Diabetic peripheral neuropathy (DPN) is one of the most prevalent and debilitating complications of type 2 diabetes mellitus (T2DM), affecting approximately half of the diabetic population and leading to significant morbidity, including pain, foot ulceration, and amputation (1, 2). The global burden of DPN is immense, contributing substantially to healthcare costs and diminished quality of life (3). The traditional pathophysiological model of DPN has centered on chronic hyperglycemia and the subsequent activation of four key metabolic pathways—increased polyol flux, advanced glycation end-product (AGE) formation, protein kinase C (PKC) activation, and hexosamine pathway flux—collectively driving oxidative stress and neuronal damage (4). While landmark trials such as the DCCT and UKPDS established the critical importance of intensive glycemic control, as measured by glycated hemoglobin (HbA1c), in reducing microvascular complications (5, 6), a significant clinical paradox remains. A substantial number of patients achieve recommended HbA1c targets yet still develop or experience progression of DPN (7, 8). This discrepancy underscores the limitation of HbA1c, a measure of average glycemia, in fully capturing the metabolic insults that lead to nerve injury, suggesting the involvement of other dynamic glycemic parameters.

In recent years, glycemic variability (GV) has emerged as a compelling candidate. GV refers to the oscillations in blood glucose levels around the mean, encompassing both hyperglycemic spikes and hypoglycemic troughs (9). The advent of continuous glucose monitoring (CGM) has enabled the precise quantification of GV through metrics such as the mean amplitude of glycemic excursions (MAGE), coefficient of variation (CV), and standard deviation (SD) (10). A growing body of experimental evidence suggests that oscillating glucose may be more detrimental to vascular and neural tissues than sustained hyperglycemia, a phenomenon often described as “glucose toxicity” (11, 12). Mechanistically, acute glucose fluctuations are hypothesized to inflict greater neuronal damage through several interconnected pathways: (1) by exacerbating oxidative stress through recurrent cycles of mitochondrial reactive oxygen species (ROS) generation (13); (2) by activating inflammatory cascades, including the NLRP3 inflammasome and nuclear factor kappa-B (NF-κB) pathways, leading to elevated pro-inflammatory cytokines such as interleukin-6 (IL-6) and tumor necrosis factor-alpha (TNF-α) (14, 15); and (3) by impairing neurotrophic support systems, including nerve growth factor (NGF) and insulin-like growth factor-1 (IGF-1), which are vital for neuronal survival and regeneration (16, 17).

Despite this strong biological plausibility, the clinical and epidemiological evidence linking GV to DPN remains inconsistent. Some cross-sectional studies have reported significant associations between GV indices and DPN prevalence or severity (18, 19), while other longitudinal analyses have failed to confirm an independent relationship after adjustment for confounders (20, 21). These discrepancies may arise from methodological heterogeneity, including differences in study design, population characteristics, GV assessment methods, and DPN diagnostic criteria. Critically, few studies have employed rigorous methods such as propensity score matching (PSM) to adequately control for confounding variables. Furthermore, the potential mediating roles of systemic inflammation and neurotrophic factors in the pathway linking GV to nerve dysfunction remain largely unexplored in clinical cohorts.

To address these critical knowledge gaps, we conducted this propensity score-matched retrospective cohort study. Our primary objectives were: (1) to determine whether GV is independently associated with the presence and severity of DPN in patients with T2DM; (2) to explore a potential dose-response relationship; (3) to investigate the interplay between GV, inflammatory cytokines (IL-6, TNF-α), and neurotrophic factors (NGF, IGF-1); and (4) to synthesize an integrative pathogenic model. We hypothesized that higher GV, independent of HbA1c, is a significant risk factor for DPN and that this association is partially mediated by inflammation and neurotrophic deficiency.

This study provides robust clinical evidence from a propensity score-matched cohort supporting the hypothesis that glycemic variability is an independent and significant risk factor for diabetic peripheral neuropathy in patients with type 2 diabetes. Our principal findings are fourfold. First, after meticulous matching for key confounders including age, diabetes duration, and HbA1c, GV parameters (MAGE, CV, and SD) remained markedly elevated in patients with DPN compared to their non-DPN counterparts, indicating an association with DPN presence. Second, and crucially for assessing severity, multivariable linear regression within this matched cohort demonstrated that both MAGE and CV were independent negative predictors of the composite NCV Z-score. Since reduced NCV is a well-established electrophysiological correlate of DPN severity, this result provides direct evidence for an independent association between higher glycemic variability and more severe neuropathy. Third, mediation analysis revealed that the neurotoxic effect of GV is partly channeled through a pro-inflammatory state, with IL-6 and TNF-α collectively mediating approximately one-third of GV’s total effect on nerve function. Fourth, concomitant neurotrophic dysregulation was observed, characterized by depleted NGF levels in patients with high GV.

Our results align with and extend a growing body of international literature. The independent association between GV and DPN corroborates findings from recent studies. For instance, a cross-sectional analysis by Firouzabadi et al. demonstrated that MAGE was an independent predictor of DPN after multivariable adjustment (18). Similarly, a prospective study by Feldman et al. found that GV metrics derived from CGM predicted the development of neuropathic symptoms over time, whereas HbA1c did not (22). The novelty of our work lies in the application of rigorous propensity score matching, which strengthens causal inference by minimizing selection bias and creating highly comparable groups—a methodological advancement over many previous observational studies. The persistence of a strong GV-DPN association post-matching underscores GV’s potential role as an intrinsic pathogenic factor, not merely a proxy for worse overall glycemic control.

Our study reinforces the growing consensus that DPN pathogenesis in T2DM is multifactorial and extends beyond chronic hyperglycemia. A compelling body of evidence suggests that neuropathy in T2DM may represent a distinct entity from that in type 1 diabetes, with a fundamentally different risk profile. Notably, while intensive glucose control, as measured by HbA1c, has a profound preventive effect on neuropathy in type 1 diabetes, its benefit in T2DM is demonstrably more modest, underscoring the significant contribution of other factors (23). Epidemiologic and laboratory studies consistently implicate components of the metabolic syndrome—including obesity, hypertension, dyslipidemia, and insulin resistance—as critical drivers of nerve injury in T2DM, likely operating through shared pathways of vascular dysfunction, inflammation, and lipotoxicity. Furthermore, the relationship between glucose control and neuropathy may not be uniform across all T2DM patients. Recent evidence using advanced imaging indicates a diametrical effect: in patients with established DPN, higher HbA1c is associated with increased nerve microvascular permeability (a marker of damage), whereas in those without DPN, the relationship is inverse (24). This heterogeneity suggests that the pathophysiological implications of hyperglycemia differ depending on the disease stage and underlying susceptibility. Collectively, these observations argue for a paradigm shift beyond HbA1c-centric management. They highlight the need to identify more precise, dynamic, and potentially stage-specific metabolic risk markers. Our investigation into GV is situated within this context. We posit that acute glucose oscillations represent a distinct and potent metabolic stressor—one that is invisible to HbA1c—and may serve as a crucial missing link in understanding DPN risk and progression in the complex milieu of T2DM.

The dose-response relationship we observed adds considerable weight to the argument for causality. The stepwise decline in nerve conduction velocity across increasing MAGE tertiles suggests a graded and cumulative detrimental effect of glucose oscillations on peripheral nerves. This finding is consistent with the concept of “metabolic memory” or “legacy effect,” where prior glycemic exposure patterns can induce lasting epigenetic and functional changes in vulnerable tissues (25). From a mechanistic standpoint, intermittent hyperglycemia may induce more severe endoplasmic reticulum stress, mitochondrial dysfunction, and impaired autophagy in neurons and Schwann cells than stable hyperglycemia, ultimately overwhelming cellular repair mechanisms (26, 27). Our integrated biomarker data provide clinical support for specific intermediary pathways.

The inflammatory mediation pathway uncovered represents a significant contribution to the field. While systemic inflammation is a recognized component of diabetic neuropathy pathogenesis (15), its specific role as a mediator between GV and nerve damage has been quantitatively underexplored in clinical studies. Our analysis indicates that ~32% of GV’s detrimental effect on NCV operates via elevated IL-6 and TNF-α. This is mechanistically plausible, as experimental models show that oscillating glucose potently activates the NLRP3 inflammasome in vascular endothelial cells and dorsal root ganglia, leading to caspase-1 activation and the release of pro-inflammatory cytokines IL-1β and IL-18 (14). TNF-α is a known direct inducer of neuronal apoptosis and an inhibitor of axonal regeneration (17). Therefore, therapeutic strategies aimed at reducing GV may confer neuroprotection partly by dampening this GV-triggered inflammatory cascade.

The observed neurotrophic dysregulation, particularly the inverse correlation between GV and NGF, provides another compelling mechanistic link. NGF is crucial for the maintenance, survival, and regenerative capacity of small sensory nerve fibers, which are often affected early in DPN (16). Our finding of lower NGF levels in the highest GV tertile suggests that glucose instability may impair the synthesis, axonal transport, or signaling of key neurotrophic factors. The U-shaped relationship observed for IGF-1 is intriguing and warrants further investigation; it may represent a compensatory hormonal response to metabolic stress or reflect altered tissue-specific clearance.

The ROC analysis yielded clinically actionable thresholds (MAGE ≥5.8 mmol/L, CV ≥32.5%) for DPN risk stratification. The superior discriminative power of GV indices over HbA1c alone suggests that incorporating CGM-derived GV metrics into routine clinical practice could significantly refine DPN risk assessment. This is particularly relevant for overcoming therapeutic inertia, as demonstrating a direct link between GV and a tangible, disabling complication may motivate both patients and clinicians to adopt or intensify therapies known to smooth glucose profiles, such as GLP-1 receptor agonists, SGLT2 inhibitors, or advanced insulin delivery systems.

Our findings support an integrated pathogenic framework in which GV is not merely an adjunct to chronic hyperglycemia, but a central driver that concurrently and synergistically amplifies both the metabolic and vascular pathways implicated in DPN (1, 2). In this model, acute glucose excursions initiate a dual assault. First, each glycemic spike generates excessive ROS within neurons and Schwann cells, exceeding the oxidative stress induced by stable hyperglycemia (11, 12). This direct metabolic insult disrupts mitochondrial function, bioenergetics, and neurotrophic support (e.g., NGF depletion). Second, GV potently activates a systemic inflammatory response (elevated IL-6 and TNF-α, as shown here), which, together with oxidant spillover, targets the vasa nervorum (15). This leads to endothelial dysfunction, impaired endoneurial blood flow, and hypoxia (28). The consequence is a vicious, self-reinforcing cycle: the axon is metabolically compromised from within, while its microvascular supply fails from without. The resulting hypoxia further exacerbates mitochondrial dysfunction and oxidative stress, accelerating axonal degeneration and demyelination 3434. This unified mechanism explains the particular potency of GV—it delivers a coordinated attack on both the neuron’s intrinsic health and its extrinsic vascular support. Our mediation analysis, indicating that systemic inflammation accounts for approximately one-third of GV’s detrimental effect on nerve conduction, provides direct clinical validation for the critical intermediary role of this inflammatory-vascular bridge (15, 28). Therefore, GV likely contributes to DPN by orchestrating a pathogenic synergy between metabolic toxicity and microvascular failure. This perspective underscores that therapeutic strategies aimed at stabilizing glucose may confer neuroprotection by mitigating injury across both key domains.

Key strengths of this study include the use of PSM to enhance comparability, objective electrophysiological confirmation of DPN, comprehensive GV assessment via standardized CGM, and the integration of mechanistic serum biomarkers. Several limitations must be acknowledged. First and foremost, the retrospective and cross-sectional nature of our study design precludes any definitive causal inference. Although propensity score matching strengthened comparability and multivariable regression adjusted for confounders, glycemic variability and DPN status were assessed concurrently. Therefore, we cannot determine the temporal sequence: whether elevated glycemic variability precedes and contributes to the development of DPN, or whether the presence of DPN (with its associated changes in autonomic function, physical activity, or disease management) leads to greater glycemic instability. It is plausible that the relationship is bidirectional. Thus, our findings demonstrate a strong and independent association, but not causation. Future prospective, longitudinal cohort studies with serial CGM and standardized neuropathy assessments are essential to establish whether GV is a predictor of DPN incidence or progression over time. Moreover, and importantly, the controlled inpatient environment during which CGM data were collected represents a significant limitation. Patients were hospitalized for routine diabetes management, workup, or stabilization of comorbid conditions. While this ensured standardized meal timing and composition, as well as reduced physical activity, it likely resulted in a more regulated and potentially attenuated glycemic profile compared to free-living conditions. The 72-hour CGM data, therefore, may not fully capture the amplitude and frequency of glucose excursions that occur in daily life with variable diet, exercise, stress, and adherence patterns. Consequently, the association between glycemic variability and DPN observed in our study might be an underestimation of the true effect size in an outpatient setting. Future prospective studies utilizing CGM in ambulatory patients over longer periods are needed to validate our findings in a more ecologically valid context. Additionally, our mechanistic analyses using serum biomarkers and mediation modeling should be interpreted with caution due to sample size and selection constraints. The biomarker subcohort (n=160), while representative in terms of matched baseline characteristics, was limited to patients with available biobanked samples. This selection based on sample availability, though not systematically biased in known clinical variables, could introduce unmeasured confounding and limits the generalizability of the biomarker findings. Consequently, the mediation analysis suggesting that inflammation accounts for approximately 32% of GV’s effect on NCV is hypothesis-generating and requires validation in larger, prospectively designed cohorts where biomarkers are systematically collected for all participants. The modest sample size for these analyses also precluded more complex modeling or exploration of other potential mediators. In addition, the lack of small fiber neuropathy assessment (e.g., via skin biopsy or corneal confocal microscopy) is a notable gap, as small fibers are often affected earlier in the disease process. Finally, regarding our electrophysiological diagnostic criteria. Diabetic peripheral neuropathy typically manifests with sensory abnormalities earlier than motor conduction slowing. Our primary diagnostic criterion relied on abnormal MNCV in at least two nerves. While this approach ensured the identification of definite, objective nerve injury and provided a standardized metric for severity (composite NCV Z-score), it may have inadvertently excluded patients with early-stage DPN whose primary electrophysiological abnormality was reduced sensory nerve action potential amplitude. Notably, our pre-planned sensitivity analysis—which employed a broader diagnostic criterion incorporating sural SNAP amplitude—yielded fully consistent results, underscoring the robustness of our primary conclusions. Nevertheless, the potential for selection bias towards patients with more advanced or motor-predominant neuropathy in the primary analysis cannot be entirely ruled out. This may limit the generalizability of our findings to the earliest, purely sensory stages of DPN. Future studies should employ comprehensive diagnostic criteria that explicitly include sensory nerve parameters to capture the full spectrum of the disease.

In conclusion, this propensity score-matched cohort study provides compelling evidence that glycemic variability is an independent and clinically significant contributor to diabetic peripheral neuropathy, with systemic inflammation serving as a key mediating pathway. Our findings advocate for a paradigm shift in diabetes management—from a singular focus on HbA1c toward a comprehensive, dual-goal strategy that equally emphasizes achieving glycemic stability. Future prospective, multicenter, interventional studies are urgently needed to determine whether therapeutic strategies specifically designed to reduce GV can prevent or retard the progression of DPN more effectively than conventional glycemic control alone.