Acetyllevocarnitine Hydrochloride for the Treatment of Diabetic Peripheral Neuropathy: A Phase 3 Randomized Clinical Trial in China

Diabetic peripheral neuropathy (DPN) most often affects the nerves of the extremities, leading to various forms of sensory dysfunction (1). Clinical manifestations of DPN include burning, tingling, progressive numbness, and pain, with DPN-related pain ranging from mild discomfort to extreme pain (2). The lifetime prevalence of DPN in adults with diabetes is ∼50% (3). Severe complications of DPN include poor wound healing, persistent ulceration, and amputation, primarily of the lower extremities (4). Such negative outcomes significantly affect patients’ quality of life and contribute to a substantial economic burden for national health care systems (5,6). The increasing incidence of type 2 diabetes (T2D) worldwide highlights the need for improved therapies for the treatment and prevention of DPN (6,7).

With no established treatment for DPN, pharmacological intervention has focused primarily on glycemic control (8), and the severity of DPN in patients with diabetes is associated with glycemic outcomes (9,10). However, the efficacy of such treatments for preventing DPN-related negative outcomes in patients with T2D remains unclear (8,11). Various symptom-based treatments have also been used for DPN (8), including serotonin-noradrenaline reuptake inhibitors, tricyclic antidepressants, calcium channel α₂-δ ligands, opioids, and topical analgesics (11). Although such medications improve quality of life for many patients, disease progression remains unaffected. α-Lipoic acid, spinal cord stimulation, dietary interventions, and physical activity–based lifestyle strategies have provided relief for some patients (12,13). However, the benefits of these nonpharmacological treatments have not been soundly demonstrated in large-scale controlled studies (12,13).

Although certain details of the underlying biological mechanisms of insulin resistance (IR) and T2D remain unclear, hyperglycemia is known to be associated with mitochondrial dysfunction (14–16). In the cytosol, L-carnitine receives acyl groups from acyl-CoA, and the acylcarnitines are imported into mitochondria (17,18). In mitochondria, acylcarnitines donate the acyl groups to free CoA to serve as substrates for oxidative metabolism, and L-carnitine is then exported out of mitochondria (17). Nonadipose cellular lipotoxicity resulting from excess serum fatty acids results in incomplete acyl-CoA oxidation, excess cellular acylcarnitines, and impaired insulin signaling (19,20), but the mechanistic links between lipotoxicity and IR remain unknown. By contrast, L-carnitine insufficiency also contributes to mitochondrial dysfunction (21) and increases the risk of T2D complications (22). Recent meta-analyses have shown that treatment with L-carnitine improves IR and reduces concentrations of fasting plasma glucose and hemoglobin A_1c (HbA_1c) (23–25).

Treatment with acetyllevocarnitine hydrochloride (ALC) has been shown to improve DPN, stimulate nerve regeneration, and provide neuroprotection (26–28). A comparison of ALC versus methylcobalamin for the management of DPN found the efficacy of each to be similar, as demonstrated by statistically similar reductions in both the neuropathy symptom score and neuropathy disability score (29). Other investigations of ALC treatment for pain reduction in DPN have also reported improvement in neurological outcomes (30,31). However, a recent systematic review concluded that the efficacies of ALC reported in these studies were not reliable as a result of various methodological shortcomings in each study (32). Further research using more stringent methods is therefore required to clarify the effects of ALC in patients with DPN.

We present herein the results of our phase 3 clinical trial investigating the efficacy of ALC for treating DPN. Our study included patients with DPN who were receiving antidiabetic therapy for ≥3 months. Previous studies of DPN diagnostic criteria have suggested that the reliability of electrophysiological variables might be greater for some patients than that of the evaluation of clinical signs, symptoms, and assessment scores (8,33). Therefore, we included both clinical assessment and electrophysiological end points in our analysis of the effects of ALC treatment.

This multicenter, randomized, double-blind, placebo-controlled phase 3 clinical trial consisted of a 2-week screening period, a 2-week placebo observation period, a 24-week treatment period, and a 2-week follow-up period. The study was conducted in accordance with the Declaration of Helsinki regarding ethical research involving human subjects. The study protocol was approved by the Ethics Committee of Beijing Hospital (approval no. 2019BJYYEC-224-02; Beijing, China) and was registered with ClinicalTrials.gov (registration no. NCT05319275). Written informed consent was obtained from all patients before participation.

Single-blind screening of patients began in July 2020. Patients meeting all of the following inclusion criteria were eligible for our study: 1) age 18–70 years with type 1 diabetes (T1D) or T2D; 2) antidiabetic therapy for ≥3 months before screening; 3) stable HbA_1c concentration <9.0%; and 4) diagnosis of DPN based on clinical assessment, such as the Michigan diabetic neuropathy score, Toronto clinical neuropathy score (TCNS), neuropathy symptom score, or neuropathy disability score, and/or based on electrophysiological testing, including nerve conduction studies, quantitative sensory testing (QST), or intraepidermal nerve fiber density (IENFD) analysis. Diagnosis of DPN was confirmed at screening and baseline by a TCNS ≥6 (34,35). Patients meeting one or more of the following criteria were excluded from our study: 1) peripheral neuropathy caused by nondiabetic diseases, 2) peripheral artery disease, or 3) acute complications of diabetes <6 months before enrollment or during the study period. The use of oral hypoglycemic drugs or insulin for diabetes and therapies for preexisting nondiabetic diseases were administered externally to our study but were monitored at each study visit.

During the 2-week placebo observation period, all enrolled participants received two placebo tablets (dummy pill) three times per day orally after meals. Patients who did not pass the placebo observation period were withdrawn from the study. The remaining patients were randomly allocated to the treatment groups at baseline. Randomization was accomplished using the stratified block method with four-digit random numbers. Patients received two placebo tablets (placebo group) or two ALC tablets (250 mg/tablet; ALC group) three times per day orally after meals for 24 weeks under double-blind conditions. Emergency unblinding was possible via an interactive web response system.

The modified TCNS (mTCNS) (34) and TCNS (35) were used to evaluate peripheral neuropathy at baseline, week 4 (±7 days), week 8 (±7 days), week 12 (±7 days), week 18 (±7 days), and week 24 (±7 days) study visits. Electrophysiological measurements were used to objectively assess DPN based on nerve conduction velocity (NCV) and response amplitude (RA). The peak amplitudes and latencies of sensory nerves were measured for the ulnar, median, peroneal, and sural nerves. The amplitudes, velocities, and onset latencies were determined only for the peroneal motor nerve. The NCV and RA of nerves in the lower limbs were measured because they were more likely affected by DPN than those innervating the upper limbs (36). Nerve conduction tests were conducted at baseline and week 12 (±7 days) and week 24 (±7 days) of treatment, including motor NCV of the bilateral ulnar, median, and common peroneal nerves and the sensory NCV of the bilateral ulnar, median, and sural nerves. Assessment of DPN-related pain was also performed at baseline using an 11-point numeric rating scale (NRS), with higher scores indicating greater levels of pain, as previously described (37).

The primary end point was overall change in mTCNS at 24 weeks from baseline, as previously described (38–40).

The secondary end points were change in overall mTCNS at 12 weeks from baseline, changes in individual items of the mTCNS at 12 and 24 weeks from baseline, and changes in NCV and RA at 24 weeks from baseline.

A subgroup analysis of TCNS at 24 weeks was performed in which stratification was based on total TCNS at baseline as follows: 6–8 points (mild), 9–11 points (moderate), and ≥12 points (severe).

Patients visited their local study center at baseline and weeks 4, 8, 12, 18, and 24 thereafter for assessment of protocol compliance; reporting of the incidence and severity of adverse events (AEs); reporting of changes in antidiabetic therapy; and evaluation of vital signs, electrocardiograph, hematology, and blood chemistry. All clinical assessments and physiological measurements were performed by trained endocrinologists. The AEs included alterations in fasting blood glucose and HbA_1c concentrations, abnormal electrocardiograph, negative changes in vital signs or laboratory measurements, and changes in background therapy. A serious AE (SAE) was defined as a case requiring hospitalization. The occurrence of an SAE did not automatically result in withdrawal from the study.

Statistical analyses were conducted using SAS (version 9.4; SAS Institute, Cary, NC). A statistical optimization design was adopted based on a previous study (41), which assumed that the difference between groups in the change from baseline in total mTCNS at last observation would be −1.1, and SD was assumed to be 4.0. Based on the reported data and clinical judgment, we assumed the change in total mTCNS at 24 weeks from baseline would be a 1.0 score reduction with an SD of 3.6, and the target sample size was calculated to be 205 patients for each group. Considering an estimated dropout rate of ∼20%, a total of 516 patients were enrolled to achieve at least 80% power with unilateral significance set at 0.025. Primary and secondary end point efficacies were evaluated using an intention-to-treat (ITT) analysis. The full analysis set (FAS) population included all randomly assigned patients who had undergone at least one major efficacy evaluation after at least one dose of the study treatment. Patients without baseline efficacy assessment data were excluded.

The efficacy of ALC treatment in the FAS population was further evaluated using an ANCOVA in which changes in the individual components of the mTCNS served as the dependent variable and baseline mTCNS, treatment group, and study center served as the independent variables in separate analyses for each component. If the upper limit of the 95% CI for the difference in efficacy between the ALC and placebo groups was <0, we concluded the efficacy of ALC treatment was significantly greater than that of the placebo. If end point data were found to be missing for any reason, the missing data were imputed as the last observation carried forward.

Numerical variables are given as mean ± SD and categorical variables as frequency and percentage. χ² and Student t tests were used to compare the mean values of categorical and continuous variables, respectively. Normally distributed baseline parameters were compared between groups using a t test. Any intragroup changes from baseline were assessed using a paired samples t test for normally distributed data and a Wilcoxon signed rank test for data with nonnormal distribution. Effect size was calculated as Cohen d using pooled SDs. All statistical tests were two sided, with the significance level set at a P < 0.05.

Data sets generated and/or analyzed during this study are available from the corresponding authors upon reasonable request.

We recruited 749 patients according to the inclusion criteria, of whom 190 were excluded, leaving 559 patients at enrollment (Fig. 1). During the placebo observation period, 43 patients were excluded for the following reasons: 1) no longer met the inclusion criteria (n = 5), 2) met at least one exclusion criterion (n = 21), 3) voluntarily withdrew from the study (n = 14), and 4) lost to follow-up (n = 3; failed to report after enrollment). A total of 516 patients from 34 institutions were randomly allocated to the ALC (n = 257) and placebo (n = 259) groups. Only eight patients with T1D were enrolled and allocated (ALC n = 2; placebo n = 6). They were removed from our analyses of the ITT and FAS populations. The average age in the ALC and placebo groups was 57.4 ± 8.4 and 56.4 ± 8.7 years, respectively. During the treatment phase of the study, a total of 25 patients were removed from the ALC group, among whom 19 voluntarily withdrew, three withdrew because of an AE, and three were lost to follow-up (Fig. 1). During treatment, a total of 29 patients were removed from the placebo group, among whom 12 voluntarily withdrew, 11 withdrew because of an AE, three were removed because of protocol violations, and three were lost to follow-up. There were 230 ALC and 224 placebo patients who completed the study. The FAS included 254 patients in the ALC group and 251 patients in the placebo group.

Patient age, sex, ethnicity, BMI, HbA_1c, total cholesterol, and triglycerides did not differ significantly between the ALC and placebo groups (P > 0.05 for all) (Table 1). Stratified by TCNS at baseline, the number of patients in the TCNS subgroups did not differ significantly between the ALC and placebo groups (Table 1). The NRS assessment of DPN-related pain at baseline showed that >50% of patients had pain scores ≤1 point and <10% of patients had pain scores ≥6 points, with no patients scoring ≥10 points on the NRS (Supplementary Table 1). At baseline, all participants (n = 505; 100%) were using one or more medications in addition to the study drug (Supplementary Table 2). Among the various medications used, antidiabetic drugs affecting the digestive or metabolic system were used by 493 patients (97.6%), among whom 448 patients (88.7%) used hypoglycemic drugs other than insulin and 239 patients (47.3%) used insulin or insulin analogs.

The mean mTCNS values in the ALC and placebo groups at baseline were 12.4 ± 5.5 and 12.2 ± 5.8 points, respectively (P = 0.728) (Table 1). At 24 weeks from baseline, the mean mTCNS values for the ALC and placebo groups were 5.5 ± 5.0 and 7.6 ± 5.0 points, respectively (P < 0.001) (Fig. 2A). The change in mTCNS at 24 weeks in the ALC group (−6.9 ± 5.3 points) was significantly greater than that in the placebo group (−4.7 ± 5.2 points; P < 0.001) (Fig. 2B). The effect size was 1.31 for the ALC group and 0.85 for the placebo group (Fig. 2C). The effect size for the ALC group represented a 0.65-fold increase over that of the placebo group.

The mean mTCNS values in the ALC and placebo groups at 12 weeks were 7.4 ± 5.0 and 7.9 ± 5.0 points, respectively (P = 0.097) (Fig. 2A). The change in mTCNS in the ALC group (−5.0 ± 4.7 points) at 12 weeks did not differ significantly from than in the placebo group (−4.3 ± 4.8 points; P = 0.120) (Fig. 2B). Effect sizes for the ALC and placebo groups at 12 weeks were 0.95 and 0.79, respectively (Fig. 2C).

The changes in scores for all components at 12 weeks did not differ significantly between the ALC and placebo groups (P > 0.05 for all). The mean changes in mTCNS component values differed significantly at 24 weeks (P < 0.05), with only the change in foot pain failing to reach statistical significance (P = 0.066) (Table 2).

Significant intergroup changes in NCV measurements occurred in the ALC group at 24 weeks from baseline for six of the 18 NCV tests performed (P < 0.05) (Supplementary Table 3). These included changes in NCV for the left ulnar nerve (n = 123; −1.6 ± 6.4 m/s; P = 0.008), left ulnar nerve with leads reversed (n = 127; 3.1 ± 7.0 m/s; P < 0.001), left median (n = 122; −1.4 ± 6.5 m/s; P = 0.028), left median with leads reversed (n = 123; 1.6 ± 5.7 m/s; P = 0.003), left sural nerve with leads reversed (n = 144; 2.1 ± 10.9 m/s; P = 0.030), and right median with leads reversed (n = 124; 1.7 ± 6.7 m/s; P = 0.009). However, only the changes in NCV for the left median with leads reversed and right sural (external ankle sural point) were significantly greater than those in the placebo group (P = 0.050 and P = 0.046, respectively). At 24 weeks from baseline, significant intergroup changes in RA measurements also occurred in the ALC group for three of the 24 RA tests performed (P < 0.05), and right ulnar (finger 5 wrist, reversed leads) RAs were significantly different between the ALC and placebo groups (P < 0.05 for both) (Supplementary Table 3).

In the subgroup analysis, the change in TCNS from baseline for each subgroup after 24 weeks of ALC treatment was significantly superior to that for the corresponding subgroup in the placebo group (P < 0.05 for all) (Supplementary Table 4). Improvement in primary end point efficacy did not coincide with any significant change in the serum concentrations of HbA_1c, total cholesterol, or triglycerides (Supplementary Table 5).

The number of patients experiencing an AE did not differ significantly between the ALC (n = 141) and placebo (n = 131) groups (P = 0.476) (Table 3). Treatment-emergent AE incidences in the ALC and placebo groups were 51.2% and 48.2%, respectively, which included urinary tract infection (5.9% and 6.8%, respectively) and hyperlipidemia (7.9% and 5.2%, respectively). Adverse drug reactions occurred in 10 (3.9%) and nine patients (3.6%) in ALC and placebo groups, respectively. No deaths occurred as a result of an AE. The SAE incidence rate was similar between the ALC group (n = 14; 5.5%) and the placebo group (n = 13; 5.2%; P = 1.000). One patient experienced an SAE that led to prolonged hospitalization, whereas the remainder of the SAEs led to brief hospitalization, none of which were related to the study drug according to the judgment of the investigators.

We investigated the use of ALC for reducing DPN severity in patients who were also receiving antidiabetic pharmacological therapy. We used the mTCNS to evaluate DPN-related symptoms and electrophysiological measurements to assess the neurodegenerative effects of DPN. Using the change in mTCNS at 24 weeks of treatment as the primary end point, our results showed that 24 weeks of ALC treatment significantly improved mTCNS in patients with T2D (Fig. 2A), which was further demonstrated by the 0.65-fold increase in the effect size of ALC treatment, relative to that of the placebo (Fig. 2C). Our findings suggest that ALC treatment in combination with antidiabetic therapy for 6 months is effective at reducing the severity of DPN-related symptoms with no ALC-related AEs (Table 3), indicating that a daily dose of 1,500 mg ALC was well tolerated by our patients.

Our results also showed that reductions in foot pain did not contribute significantly to the changes in mTCNS at 24 weeks of treatment (P > 0.05) (Table 2). These findings suggest that ALC treatment affected mechanisms other than those involved directly in pain reception. Our analysis of the results of electrophysiological measurements also showed that the effects of ALC treatment were limited to improvements in six of the 18 NCV tests and three of the 24 RA tests performed. However, four of these measurements also showed improvement in the placebo group, and six others independently showed improvement in the placebo group with no corresponding improvement in the ALC group (Supplementary Table 3). Given the role of acylcarnitine in maintaining the homeostasis of oxidative metabolism in mitochondria (17), our findings lend support to the existence of multiple intertwined biochemical mechanisms and pathways contributing to neurodegeneration and pain in DPN (42).

The effects of ALC on DPN have not been widely studied. A previous study used the neuropathy symptom score and neuropathy disability score to assess the efficacy of ALC treatment for DPN and found that both were significantly improved in patients with diabetes receiving 1,500 mg ALC per day for 24 weeks (29). Two additional studies used a visual analog scale of symptom severity combined with an analysis of electrophysiological measurements to assess ALC for DPN, in which patients received ALC for 52 weeks (30,31), which was substantially longer than the treatment period in our current study. Both studies reported improvements in DPN-related symptom severity scores. However, whereas one study reported improvements in electrophysiological parameters (30), the other study reported no such improvements (31), indicating the interpretation of electrophysiological measurements in assessing the effects anti-DPN treatment can be confounded by unknown factors.

Also of note in our study were the large effect sizes in both the ALC treatment and placebo control groups (Fig. 2C). Previous studies of ALC for DPN had not reported similar results, despite their patients also receiving antidiabetic therapies. Although mechanisms other than hyperglycemia are known to contribute to the development and severity of DPN (42,43), recent research has shown that some antidiabetic drugs and the long-term maintenance of healthy blood glucose concentrations can contribute to improvements in DPN-related symptoms in at least some patients (44,45). Most of our participants (97.6%) received one or more antidiabetic medications during the study period (Supplementary Table 2). Therefore, it is possible that the effect size for our placebo group represents the combined effects of the antidiabetic drugs used by patients in our study, some of which may not have been widely available at the time of these earlier studies. This would also help to explain the very large effect size observed in our ALC group, which thus represented the combined effects of both ALC and improved antidiabetic therapies, relative to those of previous studies of ALC for DPN.

To date, there exists no effective disease-modifying treatment strategy for DPN. With the known roles of hyperlipidemia, hypertension, obesity, aging, and hyperglycemia in the pathogenesis of DPN (45), controlling for the potential effects of all these factors in a cohort of patients with diabetes presents significant challenges to researchers. These facts suggest that a single-drug therapy intervention is unlikely to be clinically successful for DPN treatment. One recent study evaluated the effects of ALC, superoxide dismutase, vitamin B12, and α-lipoic acid in combination for DPN treatment (46) and found that most indices of DPN were improved, including sural NCV, sural nerve RA, foot pain, and quality-of-life perception (46). Given our findings, it is possible that the effect of ALC treatment constituted an important contribution to their results.

There are certain limitations to our findings. We recruited participants at study centers located throughout a large region of China, across which health care infrastructure varies greatly. We considered the use of other objective end points, including QST and IENFD, but many of our study centers did not have the necessary equipment for QST and/or lacked laboratory staff trained in IENFD biopsy evaluation. Participants’ unwillingness to accept the invasiveness of biopsy-based procedures represented another obstacle. The availability of corneal confocal microscopy for small nerve fiber analysis was also limited. Limitations also exist in our use of mTCNS as the primary end point in our study. The mTCNS is weighted toward sensory neuropathy domains, with items that evaluate large and small fiber sensory neuropathy contributing to the majority of the total possible score from clinician-rated signs.

Additional limitations to our findings include the lack of plasma carnitine measurements at baseline, as well as the absence of an assessment of IR during the study period. If multiple mechanisms are involved in DPN development, a subgroup analysis of ALC efficacy stratified by plasma carnitine concentration at baseline might suggest the physiological basis of differences in DPN pathology. A previous large-scale study of ALC for DPN reported no hypoglycemia in participants who were treated with ALC for 24 weeks (29). Our patients received routine antidiabetic drug therapy externally to our study, but glycemic control was monitored at each study visit, with no indications of substantial changes in IR. However, levocarnitine has been shown to improve IR (23,47). An analysis of the effects of ALC on biochemical markers of IR could make an important contribution to future studies examining the biochemical relationship between DPN and IR.

Our study investigated the use of ALC (1,500 mg/day) for the treatment of DPN in patients receiving antidiabetic therapy and found that ALC significantly improved mTCNS in patients with T2D, compared with that in those receiving placebo treatment, with a 0.65-fold increase in effect size at 24 weeks of treatment. However, significant improvement in the mTCNS foot pain component was not observed, and the results of NCV and RA tests did not indicate a definitive effect of ALC on neurodegeneration. The extent to which the effects of ALC treatment contributed to a clinically meaningful difference in the lives of our patients with DPN is difficult to ascertain based on our results and the current body of evidence from previous studies. However, given the lack of a single effective treatment for DPN, our findings nonetheless suggest that ALC treatment provides at least some incremental improvement in DPN-related symptoms.

Duality of Interest. This trial was supported by Haisco Pharmaceutical Group Co., Ltd. F.L. and D.L. are employees of Haisco Pharmaceutical Group Co., Ltd. No other potential conflicts of interest relevant to this article were reported.

Haisco Pharmaceutical Group Co., Ltd., was involved in the design and performance of the study; the collection, management, analysis, and interpretation of the data; and manuscript preparation, review, and submission for publication.

Author Contributions. L.G., Q.P., Z.C., Z.L., and H.J. were responsible for study conception and design and drafting of the manuscript. L.G., Q.P., Z.C., Z.L., H.J., F.Z., Y.L., W.Q., S.L., J.T., and Y.F. were responsible for the provision of study material or patients and collection and assembly of data. L.G. and H.J. provided administrative support. All authors were responsible for data analysis and interpretation, critical revision of the manuscript for important intellectual content, and final approval of the manuscript. L.G. and H.J. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.