Improvement in Neuropathy Outcomes With Normalizing HbA_1c in Patients With Type 2 Diabetes

Intensive glycemic control (GC) has shown an equivocal efficacy regarding diabetic peripheral neuropathy (DPN) in type 2 diabetes (1), mainly due to nonoptimized HbA_1c levels. Randomized trials (2,3) have not been able to establish the optimum GC level for improving neuropathy outcomes in type 2 diabetes. In type 1 diabetes, near-normoglycemia for 24 years prevented nerve function declines (4). The simultaneous pancreas-kidney transplantation (SPK) can normalize HbA_1c levels without severe hypoglycemia and was shown to improve corneal nerve fiber (CNF) measures (5). However, the long-standing normalization of HbA_1c levels by insulin-providing agents (IPAs) in poorly controlled type 2 diabetes seems infeasible owing to potential severe hypoglycemia (3). Therefore, the benefit of normalizing HbA_1c levels without hypoglycemia to neuropathy outcome measures (NOMs) in type 2 diabetes has never been investigated.

Besides mean GC levels, glycemic variability (GV) (6) and metabolic memory influence DPN and other diabetic microangiopathies (7,8). Impaired glucose tolerance (IGT) may cause neuropathy and alter CNF morphology as an indicator of small-fiber neuropathy (9).

Here, we aim to investigate the impact of normalized HbA_1c levels without hypoglycemia on neurophysiological functions and CNF measures along with retinopathy and nephropathy in patients with poorly controlled type 2 diabetes.

Among 671 patients who newly visited the Ishibashi Clinic between May 2011 and July 2013 after the diagnosis of hyperglycemia, we extracted 38 patients with type 2 diabetes fulfilling the following criteria (extensive HbA_1c control [EHC] cohort): 1) baseline HbA_1c level >7.5% (58.5 mmol/mol) (mean level 9.6% [81.4 mmol/mol]), 2) follow-up period >3 years, and 3) annual mean HbA_1c levels <6.0% (42.1 mmol/mol) for the last 2 years before the final visit. Patients visited Ishibashi Clinic after repeated diagnosis of hyperglycemia. Thirty-eight patients with type 2 diabetes matched for age, sex, baseline HbA_1c levels, and height to patients with EHC were followed without EHC and served as a control for EHC patients. Forty-eight healthy control subjects with normal glucose tolerance (NGT) and 34 subjects with IGT were enrolled and studied only at baseline.

The patients with diabetes with or without EHC were followed for 3–5 years (4.3 and 4.1 years on average, respectively) until September 2017. Patients who were treated by insulin-sensitizing agent (ISA) or IPA were given biguanides or pioglitazone or given sulfonylureas or insulin, respectively (10). The exclusion criteria were as follows: any other clinically evident causes of neuropathy apart from diabetes, vitamin B deficiency, nonproliferative severe or proliferative diabetic retinopathy as defined by the Early Treatment Diabetic Retinopathy Study (ETDRS) (11), corneal diseases, history of refractive surgery, and use of contact lenses. All subjects provided written informed consent as per the Declaration of Helsinki, and the study protocol was approved by the ethics committee of Ishibashi Clinic.

Patients with diabetes with EHC at end point, control subjects, and subjects with IGT at baseline underwent a 75-g oral glucose tolerance test (OGTT). The BMI, blood pressure, casual postprandial plasma glucose (CPPG), and HbA_1c levels were measured monthly during the terms of study. In patients with type 2 diabetes, SD and coefficient of variation (CV) of CPPG and HbA_1c levels over the follow-up period were calculated for estimating GV. The serum lipid levels and urinary creatinine and albumin levels were assessed every 3 months. An albumin-to-creatinine ratio (ACR) >30 mg/g creatinine twice a year was labeled as nephropathy (12).

At baseline and end point, bilateral retinal fundus images (45°) were captured and graded according to the ETDRS scale as no apparent retinopathy, 0; mild nonproliferative diabetic retinopathy, 1; and moderate nonproliferative diabetic retinopathy, 2 (11).

All patients with diabetes at baseline and end point, control subjects, and subjects with IGT at baseline underwent detailed neurophysiological examinations. Neurological deficits were assessed using the modified neuropathy disability score (NDS) (13), which evaluates vibration perception, pinprick, temperature perception, and ankle reflexes. The classification for evaluation of neuropathy was based on the recommendations of the Toronto Consensus on Diabetic Neuropathies (14), which considers a confirmed DPN as a combination of the presence of abnormal nerve conduction velocity (NCV) and a symptom, or a sign. As a result, patients with NDS >2 and sensory NCV (SCV) of sural nerve <42 m/s were labeled with neuropathy based on the Toronto consensus. The cutoff level of sural nerve SCV was determined according to the mean ± 2 SD (49.99 ± 8.02 m/s) of sural nerve SCV in healthy control subjects (n = 38), matched for age, sex, and height to patients with diabetes with EHC, recruited from 98 healthy control subjects at Ishibashi Clinic.

Electrophysiology and NCV studies have been measured with an electromyography instrument (Neuropak S1; Nihon Kohden, Tokyo, Japan). The motor NCV (MCV) (median nerve) and SCV (ulnar and sural nerves) and their action potential amplitudes were determined. Skin temperature was maintained above 32°C.

The vibration perception threshold (VPT) was measured at the left medial malleolus using a biothesiometer (Bio-Medical Instrument Company, Newbury, OH). Warm (WPT) and cold (CPT) perception thresholds at the dorsum of the foot were determined using a thermal stimulator (Intercross-200; Intercross Co., Tokyo, Japan). CAN was assessed by the CV of R-R intervals (CV_R-R) calculated from the R-R intervals of 200 electrocardiogram samples.

All patients with diabetes at baseline and end point, control subjects, and subjects with IGT once at baseline were examined using a Heidelberg Retina Tomograph (HRT III) in vivo corneal confocal microscope with the Rostock Corneal Module (Heidelberg Engineering, Heidelberg, Germany) (15). Based on an established protocol (16), six high-quality images per subject from Bowman’s layer were captured and analyzed to quantify the following CNF morphological parameters: 1) CNF density (CNFD): total number of major nerve fibers/mm² corneal tissue, 2) CNF length (CNFL): total length of all nerve fibers (mm/mm²), 3) corneal nerve branch density (CNBD): number of branches emanating from all major nerve trunks/mm², 4) beading frequency (per 0.1 mm), and 5) bead size (µm²) (17). Except for bead size, all measurements were performed using ImageJ (Texelcraft, Tokyo, Japan). The examiners and the team members who analyzed the images and study results were all blinded and masked to the study groups.

All statistical analyzes were performed using SPSS, version 19 (SPSS, Chicago, IL), and P value <0.05 was considered statistically significant.

A priori analysis of sample power (α = 0.05; β = 0.80) using G*Power 3.1 (http://gpower.software.informer.com/3.1/) revealed that 53 subjects per each group were required. We could not recruit enough subjects. Then a post hoc analysis of sample power was conducted with use of a one-sided ANOVA (significance of 0.05) and the Kruskal-Wallis test for CNF and neurophysiological measures. The subject’s statistical power ranged from 0.89 to 0.99. Therefore, 158 subjects in total sample size gave us adequate statistical power.

All values are presented as mean ± SEM. All data sets were tested for normality using the Shapiro-Wilk test. The differences between baseline and end point in two patient cohorts were assessed using the paired t test and Wilcoxon signed rank test for normally and nonnormally distributed continuous variables, respectively, and χ² test and McNemar test for normally and nonnormally distributed categorical variables, respectively. Values at baseline in patients with or without EHC and control subjects or between NGT and IGT subgroups categorized by end point 75-g OGTT in patients with EHC were compared by t test and Mann-Whitney test for normally and nonnormally distributed continuous variables and χ² test for categorical variables. The comparison of normally distributed variables between end point in two patient cohorts, control subjects and subjects with IGT, was performed using one-way ANOVA for continuous variables and the χ² test for categorical variables followed by Bonferroni correction. For nonnormally distributed variables, the Kruskal-Wallis test was applied followed by the Mann-Whitney U test and Bonferroni correction for continuous variables and the χ² test for categorical variables. Correlations between the changes in NOMs by GC and mean clinical factors over follow-up period, GV parameters, or baseline HbA_1c levels in patients with EHC were assessed using Spearman rank correlation coefficient or multiple regression analysis.

The sensitivity and specificity of CNF and neurophysiological measures in differentiating between control subjects, subjects with IGT, and patients with EHC were assessed at baseline and end point using receiver operating characteristic analysis.

The sex, age, height, and baseline HbA_1c levels were matched between two cohorts with diabetes with or without EHC. During follow-up period, BMI in patients with EHC decreased (mean ± SEM body weight reduction 7.3 ± 1.2 kg) but not in patients without EHC. Systolic (SBP) and diastolic (DBP) blood pressure at end point in two patient cohorts significantly decreased after GC. In patients with EHC, HbA_1c levels for the last 2 years of the follow-up period and at end point were <6.0% (42.1 mmol/mol). The mean HbA_1c levels during the whole follow-up period in patients without EHC were higher than those in patients with EHC. HbA_1c levels in control subjects were lower than those at end point in two patient cohorts and in subjects with IGT. In patients with EHC, SD and CV of CPPG and HbA_1c levels over the last 2 years from end point were significantly lower than those over the whole follow-up period. Plasma glucose levels before and after 75-g OGTT (30, 60, and 120 min) at end point in patients with EHC were 102 ± 2.2, 203 ± 7.2, 183 ± 5.0, and 152 ± 7.0 mg/dL, respectively. At end point, 120-min plasma glucose in EHC and IGT subjects was significantly higher than that in control subjects (Table 1). Of patients on EHC, 50.0% and 28.9% returned to IGT and NGT, respectively. Serum triglyceride levels in patients with EHC decreased after GC. The end point estimated glomerular filtration rate (eGFR) in two patient cohorts decreased to a level similar to that in control subjects and subjects with IGT, and ACR significantly decreased. At baseline, 86.8% and 68.4% of patients with diabetes with or without EHC, respectively, were not treated. Over the last 2 years from end point, IPAs were not prescribed for any patients with EHC (Table 1).

The smoking and alcohol consumption in the two cohorts with diabetes were similar between baseline and end point and also similar between baseline and end point in patients, control subjects, and subjects with IGT.

All baseline CNF measures in both cohorts with diabetes were similar but altered compared with control subjects. The normalized HbA_1c levels for 2 years in patients with EHC improved all CNF measures but not CNFD (P = 0.225), with no improvement in those without EHC. CNFD and CNFL at end point in patients with EHC were less than those in IGT, while other CNF measures were similar to those in subjects with IGT. All CNF parameters but CNFD in subjects with IGT were significantly less than those in control subjects (Fig. 1A–F and Table 2).

The normalized HbA_1c levels in patients with EHC significantly decreased NDS, though NDS was still higher than in control subjects and subjects with IGT. NDS in subjects with IGT was significantly higher than that in control subjects. All baseline neurophysiological measures in two cohorts with diabetes were similar. The normalized HbA_1c levels for 2 years with EHC improved results of all neurophysiological tests to be similar to those in subjects with IGT except for the median nerve amplitude; in patients without EHC, GC improved only VPT. Most neurophysiological tests in subjects with IGT were inferior to those in control subjects (Table 2). The prevalence of retinopathy and ETDRS retinopathy scale at end point increased insignificantly with (P = 0.183 and 0.180, respectively) or without (P = 0.974 and 0.317, respectively) EHC. EHC significantly decreased the prevalence of neuropathy and nephropathy, while GC without EHC decreased only nephropathy (Table 2).

The improvements in NDS (mean ± SEM −1.36 ± 0.20 vs. −0.68 ± 0.21), median nerve MCV (1.84 ± 0.26 vs. 0.89 ± 0.25), and VPT (−0.81 ± 0.06 vs. −0.26 ± 0.16) in the NGT subgroup of type 2 diabetes with EHC were significantly (P < 0.05) greater than those in the IGT subgroup. CNFD, CNFL, CNBD, and all other neurophysiological tests in the NGT subgroup improved more than those in the IGT subgroup, but the differences were insignificant (P = 0.118–0.787).

The mean HbA_1c levels during the whole follow-up period were closely related with the improvement in CNFD, CNFL, CNBD, median nerve MCV, sural nerve SCV, and WPT (Fig. 1G–L).

The multiple regression analysis indicates that the changes in all CNF and neurophysiological measures by EHC were closely associated with the mean HbA_1c level over the whole follow-up period. Higher SD and CV of HbA_1c levels and CPPG were negatively associated with some CNF and neurophysiological measures. The high baseline HbA_1c levels deteriorated many NOMs (Table 3). High DBP, LDL cholesterol, and triglycerides levels negatively influenced interval changes in median nerve MCV, ulnar and sural nerve SCV, sural nerve amplitude, CPT (standard β −0.277 to −0.411 and 0.322–0.336, P = 0.005–0.049), CNFD, CNFL, and beading frequency (standard β −0.329 to 0.588; P = 0.022–0.008).

At baseline and end point, with use of CNFD and CNFL, our ability to differentiate between control subjects, subjects with IGT, and patients with type 2 diabetes with EHC was excellent. The neurophysiological tests clearly differentiated control subjects from patients with type 2 diabetes. However, between subjects with IGT and control subjects or patients, the sensitivity and specificity of CNFD and CNFL were not always good (Supplementary Table 1).

The intensive GC in type 2 diabetes is associated with a reduction in microvascular complications (mostly albuminuria). The strict GC may play a role in preventing and ameliorating DPN. However, in the Kumamoto Study (18), the tight GC (HbA_1c 7.1% [54.1 mmol/mol]) prevented NCV but not CAN decline in type 2 diabetes. In the UK Prospective Diabetes Study (UKPDS) (2), the intensive and conventional GC (HbA_1c 7.0% vs. 7.9% [53.0 vs. 62.8 mmol/mol]) had a similar effect on DPN and CAN. In the ACCORD (Action to Control Cardiovascular Risk in Diabetes) trial (3), the intensive treatment (HbA_1c 6.3% [45.4 mmol/mol]) prevented loss of ankle jerk and light-touch sensation but increased total and CVD-related mortality and severe hypoglycemia. These randomized trials could not establish the optimum GC level for preventing the neurophysiological deterioration in type 2 diabetes.

When patients with recent-onset type 1 diabetes were followed under near-normoglycemia for 24 years, the nerve function declines were completely prevented (4). Therefore, the near-normoglycemia is prerequisite for preventing the neurophysiological deterioration. We recently showed that all neurophysiological and some CNF measures are improved by subnormal HbA_1c levels (HbA_1c 6.5% [47.5 mmol/mol]) over 4 years in patients with poorly controlled type 2 diabetes, while some nerve functions in patients with conventional treatment (mean HbA_1c 7.2% [55.6 mmol/mol]) deteriorated (19). Therefore, standard diabetes care (HbA_1c ≥7.0% [53.0 mmol/mol]) is not beneficial for DPN. In the current study, the treatment without EHC improved only VPT. Because of potential severe hypoglycemia (3), the impact of long-term normalized HbA_1c levels on improving neuropathy outcomes in type 2 diabetes had never been reported.

Tight GC needs to be maintained for >3–5 years to yield benefit (20). The current study followed patients with poorly controlled type 2 diabetes for >4 years on average, and mean HbA_1c levels in patients with EHC over the whole follow-up period and last 2 years were 6.1% (43.2 mmol/mol) and 5.8% (39.9 mmol/mol), respectively. For the last 2 years from end point IPAs were not prescribed, and no hypoglycemic episodes were reported. All neurophysiological tests and most CNF parameters were improved significantly by normalizing HbA_1c levels for 2 years. The results showed that the mean HbA_1c levels over the whole follow-up period played a key role in improving NOMs. The high baseline HbA_1c levels compromised most NOMs as a negative metabolic memory (21,22).

GV may confer additional risk for DPN independent of HbA_1c levels (23,24). In the cross-sectional study in type 2 diabetes, GV estimated by continuous glucose monitoring (CGM) had a close relationship with DPN development (25). We assessed the GV using SD and CV of monthly measured CPPG and HbA_1c levels over 4.3 years. The large GV compromised some NOMs. SD and CV of CPPG compromised most CNF parameters, suggesting that daily glucose fluctuation is harmful on small nerve fibers (SNFs). GV parameters obtained by CGM had been used in a cross-sectional study associating GV with DPN (26). For long-term follow-up study, repeated CGM is required to estimate mean GV, but it is impractical.

Normal HbA_1c levels do not necessarily mean normoglycemia. IGT may play an important role in inducing neuropathy (9,27) and influences NOMs more profoundly than impaired fasting glucose (28). Therefore, determining glucose tolerance category (GTC) by OGTT in patients with normal HbA_1c levels after EHC is required for assessing the influence of remaining glucose intolerance on NOMs. However, OGTT had never been included in clinical studies assessing the benefit of EHC for DPN. In the current study, patients with poorly controlled type 2 diabetes returned to IGT after EHC, while 28.9% of patients returned to NGT. The interval changes in NOMs by EHC were compared between NGT and IGT subgroups. The improvements by EHC in NDS, median nerve MCV, and VPT in the NGT subgroup were significantly greater than in the IGT subgroup. Although insignificant, most other NOMs in the NGT subgroup improved more than those in the IGT subgroup, suggesting that remaining IGT after normalizing HbA_1c levels may compromise NOMs. The reversal of diabetes plays a role in improving diabetes complications. The improvement of NOMs in this study was significant, but the changes were small. This might be due to remaining IGT after normalized HbA_1c levels, negative influence of GV, and metabolic memory of hyperglycemia—or perhaps longer follow-up period is required.

As the SNFs are most likely to respond to interventions for DPN (29), the morphological and functional SNF measures are essential for evaluation of the benefit of normalized HbA_1c levels on NOMs. CNF visualized by corneal confocal microscopy (CCM), predominantly SNF (30), can quantify SNF pathology in DPN (31,32). Subjects having SNF neuropathy show a significant reduction of intraepidermal nerve fiber density and CNF measures (9). In subjects with IGT, CCM reveals dynamic CNF changes related to changing GTC; CNFD and CNFL increase in IGT subjects returning to NGT, while CNFL in subjects developing type 2 diabetes reduces (33). In type 1 diabetes in the context of a normal HbA_1c level (5.9% [41.0 mmol/mol]), CNFD, CNFL, and CNBD were improved in 12 months after SPK (5). However, this study did not determine the GTC after SPK.

In the current study, EHC improved all CNF measures apart from CNFD, with no improvement in those without EHC, indicating that the normalized HbA_1c levels are more beneficial than standard care. In subjects with IGT, most CNF measures were altered compared with NGT subjects.

Normalizing HbA_1c levels for 2 years improved most neurophysiological dysfunctions to levels similar to those of subjects with IGT, while standard care improved only VPT, indicating the superiority of normalized HbA_1c level by EHC compared with standard care for ameliorating neurophysiological dysfunctions. IGT impaired NDS and most neurophysiological functions compared with results in control subjects with NGT.

Because functions of SNFs as well as large nerve fibers were impaired in subjects with IGT compared with NGT control subjects, there was no preferential impairment of SNF functions in IGT. However, we performed a single OGTT at end point in patients with EHC and at baseline in control subjects and in subjects with IGT. There are issues concerning the reproducibility and reliability of a single OGTT in establishing GTCs (34). It is noteworthy that the reproducibility of IGT is lower than for other GTCs (35) which highlights a major limitation of all studies including ours relating GTCs to impaired NOMs (35). Furthermore, we did not follow up control subjects or subjects with IGT. Therefore, we could not compare the time-dependent changes in NOMs among patients with diabetes, control subjects, and subjects with IGT, which might function to mislead us in our attempt to draw conclusions of this study.

Although chronic hyperglycemia clearly plays a causative role in development of DPN in type 2 diabetes, metabolic syndrome components are likely to cause neuropathy (36). Obesity and glucose intolerance have a definite role in neuropathy, whereas hypertension and dyslipidemia have equivocal results. In this study, BMI, SBP, DBP, and triglycerides decreased after GC. The improvements in BMI, DBP, and dyslipidemia during follow-up were significantly associated with the improvement of some CNF and neurophysiological parameters. Therefore, besides the strict GC and smaller GV, the metabolic syndrome components should be controlled for ameliorating DPN even in the context of a normal HbA_1c level. Smoking and alcohol consumption, which may influence DPN (36), had no impact on NOMs in patients with diabetes in this study. The normalized HbA_1c levels for 2 years decreased nephropathy, while retinopathy increased insignificantly. The initial extreme hyperglycemia and substantial HbA_1c reduction might develop the retinopathy (37). Although angiotensin receptor blockers (ARBs) may protect against DPN, ARBs did not improve any NOMs (standard β = 0.011 to −0.347; P = 0.918–0.062).

Assessment of the performance and validity of CCM in longitudinal studies in type 2 diabetes is not an area of study that has investigated. The current study is the only study that has measured the main DPN risk factors, including CPPG, HbA_1c levels, blood pressure, and BMI, monthly, and the lipid profile, eGFR, and ACR (every 3 months). Therefore, the mean clinical factors influencing NOMs are representative; the benefits of normalized HbA_1c levels and improvement in other risk factors to NOMs were reliably evaluated. We estimated the morphological and functional SNF measures. CNFs at baseline in patients and in subjects with IGT were already altered. The normalized HbA_1c levels for 2 years improved all neurophysiological tests of SNFs and most CNF measures.

Our study had many limitations. Firstly, we assessed only baseline and end point NOMs. Ideally, these examinations should be performed annually for assessing trends in neuropathy changes. Secondly, we used SD and CV of CPPG and HbA_1c levels as GV parameters; however, GV can be assessed by several methods, and there is little consensus regarding the optimal method. Thirdly, patients with EHC at end point, control subjects, and subjects with IGT at baseline underwent a single OGTT. Therefore, the reproducibility of the OGTT was poor. Repeated OGTTs are desirable for association of GTCs with neuropathy progress. Fourthly, the prospective follow-up study of a large number of patients over a longer follow-up period is mandatory to reinforce the present results and establish the relationship between the development of neuropathy and IGT. Finally, the patients with type 2 diabetes in the current study had a relatively short diabetes duration. Because the duration of diabetes is one of the major risk factors for DPN and other complications, and severe DPN may be less amenable to intervention, we could not extend the present results to patients with longer duration or advanced neuropathy.

In conclusion, patients with poorly controlled type 2 diabetes returned to IGT by normalizing HbA_1c level for 2 years without hypoglycemia. All neurophysiological tests and most CNF measures improved to IGT levels. However, standard care without EHC improved only VPT despite the same hypoglycemic strategies. This is due to a big difference in body weight reduction during follow-up. The high GV, negative metabolic memory, and metabolic syndrome components are harmful to most NOMs even in the context of normal HbA_1c levels. Along with normal HbA_1c levels by EHC without hypoglycemia, small GV and control of metabolic syndrome components are indicated for preventing and ameliorating DPN. The normalized HbA_1c levels are more effective than standard care for preventing the development of neuropathy but not retinopathy.

Funding. This study received no financial support.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. F.I. designed the study, researched data, and wrote the entire manuscript. M.Tan. performed CCM examinations and neurophysiological tests. A.K. and H.U. gathered the clinical and laboratory data and statistically analyzed all data. M.Tav. advised on the statistical analysis, interpreted the results, and reviewed and revised the whole manuscript. F.I. and M.Tav. are the guarantors of this work, and, as such, had full access to all data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis and interpretation.