Corneal confocal microscopy detects early nerve regeneration after pharmacological and surgical interventions: Systematic review and meta‐analysis

Corneal confocal microscopy (CCM) is an ophthalmic imaging technique that enables the identification of corneal nerve fibre degeneration and regeneration. To undertake a systematic review and meta‐analysis of studies utilizing CCM to assess for corneal nerve regeneration after pharmacological and surgical interventions in patients with peripheral neuropathy. Databases (EMBASE [Ovid], PubMed, CENTRAL and Web of Science) were searched to summarize the evidence from randomized and non‐randomized studies using CCM to detect corneal nerve regeneration after pharmacological and surgical interventions. Data synthesis was undertaken using RevMan web. Eighteen studies including 958 patients were included. CCM identified an early (1–8 months) and longer term (1–5 years) increase in corneal nerve measures in patients with peripheral neuropathy after pharmacological and surgical interventions. This meta‐analysis confirms the utility of CCM to identify nerve regeneration following pharmacological and surgical interventions. It could be utilized to show a benefit in clinical trials of disease modifying therapies for peripheral neuropathy.


| INTRODUCTION
Diabetic peripheral neuropathy (DPN) affects at least 50% of patients with diabetes and may lead to neuropathic pain and foot ulceration. 1 The diagnosis of DPN relies on symptoms, signs and electrophysiology, 2 and the food and drug administration (FDA) endorses the evaluation of symptoms, signs and nerve conduction studies (NCSs) in clinical trials of DPN. 3 Paraesthesia, burning pain, hyperalgesia, allodynia and numbness may occur with reduced warm/cold sensation and mechanical or proprioceptive function.Symptoms and signs have not proven to be robust endpoints in clinical trials of disease-modifying therapies. 4Ss are a quantitative measure of large fibre dysfunction, but we previously identified 21 double-blind randomized controlled trials of disease-modifying therapies for DPN, which had failed to demonstrate efficacy due to variability in the magnitude of change and direction of change in the placebo and active arms. 5all fibres contribute to painful neuropathy, a blunted blood flow response to tissue injury, 6 reduced pressure-induced vasodilation 7 and diabetic foot ulcers. 8Quantitative sensory testing (QST), intraepidermal nerve fibre density (IENFD) in skin biopsies and corneal confocal microscopy (CCM) can be used to assess small fibre dysfunction and damage, respectively. 9QST, as a part of the evaluation for the German Network on Neuropathic Pain protocol, 10 has shown good accuracy and reliability in single centres with expertise, 11 but in multicentre trials, high variability and poor repeatability have often led to negative outcomes. 12,13IENFD assessment is the de facto gold standard for diagnosing small fibre neuropathy (SFN) 14 and has excellent diagnostic efficiency. 15Although a recent systematic review has challenged the utility of IENFD, it was not found to be related to painful symptoms or the severity of DPN. 16In two recent studies, there was no significant change in IENFD after intravenous immune globulin 17 and supervised progressive resistance training. 18Furthermore, in a large cohort of participants undergoing bariatric surgery, there was no change in distal IENFD, QST or NCV. 19CCM is a rapid, ophthalmic imaging technique that has good diagnostic utility for DPN. 20,21[28] This meta-analysis has assessed the overall utility of CCM in detecting corneal nerve regeneration after pharmacological and surgical interventions in patients with peripheral neuropathy due to obesity, hyperlipidemia and diabetes.

| Data sources
This systematic review and meta-analysis reports according to the meta-analysis of observational studies in epidemiology guidelines. 15 registered the protocol with the International Prospective Regis- shown in Table 1.The search language was limited to English, and the grey literature was searched through ProQuest.
Randomized controlled trials (RCTs) or observational longitudinal studies reporting on at least one of the following: corneal nerve fibre density (CNFD), corneal nerve branch density (CNBD), corneal nerve fibre length (CNFL) before systemic intervention and 1-6 months or 1-5 years after intervention were included.Narrative reviews, systematic reviews, correspondence and case reports were excluded.
Study country, age, gender, duration of follow-up, intervention, CCM device, software for image analysis, CNFD, CNBD and CNFL were extracted when available.Studies using CCMetrics, ACCMetrics, Ima-geJ or morphometric software were included.Data from the study by Azmi et al. 29 were presented for the continuous subcutaneous insulin infusion (CSII) group and multiple daily insulin (MDI) injection group.
For the study by Ponirakis et al., data were presented for the exenatide and pioglitazone group and basal-bolus insulin group. 30From the study by Zhang et al., data were presented for the intramuscular injection and oral mecobalamin groups. 31Data from Misra et al. are unpublished. 32Characteristics of the included studies are shown in Table 2.
Data presented as median (range) were converted into mean ± SD using an online calculator, and data presented as mean ± SEM were converted into mean ± SD using the RevMan calculator. 33

| Study selection
After duplicate removal, all citations were screened for relevance and where there was a lack of consensus a third (senior) author was consulted (RAM).The most recent versions of the studies were reviewed for eligibility (HG and EE) according to inclusion/exclusion criteria.Full texts of all eligible citations were obtained, and HG and EE independently made the final inclusion and exclusion, but in case of disagreement, RAM was consulted to resolve conflicts.A PRISMA flow chart of the search was produced (Figure 1), and a data extraction tool was developed to extract data from the studies, which was verified by HG and EE.Corresponding authors were contacted to obtain any unpublished data.

| Data extraction and quality assessment
The Cochrane Collaboration's tool (section 8) 34 was used to categorize the risk of bias into not applicable, low, unclear or high risk for six domains: selection bias, performance bias, detection bias, attrition bias, reporting bias and other bias.On the risk of bias figure, answers with "not applicable" are empty, "low risk" are displayed as "unclear risk" are displayed as and "high risk" are displayed as .
Quality assessment was undertaken by HG and AL and if the risk of bias for a study was high or unclear, the effect of excluding the study was assessed and the relevant outcomes were reported.

| Data synthesis and analysis
Meta-analysis was undertaken in RevMan Web.Random effects meta-analysis was used to account for heterogeneity (differences in study population, intervention and duration of follow-up).The mean difference (MD) with 95% confidence interval (CI) was calculated for CNFD, CNBD and CNFL and chi-squared (χ 2 ) was used to test for a difference between subgroups.The I 2 statistic, derived from Cochrane's chi-squared test Q, was used to describe the betweenstudy variations attributed to variability in the true exposure effect 34 with an I 2 value of 0%-40% classified as not important, 30%-60% moderate, 50%-90% substantial and 75%-100% considerable.

| Sensitivity analysis
Small study effects were assessed by undertaking sensitivity analysis that examined how the results of the meta-analysis change under different assumptions.Adjustment for heterogeneity and small study effects was undertaken by comparing fixed and random effects models (10.4.4.1) and the trim and fill strategy according to Cochrane recommendations. 33

| Comparing fixed and random effects estimates
Random effects meta-analysis was used for all study variables in anticipation of heterogeneity due to differences in study design, population, intervention, duration of follow-up and software for CCM image analysis.For variables with a significant publication bias (Egger's test p < .05),we applied the fixed effects model to account for small study effects.Heterogeneity remained the same for CNFD, CNBD and CNFL, and changing the random effect to a fixed effect did not change the overall study effect or I 2 and random effect model was applied to all analyses.

| Trim and fill strategy
When removing small studies to correct for the funnel plot asymmetry (Egger's test p < .05),p-values for Egger's test remained the same for CNFD, CNBD and CNFL.All studies were included to calculate the overall effect size for the meta-analysis.Also removing studies with potentially high risk of bias due to unmasking of outcome assessments 35,36 or inability to assess the risk of bias due to lack of this ] explode all trees #8 ("peripheral nervous system disease" or "peripheral nervous system diseases" or "peripheral neuropathy" or "peripheral neuropathies" or polyneuropath* or "motor neuropathy" or "motor neuropathies" or "sensory neuropathy" or "sensory neuropathies" or "small nerve fiber" or "small nerve fibers" or "small nerve fibre" or "small nerve fibres" or "small fiber pathology" or "small fiber pathologies" or "small fibre pathology" or "small fibre pathologies" or "small fiber neuropathy" or "small fiber neuropathies" or "small fibre neuropathy" or "small fibre neuropathies" or "fibromyalgia syndrome" or "nerve fiber density" or "nerve fibre density" or "nerve density" or "nerve fiber length" or "nerve fibre length" or "neuropathic pain"): ) AND TS=("peripheral nervous system disease*" OR "peripheral neuropath*" OR "polyneuropath*" OR "motor neuropath*" OR "sensory neuropath*" OR "small nerve fiber*" OR "small fiber patholog*" OR "small fiber neuropath*" OR "fibromyalgia syndrome" OR "nerve fiber density" OR "corneal nerve density" OR "nerve fiber length" OR "neuropathic pain" OR "Obesity") ) AND noft(cornea*) AND noft(neuropath* OR polyneuropath* OR "small nerve fiber" OR "small nerve fibre" OR "small fiber" OR "small fibre" OR "fibromyalgia" OR "nerve fiber density" OR "nerve fibre density" OR "nerve density" OR "nerve fiber length" OR "nerve fibre length" OR "neuropathic pain" OR obesity) T A B L E 2 Characteristics of the included studies.information in the original studies was investigated.By removing Adam et al., 35 the overall effect of CNFD at 1-6 months changed from p = .005to p < .00001and at 1-5 years, the overall effect was changed from p = .003to p = .01.Similarly, removing Adam et al.

Study
from the CNBD analysis at 1-5 years changed the overall effect from p = .0005to p = .0008.Additionally, by removing both Adam et al.
and Chen et al. from the CNFL analysis at 1-6 months, the overall effect was changed from p = .002to p < .00001.When Adam et al.
were removed from the CNFL analysis at 1-5 years, the overall effect was changed from p = .0009to p = .002.Even by removing these two studies 35,36 with a high risk of bias due to the inability to mask the outcome assessment due to the invasive nature of both interventions, the overall effect remained significant, thus these studies were not excluded from the analysis.with a significant difference in the magnitude of CNFD improvement between studies (χ 2 = 18.73, p = .03;Figure 2).

Change at 1-5 years
Twelve studies 29,30,32,35,40,41,[44][45][46][47][48][49] with 958 participants were included in the meta-analysis.CNFL (mm/mm 2 ) increased significantly following 1-5 years of interventions (MD = 1.45, 95% CI 0.59-2.30,p = .0009),with a significant difference in the magnitude of CNFL improvement between studies (χ 2 = 50.73,p < .00001; Figure 7).requirements of a biomarker according to the NIH Biomarkers Definitions Working Group. 52 contrast, a number of FDA approved measures of neuropathy do not perform well as end-points in clinical intervention studies. 53us, we previously demonstrated early corneal nerve regeneration within 6 months, whilst neuropathic symptoms and nerve conduction improved 24 months after pancreas and kidney transplantation in patients with type 1 diabetes. 41,43,54Similarly, in patients with type 2 diabetes undergoing bariatric surgery, the neuropathy disability score and QST did not improve, whereas, CNFD, CNBD and CNFL all improved after 12 months. 35In patients with type 1 diabetes undergoing islet transplantation, motor and sensory nerve conduction improved after 1 year, with no change in sudomotor function. 55Furthermore, in diabetic patients treated with CSII, there was evidence of corneal nerve regeneration, without an improvement in NCS over 24 months. 29We have also shown corneal nerve regeneration in physically active patients with type 2 diabetes over 24 months, 56 whereas a recent meta-analysis showed no change in peroneal and sural NCV in patients who exercise. 57In an early study, we showed corneal nerve regeneration after 24 months of improvement in HbA1c, lipids and blood pressure, 48 whilst in a recent study from Japan, a decrease in HbA1c and weight was associated with improved nerve conduction and corneal nerve regeneration over 4 years. 46M shows corneal nerve regeneration as early as 28 days after treatment with an erythropoietin analogue with anti-inflammatory and neuroprotective effects 37 and after 30 days of treatment with fenofibrate 58 in patients with type 2 diabetes.Similarly, treatment of patients with type 1 diabetes with Omega-3 was associated with corneal nerve regeneration within 6 months, but with no change in neuropathic symptoms, QST and NCS. 38,49In a study from China, CCM was used to show corneal nerve regeneration after 8 weeks of both oral and intramuscular mecobalamin. 31In recent meta-analyses, there was no change in pain scores or neuropathic deficits, but there was an improvement in median and peroneal motor NCV with mecobalamin, 59 whilst treatment with vitamin B 12 lead to an improvement in neuropathic symptoms in two out of three studies. 60is is the first systematic review and meta-analysis that provides

| CONCLUSIONS
Our meta-analysis provides evidence that CCM could be a viable endpoint that identifies early nerve regeneration in clinical trials of patients with peripheral neuropathy.
ter of Systematic Reviews (PROSPERO) in August 2023 (CRD42023319565).EMBASE (Ovid), PubMed, CENTRAL and Web of Science (WoS) were searched using MESH subject headings and keywords for PubMed and CENTRAL databases and Emtree subject headings and keywords for Embase from inception to June 2023.Several terms were tested for relevance, and the final search strings are 2 TS= clinical trial* OR TS=research design OR TS=comparative stud* OR TS=evaluation stud* OR TS=controlled trial* OR TS=follow-up stud* OR TS=prospective stud* OR TS=random* OR TS=placebo* OR TS=(single blind*) OR TS=(double blind*) #3 #1 AND #2 ProQuest Dissertations & Theses noft(confocal AND (microscop* OR image*)

3 |
DISCUSSIONIn this meta-analysis of 958 participants, CCM demonstrates an improvement in all three major corneal nerve parameters (CNFD, CNBD and CNFL) within 8 months and over 1-5 years after pharmacological or surgical interventions in patients with obesity, diabetes or hyperlipidaemia, confirming the utility of CCM to detect corneal nerve regeneration.Given that CCM provides a rapid, non-invasive and highly sensitive measure of nerve regeneration,20,50,51 it fulfils the F I G U R E 1 Flowchart of the included studies.

F I G U R E 2
Forest plots of corneal nerve fibre density at baseline and after 1-8 months of interventions.CI, confidence interval; FU, follow-up.F I G U R E 3 Forest plots of corneal nerve fibre density at baseline and after 1-5 years of interventions.CI, confidence interval; FU, follow-up.
an overall estimate of the ability of CCM to identify nerve regeneration after pharmacological and surgical interventions in patients with obesity, diabetes and hyperlipidemia.The reliability of establishing a single estimate for the effect size of corneal nerve outcome measures may be affected due to the inclusion of the same subjects in different F I G U R E 4 Forest plots of corneal nerve branch density at baseline and after 1-8 months of interventions.CI, confidence interval; FU, follow-up.F I G U R E 5 Forest plots of corneal nerve branch density at baseline and after 1-5 years of interventions.CI, confidence interval; FU, follow-up.studies, type of CCM, mode of image acquisition and image analysis software used to quantify corneal nerve parameters.Indeed, there was evidence of significant heterogeneity, which was expected due to the different study populations, interventions and duration of follow-up.

F I G U R E 6
Forest plots of corneal nerve fibre length at baseline and after 1-8 months of interventions.CI, confidence interval; FU, follow-up.F I G U R E 7 Forest plots of corneal nerve fibre length at baseline and after 1-5 years of interventions.CI, confidence interval; FU, follow-up. 34 44"update review".ab.#45 (databases adj4 searched).ab.#46 (rat or rats or mouse or mice or swine or porcine or murine or sheep or lambs or pigs or piglets or rabbit or rabbits or cat or cats or dog or dogs or cattle or bovine or monkey or monkeys or trout or marmoset$1).ti.and animal experiment/ #47 Animal experiment/ not (human experiment/ or human/) #