Reduced corneal nerve fibre length in prediabetes and type 2 diabetes: The Maastricht Study

Abstract Purpose In individuals with diabetes, injury to the corneal nerve fibres predisposes to delayed corneal epithelial healing, reduced corneal sensitivity and corneal erosion. We investigated to what extent a reduction in corneal nerve fibre length (CNFL) is present in individuals with prediabetes or type 2 diabetes (DM2) compared with individuals with normal glucose metabolism (NGM). Methods Using composite images acquired by corneal confocal microscopy, we assessed total CNFL per mm2 in the subbasal nerve plexus of the cornea in 134 participants (mean age 59 ± 8 years, 49% men, 87 NGM, 20 prediabetes, 27 DM2). Multivariable linear regression was used to assess the association between CNFL and glucose metabolism status, adjusted for age and sex. Results In individuals with type 2 diabetes, the mean CNFL was significantly reduced [β = −1.86 mm/mm2 (95% CI −3.64 to −0.08), p = 0.04], as compared with individuals with normal glucose metabolism after adjustment for age and sex. Part of the reduction was present in individuals with prediabetes [β = −0.96 mm/mm2 (95% CI −2.91 to 0.99), p = 0.34], with a linear trend of corneal nerve fibre reduction with severity of glucose metabolism status (p trend = 0.04). Conclusions A significant reduction in CNFL was found in individuals with DM2 compared with individuals with NGM. A trend of reduction in CNFL was observed between individuals with NGM and prediabetes. The reduction in corneal nerve fibre length could contribute to a delayed corneal healing and an increased risk for corneal complications after surgery.


Introduction
The cornea is one of the body's most densely innervated tissues (Millodot 1984). Corneal nerves have important protective and trophic influences on the cornea . Under physiologic conditions, they play an important role in corneal epithelial cell metabolism, cell adhesion and wound healing in response to infection, trauma and surgery (Beuerman & Schimmelpfennig 1980;Linna et al. 1998;Muller et al. 2003;Gallar et al. 2004). Although one of the major ocular complications of diabetes is retinopathy, structural changes in the subbasal nerve plexus of the cornea are also observed (Rosenberg et al. 2000;Malik et al. 2003;Efron 2011). In individuals with diabetes, injury of the corneal nerve fibres results in a reduced tear secretion (Dogru et al. 2001;Saito et al. 2003;Yoon et al. 2004;Cousen et al. 2007), a delayed corneal epithelial healing (Kabosova et al. 2003) and a decreased corneal sensitivity (Nielsen & Lund 1979;Murphy et al. 2004;Tavakoli et al. 2007). Consequently, individuals with diabetes have a higher vulnerability for keratopathy (Schultz et al. 1983;Herse 1988;Ohashi 1997;Didenko et al. 1999), recurrent corneal erosions, persistent epithelial defects and neurotrophic corneal ulcers (Hyndiuk et al. 1977;Sanchez-Thorin 1998). Since damage to the corneal nerves occurs in ocular and systemic diseases, and after corneal surgery (Patel & McGhee 2009), structural changes in the corneal plexus in individuals with glucose dysregulation can further jeopardize corneal epithelial healing (Beuerman & Schimmelpfennig 1980). In addition, corneal neurodegenerative changes in individuals with diabetes are correlated with diabetic polyneuropathy (Tavakoli et al. 2010;Dehghani et al. 2014;Pritchard et al. 2015). As signs of neuronal dysfunction are not only found in individuals with diabetes, but also found in individuals with prediabetes (Papanas et al. 2011;Bongaerts et al. 2013;, neurodegenerative changes in the cornea could also be present before diabetes is clinically diagnosed (Asghar et al. 2014).
Corneal confocal microscopy (CCM) can accurately quantify the nerve fibre length in the subbasal nerve plexus of the cornea (Efron et al. 2010;Hertz et al. 2011;Petropoulos et al. 2013). Corneal confocal microscopy has a proven utility in detecting and monitoring neurodegenerative changes in individuals with diabetes (De Clerck et al. 2015). However, previous studies mainly included individuals with long-duration type 2 diabetes (DM2). Since duration of diabetes is related to severity of diabetic neuropathy (Dyck et al. 1993), it would be of interest to assess changes in corneal nerve fibre length per mm 2 (CNFL) in individuals with prediabetes. Establishing a reduction in CNFL in individuals with type 2 diabetes or prediabetes using CCM could help to identify individuals at risk for a delayed corneal healing (e.g. after routine cataract surgery), neurotrophic corneal ulcers and polyneuropathy at an earlier stage.
The aim of the present study was to investigate to what extent changes in CNFL are present in individuals with prediabetes and individuals with DM2, as compared to individuals with normal glucose metabolism (NGM).

Study population and design
We used data from The Maastricht Study, an observational prospective population-based cohort study. The rationale and methodology have been described previously (Schram et al. 2014). In brief, the study focuses on the aetiology, pathophysiology, complications and comorbidities of DM2 and is characterized by an extensive phenotyping approach. Eligible for participation were all individuals aged between 40 and 75 years and living in the southern part of the Netherlands. Participants were recruited through mass media campaigns and from the municipal registries and the regional Diabetes Patient Registry via mailings. Recruitment was stratified according to known DM2 status, with an oversampling of individuals with DM2, for reasons of efficiency. Cross-sectional data of The Maastricht Study are available from the participants who completed the baseline survey between November 2010 and September 2013. From April 2013, corneal confocal microscopy measurements were included in the measurement protocol. The present report includes cross-sectional data from all participants who completed the baseline survey between April 2013 and September 2013. The examinations of each participant were performed within a time window of 3 months. The study has been approved by the institutional medical ethical committee (NL31329.068.10) and the Minister of Health, Welfare and Sports of the Netherlands (Permit 131088-105234-PG). All participants gave written informed consent.

Glucose metabolism status
To determine glucose metabolism, all participants, except those who used insulin, underwent a standardized 2-h 75 g oral glucose tolerance test (OGTT) after an overnight fast. For safety reasons, participants with a fasting glucose level above 11.0 mmol/L, as determined by a finger prick, did not undergo the OGTT. For these individuals (n = 13), fasting glucose level and information about diabetes medication were used to determine glucose metabolism status. Glucose metabolism was defined according to the WHO criteria into NGM, impaired fasting glucose (IFG), impaired glucose tolerance (IGT), prediabetes (i.e. IFG and/or IGT), and DM2 (World Health Organization 2006). For this study, individuals with DM1, individuals with latent autoimmune diabetes of adults, steroid-induced diabetes and individuals who underwent a pancreas transplantation were excluded.

Ophthalmologic measurements
Corneal confocal microscopy (Heidelberg Retina Tomograph III, Rostock cornea module, Heidelberg Engineering, Heidelberg, Germany) was performed on the left eye to study the subbasal nerve plexus. Both eyes were anaesthetized with oxybuprocaine hydrochloride 0.4%. Both corneas were wetted with gel to prevent dry eyes and to ensure optimal contact between the cornea and the applanating cap. Participants were asked to fixate on a red light throughout the scan. The cornea was lightly touched with a TomoCap (Heidelberg Engineering) filled with gel. A charge-coupled device camera was used to image the cornea from the side to check the correct position of the cap on the apex of the cornea. A joystick was used to identify the subbasal nerve plexus, located between the basal cell layer of the epithelium and the Bowman's layer. Images were acquired in the central part of the cornea by trained research assistants according to a standard operating procedure. The recorded images were composites of multiple smaller recordings of 400 9 400 lm (384 9 384 pixels, 8 bit) assembled together by the use of a composite algorithm implemented in the HRT3 user interface (Heidelberg Engineering, Heidelberg, Germany), as previously described (Allgeier et al. 2011;Allgeier et al. 2014). Real-time mapping was performed on an area up to 1600 9 1600 µm (1536 9 1536 pixels, 8 bit), partially including the inferior whorl in some of the composite images. No measures were taken to include or exclude this region.
One two-dimensional image was acquired in each participant. The examination lasted approximately 5 min. Corneal confocal microscopy measurements were not performed in individuals with a corneal infection. All images were reviewed individually, and their quality was scored based on the contrast, the depth and the sharpness of the picture and based on the presence of epithelial cells, Langerhans cells or pressure lines using a designated protocol. The area of the composite image was automatically assessed, the nerves in each image were automatically segmented, and the CNFL, defined as the total length of all nerve fibres with respect to the image area and expressed in mm per mm 2 , was measured by the use of a custom-made software with a performance level comparable to a human grader (Guimaraes et al. 2016). Reproducibility was assessed by two observers in the left eye of six individuals (3 men, 64.5 AE 14.7 years; 1 individual with DM2) who were examined on two occasions with a 1-week interval. The intra-observer intra-class correlation coefficients were ≥0.89, and the inter-observer intra-class correlation coefficients were ≥0.97.
The presence of corneal disorders, previous surgery and the use of contact lenses were assessed by questionnaire. In addition, the presence of an intraocular lens was assessed on the Scheimpflug image of the anterior segment (Oculus Pentacam HR, Wetzlar, Germany). Participants were requested to bring all medication they used or a list from their pharmacists to the research centre. Dry eye medication use and medication for corneal infection were assessed during a medication interview where generic name, dose and frequency were registered by trained staff (Schram et al. 2014). Individuals with a small investigated area (<0.50 mm 2 ), poor quality imaging, corneal disorders, contact lenses, previous cataract or refractive surgery, and medication for dry eyes or corneal infection were excluded.

Statistical analysis
Statistical analysis was performed in SPSS Statistics 23 for Windows; SPSS, Chicago, IL. Differences between group characteristics were tested using one-way analysis of variance (ANOVA) with post hoc testing by the least significant difference (LSD) method for continuous variables and chisquare tests for categorical variables. Multivariable linear regression was used to analyse the association between glucose metabolism status (prediabetes and DM2; determinant) and CNFL (outcome). We combined the categories impaired fasting glucose (IFG) and impaired glucose tolerance (IGT) into prediabetes, because analyses did not show differences between IFG and IGT (data not shown). First, a crude analysis was performed. Next, associations were adjusted for age and sex. The results were expressed as regression coefficients (b), representing the mean difference in CNFL as compared with NGM, with their 95% confidence intervals (95% CIs) and p-values. The Wilcoxon-Mann-Whitney 2-tailed test was used for statistical power calculation to compare the CNFL in individuals with prediabetes versus individuals with NGM. Due to insufficient group size, the statistical power of 80% was not achieved. Figure 1 shows the flow diagram of the study. A total of 215 consecutive participants had a corneal confocal microscopy measurement between April 2013 and September 2013. One participant with DM1 was excluded. Participants with a small investigated area of the subbasal nerve plexus were also excluded (n = 22). We additionally excluded images of unsatisfactory quality (n = 27)that is unsatisfactory contrast (n = 5) or depth (n = 16), and/ or the presence of epithelial cells (n = 9), disruptive Langerhans cells (n = 3) or pressure lines (n = 11). In addition, individuals with corneal disorders (n = 1), contact lenses (n = 17), previous cataract or refractive surgery (n = 7), and/or medication for dry eyes or corneal infection (n = 4) were also excluded. We additionally excluded individuals in whom data on ocular disorders, previous surgery or use of contact lenses were missing (n = 2). Thus, 134 participants were available for statistical analysis. Participants who were excluded due to missing values did not differ significantly from included participants in terms of age or sex.

Results
General characteristics and the CNFL as assessed by CCM are shown in Table 1, stratified by glucose metabolism status. Of the 134 participants, 87 participants had NGM (64.9%), 20 participants had prediabetes (14.9%), and 27 participants had DM2 (20.1%). There was a statistically significant difference in age between individuals with DM2 and individuals with NGM (p < 0.01). In individuals with NGM, prediabetes and DM2, the crude CNFL was 13.4 AE 4.01 mm/mm 2 [95% CI 12.5-14.3], 12.4 AE 3.02 mm/ mm 2 [95% CI 11.0-13.9], and 11.5 AE 4.11 mm/mm 2 [95% CI 9.9-13.1], respectively (Fig. 2). The CNFL value was significantly lower in individuals with DM2 compared with individuals with NGM (p = 0.03), and a major part of this reduction was also present in prediabetes (p = 0.33). The mean investigated area of the composite images was 1.58 AE 0.66 mm 2 in individuals with NGM, 1.78 AE 0.73 mm 2 in individuals with prediabetes and 1.44 AE 0.56 mm 2 in individuals with DM2, and was not significantly different between these groups (p = 0.22). Figure 3 shows crude CNFL values according to glucose metabolism status. Reduction in CNFL was more pronounced with worsening of glucose metabolism status, with a significant pvalue for linear trend (p = 0.03).

Discussion
In the current study, we showed a significant reduction in CNFL between individuals with DM2 compared with individuals with NGM using composite CCM images. Our findings suggest a trend of reduction in CNFL from NGM to prediabetes.
To our knowledge, this is the first study that uses composite CCM images to assess the association between glucose metabolism status, as defined by an oral glucose tolerance test, and CNFL. This innovative imaging technique based on real-time mapping of several images allows visualization of a larger continuous surface of the corneal subbasal nerve plexus, increasing the level of accuracy of quantification of CNFL (Patel & McGhee 2005;Zhivov et al. 2010;    Vagenas et al. 2012;Lagali et al. 2018). The method used for the assessment of the CNFL in these images showed a high reproducibility. Standard deviations were in line with those reported in previous studies.
Larger study populations are needed to confirm our findings.
In individuals with NGM, the mean CNFL was 13.4 mm/mm 2 , which is in line with previous studies reporting a mean CNFL in the range of 11.90-24.30 mm/mm 2 (Hertz et al. 2011;Ahmed et al. 2012;Nitoda et al. 2012;Sivaskandarajah et al. 2013;Bitirgen et al. 2014;Dehghani et al. 2014;Pritchard et al. 2014;Stem et al. 2014;Lagali et al. 2017). In individuals with prediabetes, the mean CNFL was 12.4 mm/mm 2 . At present, only three studies assessed corneal neurodegenerative variables in individuals with prediabetes, reporting a mean CNFL of 14.6, 15.3 and 22.1 mm/mm 2 (Asghar et al. 2014;Azmi et al. 2015;Lagali et al. 2017, respectively). In individuals with DM2 in our study, the mean CNFL was 11.5 mm/mm 2 , which is in line with previous studies reporting a mean CNFL in the range of 9.80-21.90 mm/ mm 2 (Nitoda et al. 2012;Bitirgen et al. 2014;Stem et al. 2014;Lagali et al. 2017). The large range in CNFL in previous studies could be explained by heterogeneity in terms of type of in vivo confocal microscope, scan protocol, image selection and software for analysis of the CNFL. In our study, the CNFL in DM2 was 14% reduced compared with individuals with NGM. This is in line with the findings of previous studies, showing a reduction by 8-18% without correction for age and sex (Nitoda et al. 2012;Bitirgen et al. 2014;Stem et al. 2014;Lagali et al. 2017). However, these previous studies mainly included individuals with long-duration DM2. Our study showed no significant influence of age on the CNFL (À0.03 mm/mm 2 per year, p = 0.54), which is similar to the findings of the LANDMARK study (À0.05 mm/mm 2 per year, p > 0.05) . Another study found a significant negative correlation between age and CNFL, showing a mean decline of 0.25-0.30% per year, which agrees with the reduction of 0.22% per year found in our study (Parissi et al. 2013). Similar to previous studies, the CNFL was not significantly different between men and women Parissi et al. 2013).
Diabetes induces alterations of the corneal subbasal nerve plexus, such as a reduction in the CNFL assessed by CCM. However, the exact pathogenic mechanisms underlying the diabetesassociated reduction in CNFL are not clear. Corneal neuropathy has been suggested to be caused by paracrine signals reaching the cornea via the limbal capillaries (Leppin et al. 2014). Accumulation of advanced glycation end products (Reichard 2012), impaired endothelium-dependent vasodilation (Davidson et al. 2012), altered release of neuropeptides (Leppin et al. 2014) and downregulation of neurotrophic factors , could play a role in the pathogenesis of corneal neuropathy. This downregulation of neurotrophic factors may result in an impaired epithelial healing capacity in individuals with diabetes (You et al. 2000;Cursiefen et al. 2005).
Delayed corneal epithelial wound healing after subbasal nerve plexus injuries following refractive surgery, corneal transplantation, cataract surgery and vitrectomy leads to an increased incidence of postoperative epithelial complications and poorer refractive results (Foulks et al. 1979;Fraunfelder & Rich 2002). Postoperative regeneration of the subbasal nerves is a long-lasting process, which generally does not recover to normal, even in individuals with NGM. After laserassisted in situ keratomileusis (LASIK) (Erie et al. 2005), laser epithelial keratomileusis (LASEK) (Darwish et al. 2007) and photorefractive keratectomy (PRK) (Moilanen et al. 2003;Erie et al. 2005) healing periods up to five years have been reported, and after penetrating keratoplasty (PKP) (Niederer et al. 2007), structural changes in the subbasal nerves remain visible even 40 years after surgery. In addition, structural changes in the subbasal nerve plexus have been observed 6 months after phacoemulsification with intraocular lens insertion (Misra et al. 2015). Finally, a delayed reinnervation after corneal epithelial erosions has been observed in DM2 (Wang et al. 2012).
A trend of reduction in CNFL is now observed from NGM to prediabetes. Because a reduction in CNFL is associated with a delayed corneal epithelial healing (Kabosova et al. 2003) and a predisposition for corneal ulcers and other corneal complications, in particular after surgery (Hyndiuk et al. 1977;Sanchez-Thorin 1998;Patel & McGhee 2009), we could suppose that this group of individuals may also be at risk. Since diabetes is a contraindication for corneal laser refractive surgery (Netherlands Ophthalmologic Society 2013), one could hypothesize that individuals with prediabetes also present a relative systemic contra- indication for ocular surgery. Early diagnosis of prediabetes by the oral glucose tolerance test and the assessment of corneal neurodegenerative changes by corneal confocal microscopy may define patients at risk for corneal neuropathy. In patients with corneal neuropathy, the risk for postoperative epithelial complications could be reduced by increased lubrication and management of meibomian gland dysfunction (Krachmer 2011;Weisenthal et al. 2017). In addition, surgical epithelial debridement should be minimized and toxic medications should be avoided. Furthermore, the early observation of CNFL reduction, which is an ophthalmic marker of polyneuropathy (Ahmed et al. 2012;Sivaskandarajah et al. 2013;Dehghani et al. 2014;Pritchard et al. 2015), is in line with previous studies demonstrating neuropathic symptoms (Isak et al. 2008;Ziegler et al. 2009;Asghar et al. 2014), abnormal (semi-) quantitative sensory tests (Franklin et al. 1990;Asghar et al. 2014) and reduction in intraepidermal nerve fibre density in skin biopsy samples in individuals with prediabetes (Smith et al. 2006;Asghar et al. 2014;Azmi et al. 2015).
There are several limitations of the present study. First, the CNFL is currently the only variable calculated by our software. However, the CNFL is known to be the corneal neuronal variable with the best reproducibility compared with other corneal neuronal variables (Efron et al. 2010;Hertz et al. 2011;Petropoulos et al. 2013). Second, the cross-sectional design does not allow to address causal relationships. However, by comparing individuals with NGM, prediabetes and DM2, we mimic the pathological pathway of glucose metabolism deterioration. Third, we did not perform a slit-lamp examination and did not assess ocular sensitivity in our study. Future longitudinal studies should focus on causality and underlying mechanisms of the decrease in CNFL in prediabetes and DM2 and its predictive value for corneal pathology.
In conclusion, this study demonstrates that a significant reduction in CNFL is found in individuals with DM2 compared with individuals with NGM. A trend of reduction is observed between individuals with NGM and prediabetes. This may result in a delayed corneal healing and an increased risk for corneal complications after surgery.