Long-Term Clinical Effects of Epalrestat, an Aldose Reductase Inhibitor, on Diabetic Peripheral Neuropathy: The 3-year, multicenter, comparative Aldose Reductase Inhibitor-Diabetes Complications Trial

The ADCT was conducted at 112 medical facilities in Japan between 1997 and 2003. The protocol was approved by the Institutional Review Board of each medical facility. Informed consent was obtained after a detailed explanation of the study purpose and methods.

Adults aged ≥20 years were diagnosed in each facility as mild diabetic peripheral neuropathy based on subjective symptoms, no foot ulcer, and neurological dysfunctions (at least two parameters: MNCV [indispensable] and VPT or Achilles tendon reflex, etc.). Patients were enrolled for study if they had a median motor nerve conduction velocity (MNCV) ≥40 m/s (seemingly reversible) and stable glycemic control (HbA_1c [A1C] ≤9%, with ±0.5% variation in the previous 3 months). Subjects were excluded if their primary cause of neurologic disorder was not diabetes (alcoholic neuropathy, carpal tunnel syndrome, sequelae of cerebrovascular disease, etc.), if they had arteriosclerosis obliterans (ankle brachial pressure index of ≤0.8) or severe hepatic or renal disorder, if they were participating in other interventional studies, or if they were receiving other experimental medications for diabetic neuropathy, prostaglandin E₁ preparations, or any other medication that affects symptoms of diabetic neuropathy.

In this 3-year, open-label study, participants were randomized to epalrestat or a control group. The randomization sequence was generated by an independent organization (Bell System 24, Tokyo, Japan) and randomization was stratified by sex, age (20 to <40, 40 to <65, and ≥65 years), A1C (<7.5 and 7.5–9%), and median MNCV (40 to <50 and ≥50 m/s) using the minimization method. A central telephone and fax service was used to implement the sequence allocation. Epalrestat (50 mg) was orally administered three times daily before each meal (150 mg/day). Both groups continued conventional therapy (diet treatment, hypoglycemic agents, insulin, and hypotensive agents). With the exception of rescue medication, new medication to aid neuropathy control was prohibited.

Study visits occurred at 6 monthly intervals.

The primary end point was change from baseline to study end in median MNCV in the patient’s nondominant arm. The arm was chosen to avoid any bias due to possible lower limb impairment caused by Japanese lifestyle (a tendency to sit straight). Secondary end points included changes from baseline to study end in minimum F-wave latency (MFWL) of the median motor nerve and vibration perception threshold (VPT). Nerve conduction and MFWL were measured using electromyography at 32–34°C based on Kohara et al. (18). F-wave was measured 16 times, and the minimum was adopted. VPT was measured on the medial malleolus in the low extremities using a 128-Hz tuning fork by measuring the number of seconds until the patient could no longer feel the vibrations after the tuning fork was placed on the medial malleolus (15).

The coefficient of variation of the R-R interval at rest (CV_R-R) was measured by electrocardiograph. Variation in resting heart rate during deep breathing was also measured.

Changes in strength of Achilles tendon and quadriceps reflexes were measured. Changes in subjective symptoms of diabetic neuropathy were evaluated using a 100-mm visual analog scale at 3 years, and the percentage change from baseline was calculated. Reductions in symptom levels of ≥50% or <50% were considered an improvement or semi-improvement; no reduction was considered unchanged, and an increase was considered to be aggravation.

Retinopathy assessments were performed by an ophthalmologist using a fundus photograph. Patients’ eyes were graded as no retinopathy, simple (mild or moderate nonproliferative) retinopathy, or either preproliferative or proliferative retinopathy based on the Davis classification (19). Changes after 3 years were defined as improved, unchanged, or deteriorated. Microalbuminuric state was classified at baseline as normal, microalbuminuria, or clinical albuminuria, according to the microalbuminuria diagnostic standards of the American Diabetes Association (20). Change after 3 years was evaluated as improved, unchanged, or deteriorated.

Safety assessments included adverse event (AE) reporting and routine laboratory tests. All reported AEs were recorded with severity graded as mild, moderate, or severe, and relationship to treatment was assigned. AEs were monitored in epalrestat recipients only.

Assuming the population SD of median MNCV is 6 m/s, ∼300 subjects were needed for each treatment group to have a two-sided α level of 0.05 and a power of 0.95 to detect between-group differences.

Efficacy analyses were performed in subjects who had data for at least 1 year, with the last-observation-carried-forward method used. This method corrects bias by using the last available value for a missing value (21,22).

Statistical methods used included χ² tests for nominal scale, Mann-Whitney U tests for ordered categorical scale, two-sample t tests for comparison of mean values between groups, paired t tests for comparison of mean values within groups, and two-way repeated ANOVA for changes in glycemic control. The Cochran-Mantel-Haenszel method was performed in order to calculate the improvement of subjective symptoms when a significant group difference was seen at baseline. ANCOVA was performed for the adjusted data using baseline values as covariates. Normalization for the multiplicity of subgroup analyses was not performed. All analyses were carried out using SAS version 8.02 (SAS Institute, Cary, NC). P < 0.05 was considered statistically significant.

A total of 634 subjects were enrolled and allocated; 31 patients withdrew before starting the trial, leaving 295 and 308 patients in the epalrestat and control groups, respectively. Six (epalrestat) and three (control) patients were excluded for major protocol deviations. Thus, 289 (epalrestat) and 305 (control) patients were included in the analyses. Patient clinical characteristics are provided in Table 1. There were no significant differences between the two groups. Of these, 55 and 31 withdrew after <1 year; reasons for withdrawal were a change in hospital (12 in each group), complications in comorbid illnesses (7 in each group), amelioration of symptoms (2 epalrestat), AEs (20 epalrestat), deterioration in symptoms (7 control), or other (14 epalrestat and 5 control). Both amelioration of symptoms and AEs were likely to be experienced in the epalrestat group only, resulting in the higher withdrawal rate in this group. Additionally, 53 (epalrestat) and 59 (control) patients had ineligible data for the primary efficacy analysis, generally because of the unavailability of an electromyogram or a problem with measuring technique. Thus, the primary efficacy analysis included 181 and 215 patients in the epalrestat and control groups, respectively.

Changes in glycemic control of the patients whose median MNCV, the primary end point, were followed for 3 years. In the control group (n = 214), A1C (means ± SD) at baseline, 1, 2, and 3 years, was 7.1 ± 1.0, 7.2 ± 1.2, 7.3 ± 1.3, and 7.2 ± 1.3%, respectively. Corresponding values in the epalrestat group (n = 180) were 7.2 ± 1.1, 7.2 ± 1.1, 7.3 ± 1.2, and 7.3 ± 1.3%, respectively. One patient in each group was lost to follow-up. There were no significant differences between the two groups at any time points.

Table 2 shows changes from baseline in somatic nerve function at years 1, 2, and 3. After 3 years, no statistically significant deteriorations in median MNCV or VPT were observed in the epalrestat group, with a significant improvement in MFWL of the median motor nerve. Conversely, significant deterioration in these parameters was observed in the control group. To account for between-group differences in baseline median MNCV, an ANCOVA model using baseline values as covariate was applied to the 3-year data; there were significant differences in median MNCV, MFWL of the median motor nerve, and VPT between the groups, largely owing to deterioration of function in the control group.

Analysis of cardiovascular autonomic nerve function showed a significant deterioration in CV_R-R interval at rest after 3 years in the control group but not with epalrestat; however, the between-group difference was not statistically significant (Table 2). When subjects with a CV_R-R interval ≥5% were excluded (23), there was a significant between-group difference at year 2 (change from baseline; means ± SD: control group [n = 146] −0.18 ± 1.06% vs. epalrestat [n = 119] +0.17 ± 1.04%; P = 0.007). The same trend was observed after 3 years (data not shown).

No significant between-group differences were observed for change in variation in resting heart rate (Table 2) or Achilles tendon and quadriceps reflexes (data not shown).

At baseline, there were no significant differences between the groups, with the exception of cramping (P = 0.009, variation adjusted by the Cochran-Mantel-Haenszel method), in terms of the proportion of subjects experiencing symptoms of peripheral neuropathy. After 3 years, the epalrestat group exhibited significantly greater amelioration of the numbness of upper (P = 0.042) and lower (P = 0.030) extremities, paresthesia or hypesthesia (P = 0.015), and cramping (P = 0.024) compared with the control group. There were no statistically significant between-group differences in spontaneous pain of upper and lower extremities, dizziness, coldness, abnormal sweating, and constipation. The levels of improvement in subjective symptoms were as follows: numbness of upper extremities (control group [n = 93]: improved 28.0%, semi-improved 18.3%, unchanged 18.3%, and aggravated 35.5%; epalrestat group [n = 89]: 41.6, 19.1, 13.5, and 25.8%, respectively); numbness of lower extremities (control group [n = 190]: 33.7, 19.5, 15.3, and 31.6%, respectively; epalrestat group [n = 166]: 40.4, 23.5, 16.3, and 19.9%, respectively); paresthesia or hypesthesia (control group [n = 114]: 23.7, 17.5, 17.5, and 41.2%, respectively; epalrestat group [n = 94]: 37.2, 17.0, 19.1, and 26.6%, respectively); cramping (control group [n = 92]: 27.2, 13.0, 26.1, and 33.7%, respectively; epalrestat group [n = 89]: 42.7, 10.1, 16.9, and 30.3%, respectively).

In subjects with no or simple retinopathy at baseline, an improvement at year 3 was seen in 1 of 114 (0.88%) and 7 of 99 (7.07%) subjects in the control and epalrestat groups, respectively. In contrast, 19 (16.67%) and 10 (10.10%) subjects in the control and epalrestat groups, respectively, had a deterioration in retinopathy. The difference between groups in improvement and deterioration rate was statistically significant (P = 0.026).

Of subjects with microalbuminuria at baseline, an improvement at year 3 was seen in 6 of 69 (8.70%) and 7 of 66 (10.61%) subjects in the control and epalrestat groups, respectively, whereas 15 (21.74%) and 8 (12.12%) subjects had a deterioration in albuminuria (not statistically significant), respectively.

Stratified subgroup analyses were performed to examine the relationship between change from baseline in median MNCV and glycemic control, grade of retinopathy, and grade of proteinuria (Fig. 1).

In the control group, median MNCV deteriorated significantly over the 3 years, regardless of level of glycemic control during this study (Fig. 1A). In variation between groups, the epalrestat group showed a statistically significant prevention of deterioration of MNCV at years 2 and 3 in subjects with A1C <7% and at years 1, 2, and 3 in those with A1C ≥7% to <9% for the control group. In subjects with A1C ≥9%, there were no significant differences between the two groups. However, in the variation within groups, though median MNCV of the control group deteriorated significantly compared with baseline at years 2 and 3 (P < 0.050 and 0.010, respectively), the epalrestat group did not show such significant deterioration.

The median MNCV in control group subjects with a baseline A1C <7%, ≥7 to <8%, and ≥8% deteriorated over time (data not shown). However, in epalrestat recipients with baseline A1C <7% (n = 73), there was almost no deterioration, with significant differences compared with the control group (n = 96) at year 1 (P = 0.055), 2 (P = 0.002), and 3 (P < 0.001).

In subjects with no or simple retinopathy at baseline, the median MNCV deteriorated over time in the control group but not the epalrestat group, with significant between-group differences at years 1, 2, and 3 (Fig. 1B). Median MNCV deteriorated over time in subjects with preproliferative or proliferative retinopathy, regardless of group assignment.

In subjects with no proteinuria at baseline, median MNCV deteriorated over time in the control group but not in the epalrestat group. Significant between-group differences were reported at years 1, 2, and 3 (Fig. 1C). Median MNCV deteriorated in subjects who had microalbuminuria or clinical albuminuria at baseline, regardless of treatment. Although deterioration was less in the epalrestat group, this did not reach significance relative to the control group.

In the epalrestat group, 26 of 295 subjects (8.8%) reported AEs; these occurred within the 1st year of treatment in 22 of the 26 subjects. AEs included hepatic function abnormalities (seven cases), gastrointestinal symptoms such as nausea and diarrhea (eight cases), skin rash/eczema (two cases), and one case each of vertigo, light-headedness, dorsal pruritus, hot flushes, hand stiffness, weakness of a lower extremity, edema in a lower extremity, thirst, and cerebral infarction. There were no severe AEs, and no AEs were thought to be directly related to the long-term administration of epalrestat.

Of patients with diabetic neuropathy, 60–70% will develop serious complications that culminate in the amputation of an appendage (24). It is therefore of paramount importance to treat diabetic neuropathies appropriately.

Generally, median MNCV decreases with time in patients with diabetic neuropathy. Partanen et al. (25) reported a significant decrease of 2.9 m/s in the median MNCV over a 10-year period in patients with type 2 diabetes, which converts to a reduction over 3 years of 0.87 m/s. In our study, the control group showed a deterioration of 1.49 m/s over 3 years, exceeding the deterioration in Partanen et al. Conversely, the median MNCV of the epalrestat group has quickened slightly (+0.11 m/s over 3 years). The between-group difference of 1.6 m/s over 3 years indicates that epalrestat can prevent the deterioration of diabetic neuropathy.

MFWL may also be a useful index to evaluate diabetic neuropathy (17,18). In our study, epalrestat was associated with a significantly shorter MFWL of the median motor nerve compared with the control group, which is in agreement with another study with mild neuropathy (17). Other ARIs, including fidarestat, ranirestat, and zenarestat, have also been reported to improve MNCV and MFWL. Although a 52-week study of fidarestat did not show a significant effect on median MNCV, the median nerve F-wave conduction velocity and minimal latency were improved significantly (26). Moreover, in another 52-week study (27), zenarestat considerably improved peroneal MNCV. Similarly, raniresat was shown to improve sensory nerve conduction by ≥1 m/s over 12 weeks (28) and improved peroneal MNCV following 60 weeks of treatment (29).

Short-term treatment with fidarestat (28 weeks) remarkably improved performance in arm and leg VPT (30), an indicator of sensory nerve disturbance (31). Raniresat also improved VPT at the first toe following 60 weeks of treatment (29). In our study, VPT deteriorated over time in the control group but did not change significantly in the epalrestat group, suggesting that epalrestat is effective in preserving the sensory function. This is further supported by the improvement in subjective symptoms (numbness of limbs, sensory abnormality, and cramping) with epalrestat therapy, although such improvements are difficult to evaluate in an open-label study.

The existence of bias cannot be ruled out due to the open-label trial design. However, because the nerve function inspection by the medical technologist and the assessment of the electromyogram by the specialized physician were carried out under masked conditions, it is thought that bias has been minimized.

There was an apparent imbalance in the number of analyzed subjects between the control (n = 215) and the epalrestat (n = 181) groups resulting from more withdrawals in the epalrestat group for events deemed treatment specific.

In performing a stratified analysis using median MNCV as an index, epalrestat was most effective in subjects with good glycemic control. Hyperglycemia-induced hyperactivity of polyol pathway links to the augmentation of the metabolic disorders, like glycation, oxidative stress, and others, contributing to the deterioration of diabetic neuropathy. However, these disorders are not completely caused from the hyperactivity of polyol pathway (5–8). Therefore, our data suggest good glycemic control may be important to keep the better effect of ARI treatment. Epalrestat was also effective in no or mild diabetic retinopathy and urinary microalbumin levels <30 mg/g creatinine. We believe this is the first study to conduct such an analysis and that it provides useful criteria for the selection of subjects for ARI treatment.

ARIs may also be valuable in the treatment of other diabetes complications (5,6). Our results suggest that epalrestat may be effective for preventing the onset or delaying progression of simple retinopathy, supporting previous findings by Hotta et al. (32).

AEs attributed to epalrestat were previously reported in 3.0% of subjects in a 12-week study (13) and 129 of 5,249 subjects (2.5%) in a 3- to 12-month multicenter study (14). The higher incidence (8.8%) of AEs in this study may be due to the longer duration of the study. It should be noted, however, that no particularly severe events were observed, thus confirming the safety of epalrestat for long-term administration.

In conclusion, long-term treatment with the ARI epalrestat is well tolerated and can effectively delay the progression of diabetic neuropathy and ameliorate the associated symptoms of the disease. Effects are particularly evident in patients with good glycemic control and no or mild microangiopathies.

Dr. Yukio Shigeta (Shiga University of Medical Science, Otsu), Dr. Nobuo Sakamoto (Chubu Rosai Hospital, Nagoya), Dr. Toshio Kaneko (Yamaguchi University, Yamaguchi), Dr. Yasuo Akanuma (Institute for Adult Diseases Asahi Life Foundation, Tokyo), Dr. Yoshitomo Oka (Tohoku University Graduate School of Medicine, Sendai), Dr. Motoaki Shichiri (Research Institute of Lifestyle-Related Diseases, Osaka), Dr. Takayoshi Toyota (Tohoku Rosai Hospital, Sendai), Dr. Nigishi Hotta (Chubu Rosai Hospital, Nagoya), Dr. Kempei Matsuoka (Saiseikai Shibuya Satellite Clinic, Tokyo), Dr. Ryuzo Kawamori (Juntendo University School of Medicine, Tokyo), Dr. Yoshihito Atsumi (Saiseikai Central Hospital, Tokyo), Dr. Shinichi Oikawa (Nippon Medical School, Tokyo), Dr. Hideki Kishikawa (Health Care Center of Kumamoto University, Kumamoto), Dr. Jiro Nakamura (Nagoya University Graduate School of Medicine, Nagoya), Dr. Masayuki Baba (Hirosaki University School of Medicine, Hirosaki), Dr. Hitoshi Yasuda (Shiga University of Medical Science, Otsu), Dr. Nobuhiro Yamada (University of Tsukuba, Tsukuba), Dr. Mitsuyoshi Nakashima (Hamamatsu Institute of Clinical Pharmacology and Therapeutics, Hamamatsu), Dr. Isao Yoshimura (Tokyo University of Science, Tokyo), and Dr. Yoshio Goto (Tohoku Kosei-Nenkin Hospital, Sendai).

We are grateful to Dr. Masayuki Baba and Dr. Isamu Ozaki (Hirosaki University School of Medicine, Hirosaki, Japan) for the evaluation of the electromyogram and Dr. Masakatsu Fukuda (NTT West Osaka Hospital, Osaka, Japan) for retinopathy evaluations. We also thank doctors and coworkers at the 112 medical facilities in Japan for provision of clinical data.