT-Y, Lee IXY, Mehta JS, Liu Y-C. Tear neuromediator and corneal denervation following SMILE. J Refract Surg 2021;37:516–523 [DOI] [PubMed] [Google Scholar]
- 23. Chin JY, Liu C, Lee IXY, et al. Impact of age on the characteristics of corneal nerves and corneal epithelial cells in healthy adults. Cornea 2024;43:409–418 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. Chin JY, Yang LWY, Ji AJS, et al. Validation of the use of automated and manual quantitative analysis of corneal nerve plexus following refractive surgery. Diagnostics (Basel) 2020;10:493. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Sindt CW, Lay B, Bouchard H, Kern JR. Rapid image evaluation system for corneal in vivo confocal microscopy. Cornea 2013;32:460–465 [DOI] [PubMed] [Google Scholar]
- 26. Liu C, Lin MT-Y, Lee IXY, et al. Neuropathic corneal pain: tear proteomic and neuromediator profiles, imaging features, and clinical manifestations. Am J Ophthalmol 2024;265:6–20 [DOI] [PubMed] [Google Scholar]
- 27. Teo CHY, Ong HS, Liu Y-C, Tong L. Meibomian gland dysfunction is the primary determinant of dry eye symptoms: analysis of 2346 patients. Ocul Surf 2020;18:604–612 [DOI] [PubMed] [Google Scholar]
- 28. Tsubota K. Short tear film breakup time-type dry eye. Invest Ophthalmol Vis Sci 2018;59:DES64–DES70 [DOI] [PubMed] [Google Scholar]
- 29. Liu Y-C, Yam GH-F, Lin MT-Y, et al. Comparison of tear proteomic and neuromediator profiles changes between small incision lenticule extraction (SMILE) and femtosecond laser-assisted in-situ keratomileusis (LASIK). J Adv Res 2021;29:67–81 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Teo CHY, Lin MT-Y, Lee IXY, et al. Oral peroxisome proliferator-activated receptor-α agonist enhances corneal nerve regeneration in patients with type 2 diabetes. Diabetes 2023;72:932–946 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31. The definition and classification of dry eye disease: report of the Definition and Classification Subcommittee of the International Dry Eye WorkShop (2007). Ocul Surf 2007;5:75–92 [DOI] [PubMed] [Google Scholar]
- 32. Wolffsohn JS, Arita R, Chalmers R, et al. TFOS DEWS II diagnostic methodology report. Ocul Surf 2017;15:539–574 [DOI] [PubMed] [Google Scholar]
- 33. Yawata N, Selva KJ, Liu Y-C, et al. Dynamic change in natural killer cell type in the human ocular mucosa in situ as means of immune evasion by adenovirus infection. Mucosal Immunol 2016;9:159–170 [DOI] [PubMed] [Google Scholar]
- 34. Christensen RHB. Package ‘Ordinal’: Regression Models for Ordinal Data. R Package Version 2015.2016-2028. Vienna, Comprehensive R Network, 2015 [Google Scholar]
- 35. Yang Q-H, Zhang Y, Zhang X-M, Li X-R. Prevalence of diabetic retinopathy, proliferative diabetic retinopathy and non-proliferative diabetic retinopathy in Asian T2DM patients: a systematic review and meta-analysis. Int J Ophthalmol 2019;12:302–311 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36. Singer M, O’Brien P, Mein L, Olvera A. Corneal sensitivity is inversely correlated with severity of diabetic retinopathy in a predominantly underrepresented population. Am J Ophthalmol 2024;259:53–61 [DOI] [PubMed] [Google Scholar]
- 37. Rudraraju M, Narayanan SP, Somanath PR. Regulation of blood-retinal barrier cell-junctions in diabetic retinopathy. Pharmacol Res 2020;161:105115. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38. Yu M, Ning FTE, Liu C, Liu Y-C. Interconnections between diabetic corneal neuropathy and diabetic retinopathy: diagnostic and therapeutic implications. Neural Regen Res 2025;20:2169–2180 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39. Sayin N, Kara N, Pekel G. Ocular complications of diabetes mellitus. World J Diabetes 2015;6:92–108 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40. Colorado LH, Beecher L, Pritchard N, et al. Corneal dendritic cell dynamics are associated with clinical factors in type 1 diabetes. J Clin Med 2022;11:2611. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41. Tavakoli M, Boulton AJM, Efron N, Malik RA. Increased Langerhan cell density and corneal nerve damage in diabetic patients: role of immune mechanisms in human diabetic neuropathy. Cont Lens Anterior Eye 2011;34:7–11 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42. Chen W-L, Lin C-T, Ko P-S, et al. In vivo confocal microscopic findings of corneal wound healing after corneal epithelial debridement in diabetic vitrectomy. Ophthalmology 2009;116:1038–1047 [DOI] [PubMed] [Google Scholar]
- 43. Tomlinson DR, Gardiner NJ. Glucose neurotoxicity. Nat Rev Neurosci 2008;9:36–45 [DOI] [PubMed] [Google Scholar]
- 44. Leppin K, Behrendt A-K, Reichard M, et al. Diabetes mellitus leads to accumulation of dendritic cells and nerve fiber damage of the subbasal nerve plexus in the cornea. Invest Ophthalmol Vis Sci 2014;55:3603–3615 [DOI] [PubMed] [Google Scholar]
- 45. Shih KC, Lam KS-L, Tong L. A systematic review on the impact of diabetes mellitus on the ocular surface. Nutr Diabetes 2017;7:e251. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 46. Aggarwal S, Kheirkhah A, Cavalcanti BM, Cruzat A, Jamali A, Hamrah P. Correlation of corneal immune cell changes with clinical severity in dry eye disease: an in vivo confocal microscopy study. Ocul Surf 2021;19:183–189 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 47. Yang S, Lee HJ, Kim D-Y, Shin S, Barabino S, Chung S-H. The use of conjunctival staining to measure ocular surface inflammation in patients with dry eye. Cornea 2019;38:698–705 [DOI] [PubMed] [Google Scholar]
- 48. Iyengar MF, Soto LF, Requena D, et al. Tear biomarkers and corneal sensitivity as an indicator of neuropathy in type 2 diabetes. Diabetes Res Clin Pract 2020;163:108143. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 49. Giebel SJ, Menicucci G, McGuire PG, Das A. Matrix metalloproteinases in early diabetic retinopathy and their role in alteration of the blood-retinal barrier. Lab Invest 2005;85:597–607 [DOI] [PubMed] [Google Scholar]
- 50. Markoulli M, You J, Kim J, et al. Corneal nerve morphology and tear film substance P in diabetes. Optom Vis Sci 2017;94:726–731 [DOI] [PubMed] [Google Scholar]
- 51. Tummanapalli SS, Willcox MDP, Issar T, et al. Tear film substance P: a potential biomarker for diabetic peripheral neuropathy. Ocul Surf 2019;17:690–698 [DOI] [PubMed] [Google Scholar]
- 52. Stuard WL, Titone R, Robertson DM. Tear levels of IGFBP-3: a potential biomarker for diabetic nerve changes in the cornea. Eye Contact Lens 2020;46:319–325 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 53. Titone R, Zhu M, Robertson DM. Mutual regulation between IGF-1R and IGFBP-3 in human corneal epithelial cells. J Cell Physiol 2019;234:1426–1441 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 54. Wu Y-C, Buckner BR, Zhu M, Cavanagh HD, Robertson DM. Elevated IGFBP3 levels in diabetic tears: a negative regulator of IGF-1 signaling in the corneal epithelium. Ocul Surf 2012;10:100–107 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 55. Stuard WL, Titone R, Robertson DM. Tear levels of insulin-like growth factor binding protein 3 correlate with subbasal nerve plexus changes in patients with type 2 diabetes mellitus. Invest Ophthalmol Vis Sci 2017;58:6105–6112 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 56. Seshasai S, He F, Lam B, et al. Transition probabilities of diabetic retinopathy and death in an Asian population with diabetes. Asia Pac J Ophthalmol (Phila) 2024;13:100070. [DOI] [PubMed] [Google Scholar]
- 57. Moshfeghi A, Garmo V, Sheinson D, Ghanekar A, Abbass I. Five-year patterns of diabetic retinopathy progression in US clinical practice. Clin Ophthalmol 2020;14:3651–3659 [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.