Corneal confocal microscopy detects small nerve fibre damage in patients with painful diabetic neuropathy

Neuropathic pain is believed to arise from damage to nociceptive C fibres in diabetic neuropathy (DN). We have utilised corneal confocal microscopy (CCM) to quantify the severity of small nerve fibre damage in relation to the severity of neuropathic pain and quality of life (QoL) in patients with and without painful DN. 30 controls and patients with painful (n = 78) and painless (n = 62) DN underwent assessment of large and small nerve fibre function, CCM, neuropathic symptoms (small fibre neuropathy symptom inventory questionnaire, neuropathic pain scale) and QoL (SF-36, pre-R-ODS and hospital anxiety and depression scale). Patients with painful compared to painless DN, had comparable neurophysiology and vibration perception, but lower corneal nerve fibre density (20.1 ± 0.87 vs. 24.13 ± 0.91, P = 0.005), branch density (44.4 ± 3.31 vs. 57.74 ± 3.98, P = 0.03), length (19.61 ± 0.81 vs. 22.77 ± 0.83, P = 0.01), inferior whorl length (18.03 ± 1.46 vs. 25.1 ± 1.95, P = 0.005) and cold sensation threshold (21.35 ± 0.99 vs. 26.08 ± 0.5, P < 0.0001) and higher warm sensation threshold (43.7 ± 0.49 vs. 41.37 ± 0.51, P = 0.004) indicative of small fibre damage. There was a significant association between all CCM parameters and the severity of painful neuropathic symptoms, depression score and QoL. CCM identifies small nerve fibre loss, which correlates with the severity of neuropathic symptoms and reduced QoL in patients with painful diabetic neuropathy.

underlying nerve damage. Two large neuropathy phenotyping studies have shown that nerve conduction studies do not differentiate painful from painless neuropathy 20 , although thermal thresholds were associated with the severity of painful diabetic neuropathy 21 . With regard to structural damage to small fibres, reduced intraepidermal nerve fibre length 22 and increased regeneration and axonal swellings have been found in patients with painful DN 23 . Corneal confocal microscopy (CCM) is a rapid non-invasive surrogate marker for small nerve fibre damage in diabetic peripheral neuropathy 24 . In two small studies we and others showed greater corneal nerve fibre loss in the central cornea 22,25 and inferior whorl 26 in patients with painful compared to painless diabetic neuropathy.
This study aimed to establish if CCM can differentiate patients with painful compared to painless DN and whether the severity of corneal nerve damage was associated with the severity of neuropathic pain and quality of life.

Results
Demographic and clinical findings. Controls were age-matched with the patients with diabetes mellitus.
Blood pressure (BP) was comparable between groups. The duration of diabetes was comparable between patients with painless and painful DN. 58% of patients with PDN were on pain medications. BMI (P < 0.005) and HbA1c (P < 0.0001) were significantly higher and total cholesterol (P < 0.0001), LDL (P < 0.05) and HDL (P < 0.04) were significantly lower in patients with painful DN compared to healthy controls. Patients with painless DN had a significantly higher HbA1c (P < 0.0001) and lower total cholesterol (P < 0.0001) compared to healthy controls. Total cholesterol (P = 0.03) and triglycerides (P = 0.01) were significantly higher in painful compared to painless DN (Table 1).
Neuropathic pain and QOL. The most common descriptors of neuropathic pain among patients with painful DN assessed with the NPS questionnaire were: ~95% suffered from unpleasantness and intensity, ~89% reported deep pain, ~70% reported itchiness and ~65% cold pain. Almost every individual experienced more than one type of pain. Based on the SFN-SIQ questionnaire, the most common symptoms among patients with painful DN were dry mouth (~72%), dizziness (~67%), diaorrhea (~67%) and sensitive skin (~63%) (Fig. 3a). 83% of patients with dry mouth symptoms and 80% of patients with dizziness were on pain medications.

Discussion
Painful diabetic neuropathy has a major impact on the patient's quality of life as a result of associated anxiety and depression 27 . The exact cause of neuropathic pain remains elusive 28 , especially as pain can be experienced when there is little or no existing damage to nerves and a patient with severe neuropathy may not have any complaints 29 . Previous studies have reported that damage to Aδ and C fibres is associated with neuropathic pain 30,31 . CCM can quantify small fibre pathology and stratify the severity of DN 32 . In the present study, we demonstrate greater small fibre dysfunction (warm and cold thresholds) and damage to corneal nerve fibres but no difference in large fibre measures such as VPT and nerve conduction studies in patients with painful compared to painless DN. This agrees with our recent study showing reduced corneal nerve parameters in patients with painful compared to painless DN and an association with the severity of pain 26 .
Quattrini et al. 7 previously reported a greater reduction in corneal nerve fibre length and intra-epidermal nerve fibre length in patients with painful compared to painless DN 22 . Wang et al. 25 reported significant alterations in corneal nerve morphology and a direct correlation with the severity of pain 25 , whereas Cheng et al. 23 reported axonal swelling in patients with painful compared to painless DN 23 .
This study also demonstrates a significant reduction in cold sensation threshold and an increase in warm sensation threshold in patients with painful compared to painless DN, consistent with Raputova et al. 21 who reported significantly worse cold and warm detection thresholds in patients with mild and moderate to severe neuropathic pain compared to patients without neuropathic pain 21 . Themistocleous et al. 20 reported a significant abnormality in QST parameters in patients with neuropathic pain but found no difference in nerve conduction studies, intraepidermal nerve fibre density and cold pain threshold 20 . The current study strengthens the evidence of greater small fibre loss and dysfunction in patients with painful compared to painless neuropathy. www.nature.com/scientificreports www.nature.com/scientificreports/ Validated questionnaires are an important tool for the assessment of neuropathic pain and its effect on the patients quality of life 33 , allowing us to assess the effect on patients' lives including their mood, daily activities and social life. We show a significant correlation between measures of corneal nerve damage, especially at the inferior whorl with the presence and severity of neuropathic pain, quality of life and severity of depression. However, this weak association should be interpreted cautiously and suggests that factors other than the extent of small fibre damage contribute to a worse QoL and severity of pain. Patients with painful DN also suffered from greater anxiety and depression compared to those with painless DN. Previously, Gore et al. 34 also demonstrated higher scores of anxiety and depression in patients with more severe neuropathic pain 34 . Dizziness and dry mouth were the most common symptoms among patients with painful DN and could be attributed to concomitant use of medications to relieve neuropathic pain. The most common type of pain reported was deep pain, whilst the least www.nature.com/scientificreports www.nature.com/scientificreports/ common was itchy pain. Others using NPS have also reported sharp and deep pain to be common whilst itching and cold pain were infrequent 1,35 . Using the pre-R-ODS questionnaire we show a significant reduction in all dimensions of QoL except eating. The SF-36 average score indicated a significant reduction in QoL (e.g. physical functioning and health, wellbeing, emotional problems and general health) in patients with painful compared to painless DN.
To our knowledge this is the first study to evaluate the association between the severity of small fibre neuropathy using CCM and its relationship to the severity of painful symptoms, patient mood, daily activities, social life and overall QoL. Although the VAS is a validated scale for the assessment of the severity of pain, we acknowledge it is very subjective. Another limitation of the current study is the lack of validated neuropathy specific QoL tools such as NeuroQoL and Norfolk QoL-DN that more accurately quantify QoL in diabetic neuropathy.
In conclusion, we show that corneal confocal microscopy detects greater corneal nerve fibre loss in patients with painful diabetic neuropathy and this correlates with the severity of neuropathic pain, the patient's mood and QoL. These data encourage further studies to assess the potential utility of CCM in relation to a change in the severity of neuropathic pain and quality of life in clinical intervention trials.

Methods
Study subjects. 140 patients with type 1 (n = 41) and type 2 diabetes (n = 99) and 30 age-matched healthy controls were studied. All participants underwent demographic and clinical examination, detailed assessment of peripheral neuropathy and ophthalmic examination. Participants with a history of malignancy, vitamin B12 or folate deficiency, chronic renal and/or liver impairment, connective tissue or infectious disease, neuropathy of other cause than diabetes, current diabetic foot ulcer, corneal trauma, or systemic disease of the cornea and contact lens wear were excluded from the study. A signed consent form was obtained from each participant prior to taking part in the study. This study adhered to the tenets of the Declaration of Helsinki and was approved by Greater Manchester East Research Ethics Committee.
Demographic and clinical neuropathy assessment. BMI, blood pressure, lipid profile and glycated haemoglobin (HbA1c) were measured in each participant. Neurological deficits were measured using the simplified Neuropathy disability score (NDS) (0-10) and patients were confirmed to have DN based on an NDS > 2. Further, based on the visual analogue scale (VAS), patients were divided into two groups of painful (VAS > 4) (n = 78) and painless DN (VAS ≤ 4) (n = 62) 36 .
Vibration perception threshold (VPT) was measured with a Neurothesiometer (Scientific Laboratory Supplies, Wilford, Nottingham, UK), and cold and warm perception thresholds were assessed with the neurosensory analyzer (TSA-II NeuroSensory Analyzer, Medoc Ltd., Ramat-Yishai, Israel) on the dorsum of the left foot. In addition, sural sensory nerve amplitude and nerve conduction velocity were assessed by a consultant neurophysiologist using nerve conduction testing machine (Dantec "Keypoint" system, Dantec Dynamics Ltd., Bristol, North Somerset, UK).
Corneal confocal microscopy. All participants were examined using slit-lamp biomicroscopy to eliminate the presence of ocular surface infection, corneal ulceration and narrow angle glaucoma prior to CCM examination. Both eyes were examined with a laser-scanning CCM (Heidelberg Retinal Tomograph-III Rostock Cornea Module, Heidelberg Engineering GmbH, Heidelberg, Germany) using our previously established protocol 37 . www.nature.com/scientificreports www.nature.com/scientificreports/ Central corneal nerves and the inferior whorl were imaged at the level of sub-basal nerve plexus. CCM provides two-dimensional images with a resolution of 10μm and a size of 384 × 384 pixels.

Pain and quality of life questionnaires. QoL was assessed using the validated Short Form 36 Health
Survey (SF-36) score, preliminary Rasch-built Overall Disability Scale (pre-R-ODS), Hospital Anxiety Depression Scale (HADS) and small fibre neuropathy and symptom inventory questionnaire (SFN-SIQ). 96 patients completed all questionnaires.
SF-36 consists of 36 questions, divided into 8 dimensions and scored in a range 0-100 18 . Our analysis included a total score, which was an average value of all 8 dimensions.
A 146-item preliminary Rasch-built Overall Disability Scale (pre-R-ODS) was used to assess the different aspects of a patient's QoL 40 . Each question was scored as 0 -not possible, 1 -possible with effort and 2 -easy to perform. However, in this study, we could not use Rasch model to interpret the pre-R-ODS questionnaire, as the number of participants who completed the questionnaire was less than the number of questions, hence not fulfilling Rasch model requirements. Therefore, the pre-R-ODS questionnaire was divided into ten different dimensions: change and hold body position, walking and movement by transport, carrying, moving, handling objects, self-care, eating, household, daily tasks, meeting other people and work, studies and hobbies. For each patient, the average score for each dimension was established and expressed as a percentage.
HADS was used to score depression and anxiety levels 17 . The questionnaire consisted of 14 items and was divided into 2 subscales: 7 questions for the assessment of depression and 7 questions for anxiety. Each question was scored in a range of 0-3; the higher number represented the more severe symptoms. The subscale score was summed up from each question (range 0-21) and an average depression and anxiety score was compiled.
Other neuropathy symptoms including dry mouth, dizziness, dry eyes, changes in sweating, were assessed using the validated small fibre neuropathy and symptom inventory questionnaire (SFN-SIQ). SFN-SIQ consisted of 13 items. For each item, the scale range was 0-3 where 0=never, 1=sometimes, 2=often and 3=always 16 . For each participant, the overall score for SFN-SIQ was obtained from the sum of all items.
Neuropathic pain scale (NPS), a scale that consists of 10 items, each on a scale ranging from 0-10 was used to assess the most common types of pain among patients with painful DN 15 .

Statistical analysis. IBM SPSS statistic software (Version 22.0 for Macintosh, IBM Corporation, New
York, NY, USA) was used for the analysis. All values were calculated as mean ± standard error (SE). The significance was considered as P < 0.05. To compare values between more than two groups, One-Way Anova (post-hoc Bonferroni for parametric parameters and LSD for non-parametric parameters) was used. Independent-samples T-test was used to compare mean values between two groups (Mann-Whitney U test for non-parametric). Pearson correlation coefficient (Spearman's rank correlation coefficient for non-parametric) was measured to evaluate the association between variables. Graphs were made using GraphPad Prism software (Version 7.0c for Macintosh, GraphPad Software, La Jolla California, USA).

Data availability
The datasets generated and analysed during the study are available from the corresponding author on a reasonable request