Interventional effects of exercise on neuropathy in patients with diabetes: a systematic review with meta-analysis

Abstract

Objective

This study aims to systematically evaluate the effects of exercise interventions on neurological dysfunction in diabetic patients, addressing inconsistencies in existing research. It seeks to clarify the impact of different exercise parameters and patient characteristics on nerve function improvements, providing evidence for optimized intervention strategies.

Methods

A search was conducted using PubMed, Web of Science, The Cochrane Library, and Embase databases for randomized controlled trials related to exercise interventions for neurological dysfunction in diabetic patients. The quality of the included studies was assessed according to the Cochrane Handbook for Systematic Reviews. The study utilized RevMan 5.3 to determine effect sizes and assess heterogeneity. Stata 17.0 was employed to evaluate publication bias in the included studies.

Results

The analysis incorporated 9 randomized controlled trials. The meta-analysis demonstrated that exercise leads to improvements in neurological dysfunction among diabetes patients (SMD = 0.61). Subgroup analysis revealed that an 8-week exercise regimen with moderate intensity was particularly effective in enhancing neurological function (SMD = 1.82, 1.58). Improvements were more pronounced in lower limb nerve conduction velocity (SMD = 0.60), peroneal nerve conduction velocity (SMD = 0.86), and sensory nerve conduction velocity (SMD = 0.59). Patients with a disease duration of 5 years or less showed significant improvement (SMD = 0.62).

Conclusion

Exercise effectively improves neurological dysfunction in diabetic patients, with an 8-week, moderate-intensity program showing the greatest benefits, particularly in lower limb and sensory nerve conduction velocity. These findings offer evidence-based guidance for clinical intervention and future research.

Registration and proposal

This study was registered with PROSPEO under No. CRD42024586981; the proposal is available at www.crd.york.ac.uk.

Introduction

Diabetic peripheral neuropathy (DPN) is a prevalent complication among diabetic individuals, resulting from prolonged high blood sugar and metabolic disorders, and often causing chronic sensory or motor nerve damage [1]. Symptoms typically include limb numbness and pain. Research shows that 30–50% of diabetics experience peripheral neuropathy. If left untreated, the condition can worsen, with severe cases necessitating amputation, thereby greatly increasing disability and mortality rates [2]. Moreover, the healthcare costs for those suffering from painful peripheral neuropathy are three times higher than those for patients of the same age without the condition [3], leading to significant physical and financial burdens. Consequently, there is an urgent need to find effective ways to mitigate the severity of peripheral neuropathy in diabetic patients and relieve their suffering.

Exercise has gradually gained the attention of researchers for its convenience, low cost, and few side effects. However, there is no lack of controversy in these studies. Some studies have shown that exercise intervention can reduce the rate of deterioration of neuropathy in diabetic patients by 20% [4]. A study by Azizi et al. [5] found that moderate intensity exercise intervention can improve peroneal nerve conduction velocity in diabetic patients by improving the supply of oxygen and nutrients to nerve tissues. Meanwhile, a study by Reyhanıoglu et al. [6]concluded that balanced training does not activate Na-K-ATPase to a high degree to significantly increase oxygen and nutrient supply, and therefore, does not improve nerve conduction velocity in diabetic patients. Although, aerobic exercise and resistance exercise can improve lipid metabolism by promoting insulin binding to its receptor, improving muscle mass and the oxidative capacity of skeletal muscle, thus improving the body’s lipid metabolism, enhancing the sensitivity of insulin signaling in the skeletal muscle, and then regulating the blood glucose level of diabetic patients, promoting the regeneration of nerves in patients, and delaying the neurological deterioration of patients [7]. However, some scholars believe that exercise may increase the risk of injury in diabetic patients with severe neuropathy [8]. It can be seen that in the current field, the positive effect of exercise on neuropathy in diabetic patients has not reached the formation of a unanimous concept.

This study aims to systematically evaluate the effects of exercise interventions on peripheral nerve function in patients with diabetes mellitus through meta-analysis, providing a robust and unbiased approach to synthesizing available evidence. Although existing studies suggest that exercise may benefit diabetic neuropathy, substantial controversy remains regarding its effectiveness and clinical applicability due to inconsistent findings and methodological limitations. Differences in exercise type, intensity, frequency, and duration across studies further complicate efforts to establish clear clinical conclusions or standardized guidelines. Moreover, previous research has not sufficiently addressed how individual patient characteristics, such as disease duration, age, and comorbidities, may influence responses to exercise interventions. These factors could moderate treatment outcomes, leading to discrepancies in study results. Meta-analysis offers a powerful solution by integrating randomized controlled trial (RCT) data, systematically selecting and analyzing relevant studies to ensure reliable and unbiased conclusions. By synthesizing the current evidence, this study aims to quantify the overall effectiveness of exercise interventions, explore the impact of key intervention parameters, and assess how patient-specific factors influence treatment outcomes. The findings are expected to provide high-quality, evidence-based guidance for clinical decision-making, help refine exercise prescriptions, and shape future research directions in diabetic neuropathy management.

Subjects and methods

This study was reported in strict compliance with the PRISMA reporting checklist and was registered on the official PROSPERO website (registration number: CRD42024586981 ).

Computerized search strategy

Computer searches of PubMed, Web of Science, The Cochrane Library, and Embase English databases were conducted to collect randomized controlled trials on the effects of exercise on neurological function intervention in diabetic patients, and the search timeframe was from the establishment of each database to July 1, 2024 for all searches. The search was conducted by a combination of subject terms and free words. The search terms include Exercise, Diabetic Peripheral Neuropathy, Randomized Controlled Trial, etc., and different databases choose the corresponding combination of subject terms, free terms, and keywords. Taking PubMed as an example, its specific search strategy is as follows (See supplementary document for details):

#1 Exercise [Mesh].

#2 aerobic [Title/Abstract] OR resistance [Title/Abstract] OR training [Title/Abstract] OR sport [Title/Abstract] OR activity [Title/Abstract].

#3 #1 OR #2.

#4 Diabetes Mellitus [Mesh].

#5 diabetes [Title/Abstract] OR diabetic [Title/Abstract].

#6 #4 OR #5.

#7 Peripheral Nervous System Diseases [Mesh].

#8 Neuropathies [Title/Abstract] OR neuropathy [Title/Abstract] OR neuropathy-related symptoms [Title/Abstract] OR peripheral neuropathy [Title/Abstract] OR peripheral nerve [Title/Abstract] OR nerve conduction velocity [Title/Abstract] OR diabetic peripheral neuropathy [Title/Abstract].

#9 #7 OR #8.

#10 Randomized Controlled Trial [Publication Type].

#11 #3 AND #6 AND #9 AND #10.

Inclusion and exclusion criteria

This study followed the PICOS (Population, Intervention, Comparison, Outcome, Study Design) principle to establish clear inclusion criteria, ensuring high-quality evidence while maintaining generalizability.

Inclusion criteria

PICOS Principle	Description
Population	Patients diagnosed with diabetes mellitus, either meeting the World Health Organization (WHO) diagnostic criteria [9] or clinically diagnosed by a qualified general practitioner. Participants were required to be ≥ 18 years old to ensure a focus on the adult population.
Intervention	The intervention group received exercise intervention in addition to standard treatment, while the control group received only routine care. All included studies had to specify the diagnostic criteria used and ensure consistent baseline treatments. Exercise intensity was categorized based on maximum heart rate (HRmax) as follows: Low-intensity: HRmax < 55% Moderate-intensity: 55% ≤ HRmax ≤ 75% High-intensity: HRmax > 75% [10].
Comparison	Studies comparing an exercise intervention group to a control group (routine care without structured exercise).
Outcome	The primary outcome was neurological function, assessed using validated clinical tools such as: Michigan Diabetes Neuropathy Score (MDNS) [11] Michigan Neuropathy Screening Instrument (MNSI) [12] Nerve conduction velocity (NCV) [13] Other clinically recognized neuropathy assessment tools measuring nerve function.
Study design	Only randomized controlled trials (RCTs) were included to ensure high-level evidence quality.

1. Abbreviations: HR_max, Maximum heart rate

Exclusion criteria

To minimize potential bias and ensure data integrity, studies were excluded if they met any of the following criteria:

Study design limitations

Non-randomized controlled trials (e.g., observational studies, case reports).

Reviews, commentaries, editorials, and animal experiments.

Duplicate publications of the same study.

Data quality issues

Studies with incomplete, non-extractable, or missing data.

Studies that failed to report neurological function outcomes.

Intervention and Outcome Irrelevance:

Studies that did not specifically investigate exercise intervention as an independent factor.

Studies that lacked clearly defined neurological assessments or used non-standardized measurement tools.

Screening of literature and data extraction

The 2 researchers screened literature information, extracted information, data calculation and cross-checked independently in accordance with the inclusion and exclusion criteria of the 2 researchers, and in case of disagreement, they went through discussion or third-party opinion until consensus was reached. The researcher extracted relevant information using a pre-established form, including authors, years of experience, basic characteristics of the study population, detailed interventions in the experimental group, and outcome indicators.

Quality assessment of included literature

The researcher assessed the risk of bias of the included literature according to the Cochrane Handbook for Systematic Reviews of Interventions version 6.3, 2022. Items assessed included (1) bias arising from the randomization process, (2) bias arising from deviation from the intended intervention, (3) bias arising from missing outcome data, (4) bias in outcome measurement, and (5) bias in selective reporting of outcomes. The quality of studies included in each domain was assessed on the basis of three options: “high risk”, “low risk”, and “some risk” [14]. If all domains were “low risk”, the outcome was “low risk” and was assessed as (A) If some domains were “some risk” and there was no “high risk”, the outcome was “low risk”. If any of the areas are “some risk” and there is no “high risk”, then the result is “medium risk” and a grade of (B) If one of the areas is “high risk”, then the result is “high risk If one of the domains is “high risk”, the result is “high risk” and the judgment is level (C) The Revised Cochrane risk-of-bias tool for randomized trials (ROB2) was used to summarize the risk of bias for each sub-dimension to obtain the total risk of bias for the included literature [15].

The GRADE profiler software was used to evaluate the quality of evidence [16], and the quality of evidence evaluation for the outcome indicators included five downgrading factors: study limitations, inconsistency of results, indirectness of evidence, imprecision, publication bias. Among them, a downgrade of 3 was considered as very low level of evidence, a downgrade of 2 was considered as low level of evidence, a downgrade of 1 was considered as intermediate level of evidence, and no downgrade was considered as high level of evidence, and the final grade of evidence was categorized into 4 grades: high level, intermediate level, low level, and very low level. Quality ratings were evaluated independently by 2 researchers, and in case of disagreement, a third researcher was asked to adjudicate collaboratively.

Statistical methods

We used RevMan 5.4 software for effect size pooling and subgroup analysis and Stata 17.0 software for sensitivity analysis and publication bias assessment. Due to variations in neurological function measurement methods and units across studies, we selected Standardized Mean Difference (SMD) as the effect size metric. We assessed heterogeneity using P-values and I² statistics and selected either a fixed-effects model or a random-effects model based on the degree of heterogeneity: If P > 0.1 and I² < 50%, indicating low heterogeneity among studies (i.e., differences are primarily due to sampling error rather than true effect variation), the Fixed-Effects Model (FEM) was applied for meta-analysis. If P ≤ 0.1 or I² ≥ 50%, suggesting significant heterogeneity among studies, the Random-Effects Model (REM) was used to account for potential differences in true effects across studies. When significant heterogeneity was present, we conducted subgroup analysis and sensitivity analysis to explore potential sources of heterogeneity [17]. Subgroup analysis was performed based on exercise factors (form, duration, intensity, etc.), different types of neurological function (motor nerve conduction velocity, sensory nerve conduction velocity, and nerve damage), different measurement sites, age groups, disease duration, and evaluation methods. However, due to insufficient classification of exercise duration and frequency in the included studies, this factor was not analyzed in subgroups. Sensitivity analysis was conducted using the Leave-One-Out method, systematically removing individual studies to assess their impact on the overall results and to verify the robustness of the findings. Heterogeneity quantification was based on I² statistics, with 75%, 50%, and 25% representing high, moderate, and low heterogeneity, respectively [18]. Cohen’s d interpretive criteria: 0.2 is a small effect, 0.5 is a moderate effect, and 0.8 is a large effect [19].

Results

Literature search results

A comprehensive search was conducted across PubMed, Web of Science, The Cochrane Library, and Embase, yielding a total of 8,955 records. Following duplicate removal, the remaining studies underwent title and abstract screening to eliminate irrelevant literature. Subsequently, a full-text review was performed to assess eligibility based on the inclusion criteria. After excluding non-compliant studies, a total of 8 studies were finally included in the analysis. A detailed overview of the study selection process is presented in Fig. 1.

Fig. 1

Literature selection flow diagram

Basic characteristics of the included literature

A total of 8 papers with 8 randomized controlled studies were included in the meta-analysis (Table 1) [20,21,22,23,24,25,26,27], involving 508 patients with diabetes and a control group containing both healthy and diabetic populations. The experimental group’s intervention was exercise or exercise combined with medication, and the control group maintained a routine lifestyle without regular exercise. Three studies [22, 24, 27]used scales to assess neurological function, including the Michigan Neuropathy Screening Instrument score (MNSI) [12], the Michigan Diabetes Neuropathy Score (MDNS) [11], and five studies [20, 21, 23, 25, 26]used nerve conduction measurements to assess neurological function. Six studies [20, 22,23,24, 26, 27]reported the duration of the patients’ disease and the others did not.

Table 1 Basic information of included literature

Evaluation of literature quality

The included 8 studies were all randomized controlled trials (RCTs) and explicitly reported the randomization process(Fig. 2). The results indicate that all 8 studies provided detailed descriptions of the randomization procedure, with no observed missing outcome data, measurement bias, or selective outcome reporting. Additionally, 3 studies [20, 24, 26] clearly described the expected intervention measures, with a relatively low risk of bias. Based on a systematic assessment using the ROB2 tool, 3 studies were rated as low risk, while 5 were rated as moderate risk, indicating that the overall study quality is relatively reliable. “Medium-risk” studies typically have certain issues related to blinding and randomization. However, since this study focuses on exercise interventions, it is not feasible for participants to remain completely blinded to whether they are receiving the intervention. As a result, implementing a fully double-blind design presents practical challenges in some studies. Moreover, while some studies may have limitations in allocation concealment or blinding of researchers, the effects of exercise interventions are primarily assessed using objective measures (e.g., nerve conduction velocity, diabetic peripheral neuropathy scores) or standardized evaluation tools, ensuring a relatively high level of reliability in the study results.

Fig. 2

Evaluation of the quality of the included literature

Evidence quality level evaluation

This paper was evaluated for evidence quality level by GRADE profiler software. See Fig. 3.This paper is high-level evidence if it does not meet the downgrading conditions in 5 factors: study limitations, inconsistency of results, indirectness of evidence, imprecision, publication bias. In terms of study limitations, based on the results of the Cochrane risk of bias assessment, most of the information in the literature included in this study was from low risk, and therefore, it was not downgraded. As far as inconsistency is concerned, the results of this study showed a high degree of heterogeneity, and the heterogeneity may derive from different motor elements, different types of neurological functions, different measurement sites, patient age, disease duration, and differences in evaluation methods; therefore, it is not downgraded. As far as indirectness is concerned, this study was included in the diabetic population to systematically evaluate the intervention effect of exercise intervention on neurological function in diabetic patients, therefore, not downgraded. As far as imprecision is concerned, the total number of cases in the literature included in this study was 431, therefore, not downgraded. As far as publication bias is concerned, this study included RCT studies that included non-small samples, therefore, not downgraded.

Fig. 3

Quality level of evidence

Meta-analysis results

Results of nerve function evaluation

Exercise was effective in improving nerve function in patients with diabetic peripheral neuropathy (SMD = 0.61, 95% CI: 0.28–0.95, P < 0.001), reaching a moderate effect size (Fig. 4). The results of the heterogeneity test showed high statistical heterogeneity among the studies (I² = 85%, P < 0.001). SMD = 0.61 indicates that exercise interventions are moderately to highly effective in improving neurological function, suggesting that there is some value in their clinical application.

Fig. 4

Forest plot of meta-analysis on neurofunctional outcomes

Sensitivity analysis

A sensitivity analysis was performed using RevMan 5.4.1 to determine whether the heterogeneity among studies was influenced by a single study. As shown in Fig. 5, the analysis involved systematically excluding each study one at a time. The results indicated that Snehil Dixit 2013 and Snehil Dixit 2014 were likely contributors to the observed heterogeneity, as the I² value increased from 13 to 86% upon their exclusion. A possible explanation for this heterogeneity is the higher frequency of exercise interventions in these two studies, which ranged from 3 to 6 times per week, whereas in other included studies, the exercise frequency did not exceed 3 times per week. Additionally, the age of participants in these two studies was not reported, which may have further contributed to the variability in results. These factors suggest that exercise frequency and participant age could be potential sources of heterogeneity in this meta-analysis.

Fig. 5

Sensitivity analysis of the impact of exercise interventions on neurofunctional outcomes in patients with diabetes

Publication bias of included literature

Observation of the funnel plot (Fig. 6) shows a relatively symmetrical distribution. However, the Egger’s test result (t = 3.45, P = 0.003, P < 0.05) suggests the potential presence of publication bias, language bias, or small-study effects. To further assess the impact of bias on the meta-analysis results, we applied the Trim and Fill method and conducted two iterations of meta-analysis. The results indicated that the adjusted effect size and confidence intervals remained largely unchanged, suggesting that although some publication bias may exist, its impact on the overall study results is minimal, and the conclusions remain robust.

Fig. 6

Funnel plot of publication bias

Subgroup analysis

To further explore the sources of heterogeneity, subgroup analyses were performed for exercise elements such as form, period, and intensity (the classification of exercise duration and frequency in the included studies herein was insufficient for subgroup analyses), different types of neurological function (motor nerve conduction velocity, sensory nerve conduction velocity, and neurological injury), different sites of measurements, and different ages, disease duration, and modes of evaluation (Table 2).

Table 2 Subgroup analysis results for heterogeneity and effect size

Different exercise elements

Subgroup analyses were performed by grouping exercise cycles and exercise intensity. The results showed negligible heterogeneity for, 12 weeks (I² = 0%, P = 0.69), light-moderate intensity (I² = 0%, P = 0.87), and high-intensity (I² = 0%, P = 0.89). higher heterogeneity was observed for 8 weeks (I² = 95%, P < 0.001), and moderate intensity (I² = 92%, P < 0.001). Thus, exercise period as well as exercise intensity may be the source of heterogeneity. In addition, the intervention effects at 8 weeks (SMD = 1.82, 95% CI:0.36–3.28) and 12 weeks (SMD = 0.22, 95% CI:0.07–0.37), moderate intensity (SMD = 1.58, 95% CI:0.49–2.67) and high intensity (SMD = 0.25, 95% CI: 0.01–0.50) were statistically significant.

Different types of nerve functions

Subgroup analysis with different types of nerve function showed negligible heterogeneity in motor nerve conduction velocity (I² = 42%, P = 0.06). Sensory nerve conduction velocity (I² = 80%, P < 0.01) and nerve damage (I² = 96%, P < 0.01), therefore, different types of nerve function were the source of heterogeneity. The intervention effect of exercise on motor nerve conduction velocity (SMD = 0.25, 95% CI:0.02–0.49) and sensory nerve conduction velocity (SMD = 0.59, 95% CI:0.04–1.15) was statistically significant.

Measurement sites

Subgroup analyses were performed by grouping different nerve sites. The results showed that the heterogeneity of nerve conduction velocity in the upper limb was negligible (I² = 0%, P = 0.89), and the heterogeneity of nerve conduction velocity in the lower limb was higher (I² = 80%, P < 0.001). And the intervention effect of exercise on lower limb nerve conduction velocity (SMD = 0.60, 95% CI:0.11–1.09) was statistically significant. Further analysis showed that the heterogeneity of median nerve (I² = 0%, P = 0.33), ulnar nerve (I² = 0%, P = 0.86), tibial nerve (I² = 0%, P = 0.86) was negligible, and the heterogeneity of peroneal nerve was higher (I² = 86%, P < 0.001). And the intervention effect on peroneal nerve (SMD = 0.86, 95% CI:0.11–1.60) was statistically significant.

Age

Subgroup analyses were performed with age groupings. The results showed that the heterogeneity was negligible in age 40–49 (I² = 0%, P = 0.77) and 50–59 (I² = 33%, P = 0.18), and some heterogeneity existed in age 60–69 (I² = 56%, P = 0.05), thus, age was the source of heterogeneity. However, the intervention effect of exercise on neurologic function in patients with diabetic peripheral nerve pain in the three age groups was not statistically significant.

Disease course

Subgroup analyses were performed with disease duration subgroups. The results showed negligible heterogeneity for disease duration of 11–15 years (I² = 0%, P = 0.72), and higher heterogeneity for disease duration of ≤ 5 years (I² = 85%, P < 0.01) and 6–10 years (I² = 93%, P < 0.01), thus, disease duration was the source of heterogeneity. The effect of exercise on the intervention was statistically significant for disease duration of ≤ 5 years (SMD = 0.62, 95% CI:0.07–1.17).

Evaluation approach

Subgroup analyses were performed by grouping the evaluation modalities. The results showed that there was heterogeneity in both the dosage scale (I² = 96%, P < 0.001) and the nerve conduction velocity meter (I² = 64%, P < 0.001), and therefore, the evaluation modality was the source of heterogeneity. The sensitivity of the nerve conduction velocity meter (SMD = 0.35, 95% CI:0.11–0.59) to measures of nerve function in patients was statistically significant.

Discussion

Key findings

This study demonstrates that exercise is effective in improving nerve function in diabetic patients, aligning with previous findings [28]. The beneficial effects of exercise may be attributed to its role in enhancing mitochondrial function [29], promoting axonal [30] and myelin sheath regeneration [31] improving neurotrophic support [32], and increasing muscle fiber cross-sectional area [33], all of which are crucial for mitigating peripheral neuropathy. However, in contrast to some previous studies, our results suggest that an 8-week exercise intervention yielded greater improvements, whereas Nadi et al. [34]. reported that a 12-week intervention was more effective. This discrepancy may be attributed to differences in study populations, as Nadi et al.’s study included only female patients with diabetic neuropathy. Since hormonal variations and physiological differences in women can influence functional recovery following exercise interventions [35]. female patients may require a longer adaptation period to fully benefit from exercise compared to males. These findings highlight the importance of individualized exercise prescriptions that consider sex differences and physiological adaptation time to optimize therapeutic effects.

This study systematically evaluated the effects of exercise interventions on peripheral neuropathy in diabetic patients by analyzing a total of eight studies. The quality of the included studies was assessed using the ROB2 tool, resulting in four studies classified as A-grade and five as B-grade, indicating generally good overall quality. In evaluating the quality of evidence, no significant downgrading factors were identified among publication bias, inconsistency, indirectness, imprecision, and study limitations; thus, the overall evidence quality was rated as high. However, several limitations existed: due to the limited number of studies included, subgroup analyses based on patient age did not show significant moderating effects, though the overall effect size remained significant, potentially impacting result reliability. Moreover, substantial heterogeneity (I² = 85%, P < 0.001) existed among studies, likely arising from differences in exercise parameters, types of neurological function assessed, measurement sites, disease duration, and evaluation methodologies. These factors should be considered cautiously, as exercise intervention effects may vary across diabetic neuropathy patients of different ages.

Optimal exercise duration and intensity for neuropathy improvement

Our analysis of the exercise elements revealed that an exercise cycle of 8 weeks and a moderate intensity of exercise may be better for improving peripheral nerve function improvement in diabetic patients. For the exercise cycle, studies have shown that exercise may regulate blood glucose changes by improving the sensitivity of insulin target tissues in diabetic patients [36]. Importantly, the researchers concluded that continuous exercise for at least 8 weeks may help patients to better control the abnormal changes in blood glucose [37], and to improve neurological neuropathy in diabetic patients under the dual effect of exercise intervention and good blood glucose control. Diabetic patients usually have dysregulation of the PI3K/AKT signaling pathway [38], leading to a decrease in its insulin-mediated function, and the PI3K/AKT signaling pathway is closely related to PPARγ [39], and moderate-intensity exercise may regulate patients’ glucose-lipid metabolism and inflammation through the PPARγ/ Pl3K/AKT signaling pathway [40], thus helping to improve neurological neuropathy in diabetic patients. Fisher et al. [41], showed that an 8-week exercise intervention was effective in improving neurologic function in patients with diabetic neuropathy. In animal studies, it was found that hippocampal FNDC5 protein expression was increased in diabetic mice after 8 weeks of exercise intervention, hippocampal neuronal damage was significantly reduced, neuroinflammatory response was decreased, and the prominent plasticity was enhanced, which in turn slowed down the pathologic changes in nerve function [42].

Exercise-induced improvements in nerve conduction velocity

Further detailed analysis of nerve function revealed that exercise interventions improved lower limb nerve conduction velocities in diabetic patients, with the peroneal nerve conduction velocity showing the most significant improvement. Previous studies suggest that sensory nerves may be more responsive to metabolic and circulatory changes and exhibit higher plasticity [43], potentially leading to greater sensitivity to exercise stimuli. Consistently, the enhancement of sensory nerve conduction velocity by exercise appeared more pronounced compared to motor nerve conduction velocity. Specifically, improvements were observed in peroneal nerve conduction velocity, while the improvement in tibial nerve conduction velocity was less clear. As the peroneal nerve primarily controls sensation and movement in the leg and dorsal foot, it may experience more frequent stimulation during exercise interventions, potentially explaining its superior responsiveness to exercise [44].

Impact of disease duration on exercise efficacy

Additionally, exercise-induced improvements in neurological function were more significant in diabetic patients with a disease duration of ≤ 5 years. Diabetic neuropathy is associated with increased deposition of Advanced Glycosylate-End Products (AGE) [45] and elevated concentrations of low-affinity nerve growth factor (NGF) receptors [46], both implicated in nerve damage. In patients with shorter disease duration, the deposition of AGEs and NGF concentrations might be comparatively lower, enhancing the efficacy of exercise in reducing these factors. Early-stage diabetic patients, typically exhibiting less severe neurological impairment, may respond more readily to exercise interventions. Their nervous systems may have also gradually adapted to prolonged hyperglycemia through prior lifestyle adjustments and medication, potentially enhancing their tolerance and adaptability to exercise interventions [47, 48]. Conversely, diabetic patients with longer disease durations often exhibit impaired immune responses and reduced recovery capacities [49], possibly diminishing their responsiveness to exercise interventions.

Clinical significance

This study demonstrates that exercise intervention has a significant effect on improving nerve function in patients with diabetes, providing important evidence for non-pharmacological interventions in diabetic peripheral neuropathy (DPN). Exercise not only helps to slow the progression of nerve damage but also enhances sensory function, improves nerve conduction, and alleviates neuropathic symptoms, ultimately leading to better motor function and quality of life for patients.

Furthermore, the study highlights the clinical value of optimizing exercise parameters, revealing that an 8-week, moderate-intensity exercise program yields the best outcomes. These findings provide a scientific basis for personalized exercise prescriptions, assisting clinicians in designing safe and effective exercise plans for diabetic patients to improve adherence and maximize the neuroprotective effects of exercise. Therefore, exercise intervention should be integrated into diabetes management as a key strategy to enhance nerve function and reduce the risk of diabetic complications, making it highly valuable for clinical application and public health promotion.

Study limitations

Despite the rigorous systematic review and meta-analysis methods used to evaluate the effects of exercise interventions on patients with diabetic peripheral neuropathy (DPN), this study still has the following limitations:

• (1) High Heterogeneity Among Studies.

There is significant variation in the type, intensity, frequency, and duration of exercise interventions, which may affect the stability of the overall effect. Differences in participants’ age, duration of diabetes, severity of DPN, and baseline physical activity levels may influence their response to exercise interventions. Inconsistencies exist in outcome measurement methods; some studies use nerve conduction velocity (NCV), while others rely on clinical scoring scales, potentially leading to heterogeneity in effect estimates. Differences in study design and methodology, such as randomization methods, control group settings, and blinding implementation, may affect study quality and comparability.

• (2) Variability in Overall Study Quality.

Some studies have a risk of bias, including unclear descriptions of randomization methods, inadequate allocation concealment, or imperfect blinding, which may impact the internal validity of the results. Due to the nature of exercise intervention studies, participants cannot be completely blinded, which may lead to measurement bias in assessment indicators.

Future perspectives

Future research should focus on optimizing specific exercise parameters, including the ideal type, intensity, frequency, and duration of exercise interventions to maximize their effectiveness in patients with diabetic peripheral neuropathy (DPN). Additionally, large-scale, long-term randomized controlled trials (RCTs) are needed to verify the sustained benefits of exercise and explore its applicability across different subtypes of diabetic patients. Moreover, integrating multimodal objective assessment tools (such as neurophysiological testing and biomarkers) could enhance the scientific rigor of studies, providing a more precise understanding of the mechanisms underlying exercise-induced improvements in nerve function and offering stronger evidence for clinical applications.

Conclusion

Exercise has a significant effect on improving peripheral neuropathy in patients with diabetes mellitus. Its efficacy is influenced by the duration of the disease, demonstrating a dose-response relationship with exercise duration and intensity, and yielding differentiated effects on various aspects of nerve function. These findings provide evidence-based support for clinical interventions in diabetic patients. However, due to the high heterogeneity among studies and potential biases, the generalizability of these conclusions remains limited. Additionally, research on the effects of exercise on nerve function in different subtypes of diabetic patients is still insufficient. Therefore, more high-quality, rigorously controlled studies are needed to further refine exercise parameters and verify their long-term effectiveness and clinical applicability.