Corneal Confocal Microscopy Predicts Cardiovascular and Cerebrovascular Events and Demonstrates Greater Peripheral Neuropathy in Patients with Type 1 Diabetes and Foot Ulcers

Objective: In this study, we evaluate small and large nerve fibre pathology in relation to diabetic foot ulceration (DFU) and incident cardiovascular and cerebrovascular events in type 1 diabetes (T1D). Methods: A prospective observational study was conducted on people with T1D without diabetic peripheral neuropathy (DPN) (n = 25), T1D with DPN (n = 28), T1D with DFU (n = 25) and 32 healthy volunteers. ROC analysis of parameters was conducted to diagnose DPN and DFU, and multivariate Cox regression analysis was performed to evaluate the predictive ability of corneal nerves for cardiac and cerebrovascular events over 3 years. Results: Corneal nerve fibre length (CNFL), fibre density (CNFD) and branch density (CNBD) were lower in T1D-DPN and T1D-DFU vs. T1D (all p < 0.001). In ROC analysis, CNFD (sensitivity 88%, specificity 87%; AUC 0.93; p < 0.001; optimal cut-off 7.35 no/mm2) and CNFL (sensitivity 76%, specificity 77%; AUC 0.90; p < 0.001; optimal cut-off 7.01 mm/mm2) had good ability to differentiate T1D with and without DFU. Incident cardiovascular events (p < 0.001) and cerebrovascular events (p < 0.001) were significantly higher in T1D-DPN and T1D-DFU. Corneal nerve loss, specifically CNFD predicted incident cardiovascular (HR 1.67, 95% CI 1.12 to 2.50, p = 0.01) and cerebrovascular (HR 1.55, 95% CI 1.06 to 2.26, p = 0.02) events. Conclusions: Our study provides threshold values for corneal nerve fibre metrics for neuropathic foot at risk of DFU and further demonstrates that lower CNFD predicts incident cardiovascular and cerebrovascular events in T1D.


Introduction
Globally, 537 million adults are living with diabetes, which is a prevalence of 10.5% with a projected rise to 12.2% by 2045 [1].Diabetic foot ulcers (DFUs) are a major cause of morbidity, mortality and health care expenditure due to hospitalization.The incidence of cardiovascular events is increased in people with DFU, especially those with non-healing or recurrent ulceration [2], with up to a two-fold increase in mortality [3].
Small nerve fibre damage precedes large nerve fibre damage in diabetic peripheral neuropathy (DPN) [4].However, there remains a paucity of data on the natural history and extent of small nerve fibre damage in relation to DFU and there is a lack of consensus as to which measure of DPN best identifies individuals at risk of DFU.Several tests can detect DPN, including vibration perception threshold (VPT), thermal thresholds and nerve conduction studies.Whilst VPT detects DPN and predicts DFU [5], thermal thresholds poorly differentiate patients with and without DFU [6].The most utilized bedside tests are the 10 g monofilament and 128 Hz tuning fork, both of which identify end-stage neuropathy and those at high risk of DFU.Corneal confocal microscopy (CCM) is a noninvasive ophthalmic test which quantifies small nerve fibre degeneration with reliable and accurate diagnostic ability for DPN [7][8][9].Previous longitudinal studies have demonstrated the association between corneal nerve fibre loss and progressive large fibre dysfunction [10], suggesting that CCM may predict the development of DFU [11].Hyperglycaemia is a well-established independent risk factor for cardiovascular disease.However, this risk may be mitigated in people with type 1 diabetes (T1D) with good glycaemic control [12].Whilst observational studies suggest that a history of DFU increases the risk of cardiovascular disease [13], there is a paucity of evidence on the association between cardiovascular disease and neuropathy.Specifically, there is a lack of high-quality studies showing that CCM-based quantification of small nerve fibre damage is associated with cardiovascular disease.
In this paper, we aimed to determine the diagnostic utility of CCM for DFU and its predictive ability for incident cardiovascular and cerebrovascular events and mortality in patients with T1D.

Study Subjects
This was a prospective observational study of people with T1D attending diabetes outpatient clinics in secondary care.The study was approved by the Preston Research Ethics Committee (REC 18/NW/0532) and sponsored by the University of Liverpool, UK.Written informed consent was obtained according to the Declaration of Helsinki.Participants underwent assessment of the neuropathy symptom profile, neurological examination and nerve conduction studies to determine the presence of neuropathy and active foot ulceration alongside age-and sex-matched healthy volunteers without diabetes.Peripheral neuropathy was identified according to the Toronto criteria [14], and participants in the DFU group had at least one confirmed active diabetic foot ulcer.Participants with other causes of neuropathy (except diabetes), history of neurological disease or previous ocular trauma or ocular surgery were excluded.

Clinical Assessments
All participants completed a study questionnaire including detailed past medical history and current medications.Participants underwent anthropometric (height, weight and body mass index), and standard clinical care biochemistry as part of standard practice; glycated haemoglobin (HbA 1c ), total cholesterol, HDL cholesterol, triglycerides, renal profile, and urine albumin-to-creatinine ratio (ACR) assessments.

Assessment of Micro-and Macrovascular Complications
Screening of both the primary and secondary care health records was undertaken.Retinopathy and maculopathy were identified from the Diabetic Retinopathy National Screening records.Nephropathy was defined according to the estimated glomerular filtration rate (eGFR) and urine albumin-creatinine ratio.Cardiovascular (CV) events were defined based on a history of physician-diagnosed myocardial infarction, angina, coronary artery disease, percutaneous coronary angiography, or coronary artery bypass grafting.
Cerebrovascular accidents (CVA) were defined by any new occurrence of ischaemic or haemorrhagic stroke.Mortality data were recorded from the medical records.Macrovascular outcomes were analysed over 3.0 ± 0.6 years of follow-up in all the participants with T1D.

Assessment of Neuropathy
Peripheral neuropathy was defined according to the Toronto criteria [14] by the presence of an abnormal nerve conduction study and symptom/symptoms and/or sign/signs of peripheral neuropathy.All participants underwent evaluation of neurologic symptoms according to the neuropathy symptom profile (NSP), and the McGill visual analogue score (VAS) was used to assess the severity of painful neuropathy.Neurologic deficits were assessed using the modified neuropathy disability score (NDS), which includes the evaluation of vibration, pin prick, temperature perception and ankle reflexes.Quantitative sensory testing (QST) was performed to determine the cold threshold (CT) (Aδ fibres), warm threshold (WT) (C fibres), cold (CIP) and warm-induced pain (WIP) thresholds using the method of limits with the MEDOC TSA II (Medoc, Ramat Yishai, Israel) on the dorsum of the left foot.Vibration perception threshold (VPT) was measured from an average of three values on the large toe using a neurothesiometer (Horwell, Scientific Laboratory Supplies, Wilford, Nottingham, UK).Participants with a VPT > 15-24 V were considered to have DPN, whilst those with a VPT ≥ 25 V were deemed at high risk for DFU [14].The sural nerve conduction velocity (SNCV) and sural nerve amplitude (SNAP) were evaluated using the NC-Stat ® DPN Check system (Neurometrix, Waltham, USA).If SNCV and SNAP were unrecordable, the lowest possible recordable value detectable with NC-Stat ® DPN Check system was used (SNCV reading of 28 m/s and SNAP reading of 1.5 µV).

Corneal Confocal Microscopy
All participants underwent CCM for which a laser scanning corneal confocal microscope (Heidelberg Retina Tomograph III; Heidelberg Engineering, Heidelberg, Germany) was used.Several scans of the entire depth of the cornea were recorded using the section mode to acquire and store two-dimensional images with a final resolution of ~2 µm/pixel and an image size of 400 × 400 pixels of sub-basal nerve plexus of the cornea.Eight images (four images from each eye per participant) from the central cornea were selected and examined in a masked and randomised fashion.Automated corneal nerve fibre quantification (ACCMetrics software, version 2.0, University of Manchester, Manchester, UK) was undertaken to derive (1) corneal nerve fibre density (CNFD), number of main nerves/mm 2 of corneal tissue; (2) corneal nerve branch density (CNBD), number of primary nerve branches/mm 2 ); and (3) corneal nerve fibre length (CNFL), length of main nerves and nerve branches (mm/mm 2 ) from the eight central cornea nerve images per participant.

Power Calculation
Corneal nerve fibre density was selected as the primary corneal nerve outcome to assess small nerve fibre structure.With an SD between groups of 9 nerves/mm 2 , we estimated that a minimum of 25 participants for each group would provide an 80% chance to detect a clinically meaningful difference in the CNFD of 5 nerves/mm 2 and an assumption of a type 1 error (alpha-level) of 0.05.

Statistical Analysis
Normally distributed data were expressed as mean (standard deviation) (SD).Nonnormally distributed data were expressed as median and interquartile range (IQR).Correlations were undertaken using Pearson's test for normally distributed and Spearman's rank test for non-normally distributed data.The ANOVA method or a non-parametric counterpart, Kruskal-Wallis, was used to assess differences between groups depending on the normality of the data.Overall, the p value was maintained at 0.05 for multiple comparison tests (Bonferroni adjustment or non-parametric counterpart-0.05/4).
Chi squared tests (r by c Chi squared tests) were undertaken to examine the differences between the microvascular complications at baseline, and incident cardiovascular and cerebrovascular events during the follow-up period.We performed the multivariate Cox regression analysis for neuropathy parameters to evaluate the contribution of each of the categorical and continuous parameter for predicting incident cardiovascular and cerebrovascular events.All selected variables that were predictive, including small and large fibre tests, were entered into the multivariate Cox regression analysis to ascertain the impact on incident cardiovascular and cerebrovascular events.The regression coefficients, hazard ratios, and their corresponding upper and lower 95% confidence interval (CI) were estimated.Receiver operating characteristic (ROC) curve analyses were used to define the Wilcoxon estimate of an area under ROC curve, optimal cut-offs with associated sensitivity and specificity for CCM parameters and small and large fibre tests to identify those with DFU.Statistical analyses were undertaken on SPSS Statistics 25 (IBM Corporation, Armonk, NY, USA).

Predictors of Cardiovascular and Cerebrovascular Events in Patients with T1D
Over a 3-year follow-up, the incidence of new cardiovascular events (p < 0.001), cerebrovascular events (p < 0.001) and lower extremity amputation (p < 0.001) were higher in T1D-DPN and T1D-DFU (Table 4).There was no difference in the 3-year mortality rates (p = 0.25) between groups.In the Cox-based regression analysis, we evaluated the key parameters and characteristics of T1D participants (without DPN, with DPN and DFU) for developing new incident cardiovascular and cerebrovascular events, respectively (Supplementary Table S2a,b).In the multivariate Cox regression analysis, decreased CNFD predicted greater incident cardiovascular (HR 1.67, 95% CI 1.12-2.50,p = 0.01) and cerebrovascular (HR 1.55, 95% CI 1.06-2.26,p = 0.02) events (Table 5).α -increased/higher value was associated with increased incidence of cardiovascular and cerebrovascular events.β -decreased/lower value was associated with increased incidence of cardiovascular and cerebrovascular events.

Discussion
In this study, we report greater small and large nerve fibre damage in T1D patients with DFU, with a clear association between corneal nerve fibre loss and the severity of DPN.We further demonstrate for the first time that corneal nerve loss predicts incident cardiovascular and cerebrovascular events, more so than age, duration of diabetes, dyslipidaemia, and other measures of neuropathy, including nerve conduction studies and VPT.Our data suggest that degeneration of corneal nerve fibres may serve as a surrogate biomarker for detecting the at-risk diabetic foot and higher incidence of cardiovascular and cerebrovascular events, though the mechanistic association remains to be established.
Both large and small nerve fibre degeneration occurs in DPN [4,6], though small fibres are affected earlier than larger fibres [15].Neuropathy and insensitivity to trauma is key to the development of DFU [11,16].Indeed, reduced motor nerve conduction velocity has been shown to predict foot ulceration and increased mortality in diabetes [17] and increased VPT has a good predictive value for the development of DFU [14].Small nerve fibre damage with impaired heat and pain sensation exposes patients to unperceived foot trauma [18] and increased thermal thresholds are associated with an increased likelihood of DFU [19].Therefore, sensory and autonomic neuropathy independently influence the risk of DFU and amputation [20,21].However, impaired small fibre-mediated pressure-induced vasodilation, the hyperaemic response to tissue injury [19,22] and deficiency in nerve growth factor and substance P are key mediators of skin breakdown and blunted healing of DFU [21,23,24].The evaluation of small fibre neuropathy has relied on quantitative sensory testing (QST) and contact heat-evoked potential tests, but they are time consuming and subject to large intra-individual variability.Skin biopsy and assessment of intraepidermal nerve fibre density (IENFD) allows to conduct a direct measure of small fibre pathology in DPN, but it is an invasive technique [25].We have previously shown that CNFD and IENFD have comparable diagnostic utility for DPN [26].This study now demonstrates that CNFD, CNBD and CNFL have good diagnostic utility for patients with DFU, consistent with our data showing that a reduction in CNFL preceded the development of DFU [11].A recent meta-analysis has confirmed that CCM has good diagnostic utility for DPN [9].Furthermore, lower CNFL predicts 4-year incident DPN [27] and a more rapid CNFL decline is associated with the development of DPN [28] and foot ulceration [11,16].We now show that CNFL is a superior diagnostic biomarker for DPN (Sn 0.96, Sp 0.92; AUC 0.98; (95% CI) (0.96-1.00); p < 0.001), whilst CNFD is superior for DFU (Sn 0.88, Sp 0.87; AUC 0.93; (95% CI) (0.88-0.98); p < 0.001).This suggests an initial length dependent process followed by more global proximal nerve fibre loss with more severe DPN, reflecting a similar process in intra-epidermal nerve fibres.
Inflammation and premature atherosclerosis remain the predominant cause of excess mortality in T1D.Impaired flow-mediated dilation (FMD) and reactive hyperaemia peripheral artery tonometry (RH-PAT) are associated with small fibre denervation and dysfunction.Small fibre deficits are also associated with a higher incidence of CAN [29] with an increased risk of myocardial dysfunction, silent myocardial ischaemia and cardiac arrhythmias [29].In our previous studies we have shown greater corneal nerve loss in patients with acute ischemic stroke [15], especially those with recurrent stroke [30].In the Canadian cohort study of people living with T1D for over 50 years, greater coronary artery calcification [31] was associated with large fibre neuropathy and retinopathy, suggesting common inflammatory pathways.The mechanisms underpinning the association between small nerve fibre degeneration and atherosclerotic cardiovascular disease remain unclear.Atherosclerotic plaque inflammation leads to the transmission of impulses via sensory afferent fibres to the medullary and hypothalamic neurones, thus increases sympathetic efferent activity in the vessel wall (artery-brain circuit), resulting in increased lymphocyte and cytokine activity and inflammation [32].Ganglionectomy in experimental models has been shown to attenuate this neuroimmune activation by the peripheral nervous system and reduce disease progression [32].
Limitations of our study include the potential influence of confounding factors such as the long duration of diabetes, predisposing to increased risk of DFU and cardiovascular disease.We also investigated relatively small numbers of patients with minor disparities in age between the cohorts.Studies to assess the utility of CCM in the prediction of cardiovascular disease and recurrent DFU are warranted.Further studies involving the use of artificial intelligence (AI)-based algorithms and end-to-end classification of CCM images [7] may strengthen the utility of CCM.Recently, machine learning algorithm was used to develop systems based on several discriminative patient parameters, helping identify patients at high risk of cardiovascular events [33].
In summary, we show that small nerve fibre damage is prevalent in patients with DFU and present diagnostic cut-offs for CCM to identify the at-risk neuropathic foot.Furthermore, we show that corneal nerve fibre density predicts incident cardiovascular and cerebrovascular events in patients with T1D.

Supplementary Materials:
The following supporting information can be downloaded at: https:// www.mdpi.com/article/10.3390/diagnostics13172793/s1, Figure S1: ROC curves for CNFD, CNBD, CNFL, VPT, CT and WT for T1D-DPN; Table S1: ROC analysis with area under the curve, optimal cut off and respective sensitivity and specificity with 95% confidence interval in T1D without DPN versus T1D-DPN for CNFD, CNBD, CNFL, VPT, CT and WT, Table S2a: Univariate associations with incident cardiovascular events over 3 years in patients with T1D based on individual parameters with Cox regression analysis, Table S2b: Univariate associations with incident cerebrovascular events over 3 years in patients with T1D based on individual parameters with Cox regression analysis.

Figure 1 .
Figure 1.Corneal confocal microscopy images of the sub-basal nerve plexus showing a progressive loss of corneal nerves in a patient with T1D (top right), T1D-DPN (bottom left) and T1D-DFU (bottom right) compared to a healthy control (top left).T1D-type 1 diabetes; T1D-DPN-type 1 diabetes with diabetic peripheral neuropathy; T1D-DFU-type 1 diabetes with diabetic foot ulcer.

Table 1 .
Demographics and metabolic profile of the participants.

Table 2 .
Neuropathy symptoms, neurological deficits, quantitative sensory testing, neurophysiology and corneal confocal microscopy in all participants.

Table 3 .
ROC analysis with AUC, optimal cut off, sensitivity and specificity with 95% confidence interval in T1D with and without DFU.

Table 4 .
Morbidity and mortality outcomes in participants with T1D.

Table 5 .
Associations with incident cardiovascular and cerebrovascular events over 3 years in patients with T1D based on the multivariate Cox regression analysis.