Corneal Confocal Microscopy Identifies People with Type 1 Diabetes with More Rapid Corneal Nerve Fibre Loss and Progression of Neuropathy

There is a need to accurately identify patients with diabetes at higher risk of developing and progressing diabetic peripheral neuropathy (DPN). Fifty subjects with Type 1 Diabetes Mellitus (T1DM) and sixteen age matched healthy controls underwent detailed neuropathy assessments including symptoms, signs, quantitative sensory testing (QST), nerve conduction studies (NCS), intra epidermal nerve fiber density (IENFD) and corneal confocal microscopy (CCM) at baseline and after 2 years of follow-up. Overall, people with type 1 diabetes mellitus showed no significant change in HbA1c, blood pressure, lipids or neuropathic symptoms, signs, QST, neurophysiology, IENFD and CCM over 2 years. However, a sub-group ( n = 11, 22%) referred to as progressors, demonstrated rapid corneal nerve fiber loss (RCNFL) with a reduction in corneal nerve fiber density (CNFD) ( p = 0.0006), branch density (CNBD) ( p = 0.0002), fiber length (CNFL) ( p = 0.0002) and sural ( p = 0.04) and peroneal ( p = 0.05) nerve conduction velocities, which was not related to a change in HbA1c or cardiovascular risk factors. The majority of people with T1DM and good risk factor control do not show worsening of neuropathy over 2 years. However, CCM identifies a sub-group of people with T1DM who show a more rapid decline in corneal nerve fibers and nerve conduction velocity.

2.1. Selection of Patients Patients with T1DM ( n = 50) and age and sex matched non-diabetic healthy control participants (C) ( n = 16) were assessed at baseline and at two years follow-up. Participants with other causes of peripheral neuropathy excluded by a medical history and laboratory panel, e.g., vitamin B12, folate, autoimmune antibodies, TFTs, renal profile, etc., and those with a history of ocular trauma or previous ocular surgery were excluded from the study. This study was observational with no active intervention. The study was approved by the North Manchester Research Ethics committee (Ref: 09/H1006/38; Date: July 2009), and written informed consent was obtained according to the Declaration of Helsinki.

All participants underwent a detailed evaluation of neuropathic symptoms using the neuropathy symptom profile (NSP) and the McGill VAS to assess the severity of painful neuropathy. Clinical neurologic deficits were assessed using the modified neuropathy disability score (NDS) [ 26 ]. Diabetic neuropathy was defined according to the Toronto criteria by the presence of an abnormality of nerve conduction and a symptom or symptoms or a sign or signs of neuropathy [ 27 ].

NCS were undertaken using a Dantec “Keypoint” system (Dantec Dynamics, Bristol, UK) equipped with a DISA temperature regulator to keep limb temperature constantly between 32 °C and 35 °C. Peroneal motor and sural sensory nerves were assessed in the right lower limb by a consultant neurophysiologist. The motor study was performed using silver–silver chloride surface electrodes at standardized sites defined by anatomical landmarks, and recordings for the sural nerve were taken using antidromic stimulation over a distance of 100 mm.

Participants underwent examination with the Heidelberg retina tomography III in vivo corneal confocal microscope employing our established methodology for image acquisition [ 29 ]. A total of 6 CCM images per participant (3 per eye) from the sub-basal nerve plexus in the central cornea (apex) were captured in a masked fashion with a lateral resolution of ~2 mm/pixel and final image size of 400 × 400 pixels. Three corneal nerve parameters were quantified: (1) corneal nerve fiber density (CNFD) (number of main nerve fibers (no./mm 2 )); (2) branch density (CNBD) (number of nerve branches (no./mm 2 )); (3) CNFL (length of nerve fibers (mm/mm 2 )). Automated corneal nerve fiber quantification (ACCMetrics software, University of Manchester, Manchester, UK) was undertaken and consists of two steps: (1) CCM image enhancement and nerve fiber detection and (2) quantification of CNFD, CNBD and CNFL [ 30 , 31 ]. ACCMetrics is available to all potential collaborators solely for research purposes (non-for-profit/non-commercial) and is protected by the University of Manchester in the form of a license agreement, which can be requested online ( https://sites.manchester.ac.uk/ccm-image-analysis/ (accessed on 11 April 2022)).

Previously, QST abnormal ranges have been defined as values outside the 95% confidence interval of healthy subjects [ 34 ]. The “normal range” of CNFL change over 2 years is defined as values falling within 2 standard deviations of the mean in controls. Therefore, patients who demonstrated a >2SD reduction in CNFL over 2 years were considered to have rapid nerve fiber loss (RCNFL).

The sample size analysis was based on putative small nerve fiber decline in the cohort with diabetes. An assumption of paired groups was used to calculate the sample size considering a change in CNFD over 2 years. Recruiting a minimum of 14 participants (for the diabetic group) provided 80% power to detect a clinically meaningful change in CNFD of 7.5 nerves/mm 2 with a standard deviation of 9 nerves/mm 2 and an assumption of a type 1 error (α-level) of 0.05. Lewis et al. [ 25 ] previously detailed 30% RCNFL in those without overt diabetic neuropathy. Therefore, a minimum of 47 participants with diabetes were required.

There was no significant change in clinical and metabolic variables between baseline and year-2 follow-up in controls and people with T1DM, apart from a significant decrease in HbA1c and systolic BP ( p = 0.04) in controls and diastolic BP ( p = 0.004) and eGFR ( p = 0.02) in T1DM.

There were no changes in VAS, NSP, NDS, QST (apart from CST), IENFD, NCCA or CCM parameters in either group over the 2-year follow-up.

Our study showed no progression of diabetic neuropathy in a cohort of patients with T1DM with stable moderate glycemic control and good blood pressure and lipid control over 2 years. However, we have identified that 22% of T1DM patients show a rapid loss of corneal nerve length over 2 years which was accompanied by a reduction in sural and peroneal NCV that could not be attributed to a change in clinical, metabolic or vascular risk factors. Sural nerve conduction velocity predicts incident DPN over 4 years [ 35 ] and a number of clinical trials have utilized sural NCV as a primary endpoint [ 36 ]. In addition, peroneal nerve conduction velocity predicts both foot ulceration and death in diabetes [ 37 ]. Our study shows minimal progression of DPN [ 10 , 11 , 22 ] which is lower than previous studies demonstrating a reduction in nerve conduction velocity of ~0.5 m/s/year in T1DM [ 10 ]. Indeed, we have previously reported that CNFD or CNFL does not change, and CNBD increases over 4 years in a cohort of T1DM patients with moderate glycemic control and good blood pressure and lipid control [ 22 ]. More recently, Lewis et al. [ 25 ] showed that RCFNL was more common in people with baseline DPN, and indeed, our cohort of participants with T1DM had borderline NCS values for DPN. However, progressive corneal nerve fiber loss was not associated with baseline or change in HbA1c or BMI [ 25 ]. Lewis et al. [ 25 ] showed that rapid loss of corneal nerve fiber length appeared to be the most sensitive measure for identifying progressors [ 38 , 39 ]. The present study shows that a decline in CNFD, CNBD and CNFL could identify progressors. Several studies of CCM have demonstrated that a single measure of CNFL (<15 mm/mm 2 ) can predict incident DPN, and those with CNFL (<11 mm/mm 2 ) are likely to have prevalent DPN [ 38 , 39 ]. Thus, identifying patients with RCNFL may have utility in clinical trials to identify and enroll subjects with progressive rather than stable disease to enable participant enrichment for trials of new disease modifying therapies for DPN. Recent studies indicate that small nerve fiber damage, in particular a reduction in CNFL, precedes an abnormality in NCS [ 27 , 40 ]. Gibbons et al. reported that both NCS and QST showed no progression in patients with T2DM over 3 years, whereas there was a significant worsening in the nerve axon reflex, a measure of small fiber neuropathy [ 41 ]. In the present study, 22% of patients with T1DM showed a more rapid decline in corneal nerve fibers and peroneal and sural nerve conduction velocity. Lewis et al. showed similar outcomes and in regression analysis established a significant association between CNFL loss and other measures of diabetic neuropathy [ 25 ]. Such a rapid decline in CNFL may have clinical relevance as we have shown that it may be temporally related to the development of foot ulceration and Charcot foot [ 42 ]. Natural history studies have demonstrated early corneal nerve fiber regeneration within twelve months of simultaneous kidney and pancreas transplantation [ 21 , 43 ], continuous subcutaneous insulin infusion [ 44 ], and treatment with the non-erythropoietic peptide Cibenitide [ 45 , 46 ], with no change in other measures of neuropathy. Another study has demonstrated a significant 29% improvement in CNFL after 12 months of treatment with omega-3 polyunsaturated fatty acids, with no change in neurophysiology or QST [ 47 ]. In a randomized placebo-controlled trial of Omega-3 fatty acid in patients with type 1 diabetes, after 180 days, there was a significant increase in CNFL, but no change in thermal thresholds, autonomic function or nerve conduction studies [ 48 ]. This suggests that we either need longer clinical trial durations in excess of 3 years when utilizing current FDA-accepted endpoints of symptoms, signs and neurophysiology or that CCM could be used as a co-primary endpoint to provide an early go/no-go signal for clinical trials of disease-modifying therapies in diabetic neuropathy. We acknowledge that the size of the cohort in this study was small, and therefore, detailed multivariate regression/correlation analyses were not possible. A future larger cohort analysis would benefit from a correlation analysis and subsequent matrix of CCM vs. NCS measures. Our data add to the current evidence that CCM can detect incident neuropathy and rapid nerve fiber decline. Indeed, long duration of diabetes and the presence of advanced neuropathy have been suggested to contribute to the failure of clinical trials. Thus, our finding that we are able to identify a sub-group of T1D patients with more rapid progression adds to the utility of CCM. These data compliment the larger studies by Lovblom et al. [ 38 ] and Lewis et al. [ 25 ], which also showed that CCM can effectively quantify neuropathy progression to risk stratify individuals and enable selection of appropriate cohorts for future clinical trials of pathogenetic treatments. Additiona