Association between hypoglycemia unawareness and cardiovascular autonomic neuropathy in patients with type 2 diabetes mellitus: a prospective cohort study

Abstract

Background

Diabetes has emerged as the third most significant global public health challenge. Among patients with type 2 diabetes mellitus (T2DM), cardiovascular autonomic neuropathy (CAN) represents a prevalent yet frequently overlooked complication. Hypoglycemia unawareness (HU) poses a persistent challenge in glycemic management for T2DM patients, yet its relationship with CAN remains inadequately understood. The aim of this study was to explore the correlation between HU and CAN through a prospective cohort study to provide a basis for early screening and intervention of CAN.

Methods

This study was a prospective cohort study, which finally included 223 T2DM patients who were admitted to the Department of Endocrinology of Hefei Hospital of Anhui Medical University from December 2020 to December 2024.Based on the subgroups with and without HU, Kaplan-Meier survival analysis model was constructed to clarify the variability of new-onset CAN between groups, and the log-rank test was used to assess the differences between groups. And further landmark analysis was performed on the survival curves. The correlation between HU and CAN was assessed using the COX proportional risk model with the no hypoglycemia group as the reference group.

Results

Among the 223 patients analyzed, 143 (64.1%) subsequently developed CAN. Compared to those without CAN, patients in the CAN group exhibited significantly higher rates of diabetic peripheral neuropathy (DPN), a history of stroke, and smoking, alongside increased glycemic variability (SD). The incidence of new-onset CAN was markedly higher in the hypoglycemia unawareness group than in the non-hypoglycemia group. COX regression analysis revealed that HU is an independent risk factor for CAN in T2DM patient. Subgroup analyses and sensitivity analyses further validated the results.

Conclusion

The HU is an independent risk factor for CAN in patients with T2DM, and the effect of HU on CAN is more pronounced with the prolongation of the disease course. Early screening and intervention for CAN should be carried out in patients with HU in order to reduce the death rate associated with CAN.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13098-025-01794-6.

Introduction

Diabetes has ascended to become the third most formidable global public health issue, following cancer and cardiovascular diseases. According to data released by the International Diabetes Federation (IDF) Global Diabetes Overview, 10th Edition (2021) [1]approximately 537 million adults (aged 20–79 years) worldwide will have diabetes in 2021; that number is projected to rise to 643 million by 2030; and to 783 million by 2045.Total global diabetes-related health expenditures will continue to grow in the future, reaching a projected 1.03 trillion US dollars by 2030 and 1.05 trillion US dollars by 2045. Prolonged hyperglycemia triggers a spectrum of chronic complications, which significantly contribute to the increased morbidity and mortality rates among T2DM patients. Among these, cardiac autonomic neuropathy (CAN) is a common yet often overlooked condition, with its prevalence exhibiting notable variability [2]. The primary reason for its neglect lies in its subtle and nonspecific early clinical manifestations, which often only become apparent in advanced stages [3]. Symptoms may include orthostatic tachycardia, resting tachycardia, exercise intolerance, and even silent myocardial infarction, with severe cases potentially leading to sudden cardiac death. Current clinical practices for CAN assessment rely on symptom and sign evaluation, which are not easily scalable. Identifying early predictive factors is thus crucial to prevent the progression of CAN to symptomatic stages.

Hypoglycemia remains a critical limiting factor in diabetes management [4]. Its symptoms serve as warnings for patients to take corrective measures to avoid severe hypoglycemia, which can lead to hypoglycemic coma or delirium. According to the 2025 edition of the ADA’s most recent guidelines for the hypoglycemia Sect. [5]hypoglycemia is classified into 3 grades, which are grade 1 hypoglycemia (glucose concentrations < 70 mg/dL (< 3.9 mmol/L) but ≥ 54 mg/dL (≥ 3.0 mmol/L)), grade 2 hypoglycemia (defined as glucose concentrations < 54 mg/dL [< 3.0 mmol/L]), grade 3hypoglycaemia (defined as altered mental and/or physical functioning that requires assistance from others to recover, regardless of blood glucose level). Hypoglycemia unawareness (HU) may occur in all 3 grades of hypoglycemia due to patients exhibiting impaired counter-regulatory responses to hypoglycemia and/or impaired awareness of hypoglycemia. HU involves a diminished or absent perception of low blood glucose, often resulting from recurrent or severe hypoglycemia, which can further lead to autonomic system failure (hypoglycemia-associated autonomic failure, HAAF) [6]. Traditional perspectives suggest that hypoglycemia can cause neurological impairment and cardiovascular consequences [7–9].Additionally, studies have indicated [10] that CAN is independently associated with a higher risk of recurrent hypoglycemia in T2DM patients. To our knowledge, research on the relationship between HU and CAN in T2DM patients remains scarce.

This study leverages prospective cohort data from T2DM patients monitored using continuous glucose monitoring systems (CGMS) to analyze the progression of CAN over time and explore the relationship between HU and CAN.

Materials and methods

Study population

This prospective cohort study was conducted among patients consecutively recruited from the Department of Endocrinology and Metabolism at Hefei Hospital, Affiliated with Anhui Medical University, between December 2020 and December 2024. Participants were selected based on the following criteria. (The specific process is shown in Fig. 1.)

Fig. 1.

Flowchart

Inclusion Criteria:

1. Age ≥ 18 years, meeting the 2025 version of the American Diabetes Association (ADA) [11]diagnostic criteria for T2DM.”

2. Completion of CGMS monitoring, the Ewing test, and provision of complete data on glycated hemoglobin A1c (HbA1c) and fasting blood glucose (FBG) during the initial hospitalization.

3. A stable antidiabetic treatment regimen for the past three months.

Exclusion Criteria:

1. Other types of diabetes mellitus (e.g., gestational diabetes mellitus or type 1 diabetes mellitus).

2. Recent acute complications of diabetes mellitus, such as diabetic ketoacidosis, hyperosmolar coma, lactic acidosis, and hypoglycemic coma.

3. Patients with a history of gastroparesis/constipation/urinary retention/beta- and alpha-blockers/tricyclic antidepressants/stress.

4. Abnormalities of the thyroid function.

5. Patients with proliferative retinopathy who are at risk for performing the ewing test.

HU and CAN assessment and grouping

All patients underwent continuous 72-hour subcutaneous interstitial glucose monitoring using CGMS. The CGMS sensor generated 288 continuous glucose readings per day, which were calibrated at least four times daily using subcutaneous glucose measurements. Parameters such as the coefficient of variation (CV), mean glucose (MG), and standard deviation (SD) were derived from the CGMS data. According to the 10th International Conference on Advanced Technologies & Treatments for Diabetes [12]a hypoglycemic event is defined as follows: the event begins when the CGMS reading remains below 3.9 mmol/L (70 mg/dL) for at least 15 min and ends when the reading returns to ≥ 3.9 mmol/L (≥ 70 mg/dL) for at least 15 min. The diagnosis of HU was determined by combining CGMS-derived glucose data with the assessment of an endocrinologist with many years of clinical experience, where hypoglycemia-associated symptom [13] included hunger pangs, panic attacks and hand tremors, weakness of the limbs, profuse sweating, and even altered sanity. During CGMS monitoring, all participants were instructed to adhere to a standardized diet and regular meal times: breakfast between 6:30–7:30, lunch between 11:00–12:00, and dinner between 17:00–18:00.The standardized diet aimed to provide a daily caloric intake of 25 kcal/kg, with 55% carbohydrates, 17% protein, and 28% fat¹⁴.

The diagnosis of CAN was based on the Ewing test, which included the following components:

1. Resting heart rate: A resting heart rate > 100 beats per minute was considered abnormal (excluding arrhythmias and cardiac insufficiency).

2. Handgrip test: Blood pressure was measured immediately after sustained forceful handgrip for 3 min. A systolic blood pressure increase of ≥ 16 mmHg (1 mmHg = 0.133 kPa) was considered normal, while an increase of ≤ 10 mmHg was abnormal.

3. Orthostatic blood pressure difference: Blood pressure was measured in the supine position at rest, followed by immediate standing. Blood pressure was measured again within 1 min. A systolic blood pressure drop of > 20 mmHg or a diastolic drop of > 10 mmHg was considered abnormal.

4. Orthostatic heart rate difference: A heart rate increase of > 10 beats per minute was normal, while an increase of ≤ 10 beats per minute was abnormal.

After each Ewing test component, participants rested for at least 2 min before proceeding to the next test. Abnormal results in components (1), (2), and (4) were each scored as 1 point, while an abnormal result in component (3) was scored as 2 points. CAN was diagnosed if two or more components were abnormal, scoring ≥ 2 points [15].

It is also categorised into CAN + group (with new CANs) and CAN- group (without new CANs) based on the subsequent presence or absence of new CANs.

Baseline measurements

Basic clinical information, including age, gender, duration of diabetes, smoking and alcohol history, stroke, diabetic kidney disease(DKD), diabetic retinopathy(DR), diabetic peripheral neuropathy(DPN), and medication prescriptions, was extracted from standardized electronic inpatient medical records. Training personnel measured height, weight, and blood pressure upon admission using standardized protocols. Height and weight were measured with a stadiometer to the nearest 0.1 cm, with participants wearing light clothing and no shoes. Body mass index (BMI) was calculated as weight in kilograms divided by the square of height in meters. Blood pressure was measured three times using a standard mercury sphygmomanometer after a 5-minute rest, and the average readings were recorded. Blood samples were collected the following morning after at least 10 h of fasting. Biochemical parameters included FBG, fasting C-peptide (FCP), HbA1c, total cholesterol (TC), triglycerides (TG), low-density lipoprotein cholesterol (LDL-C), high-density lipoprotein cholesterol (HDL-C), and urinary albumin-to-creatinine ratio (UACR). Imaging indicators included macrovascular plaque (defined as carotid plaque or (and) lower limb arterial plaque), which was determined by imaging. (See supplementary material S1 for details of hypertension, diabetic nephropathy, diabetic retinopathy and macrovascular plaques)

Prospective follow-up

Patients initially enrolled in the study were followed up every three months, during which trained personnel conducted the Ewing test to assess the progression of CAN.

Outcome

The primary outcome was the development of a new-onset CAN. New-onset CAN was defined as participants who did not meet the diagnostic criteria for CAN at baseline but were subsequently diagnosed with CAN during follow-up. The follow-up period began at enrollment and continued until the occurrence of new-onset CAN, withdrawal from the study, or the end of the study.

Follow-up event definitions

Effective follow-up: defined as non-CAN patients who had complete collection of CGMS-derived glycemic data at the beginning of the study and who had undergone the Ewing test and maintained follow-up until the end of the study.

Ineffective follow-up: defined as (1) patients who voluntarily withdrew, were lost for more than 3 months, or could not be followed up due to other unavoidable circumstances; and (2) patients who were missing specific information about the Ewing test during the follow-up process, or information about the outcome of whether or not CAN occurred.

Statistical methods

Normality tests were performed for continuous numerical variables. Normally distributed data were expressed as Mean ± SD, while non-normally distributed data were presented as Median (IQR). Categorical data were expressed as n (%). Based on the presence or absence of HU, a binary variable was established, with the non-hypoglycemia group as the reference and the HU group as the observation group. Differences between groups with and without new-onset CAN were assessed using the chi-square test (for categorical variables) and the Kruskal-Wallis H test (for skewed data). Kaplan-Meier survival analysis curves were used to analyze the incidence of the primary outcome between the two groups, and landmark analysis was further applied to the results. The COX proportional hazards model was employed to evaluate HU’s hazard ratio (HR) as a risk factor for the outcome event. In the multivariate COX regression model, age, gender, BMI, duration of diabetes, diabetic kidney disease (DKD), diabetic retinopathy (DR), diabetic peripheral neuropathy (DPN), stroke, hypertension, smoking history, oral hypoglycemic agent use, insulin use, fasting C-peptide, HbA1c, LDL-C, and coefficient of variation (CV) were included as confounding factors. Subgroup analyses were further conducted based on age, gender, history of stroke, hypertension, smoking, and insulin use, with the likelihood ratio test used to assess interactions between HU and subgroup factors. Additionally, sensitivity analysis was performed on 209 T2DM patients who were not using β-blockers to validate the robustness of the results. Statistical analyses were performed using R Studio (version R4.4.2) and Free Statistics V2.0. All tests were two-tailed, and a P-value < 0.05 was considered statistically significant.

Results

Clinical characteristics of the study population

In the final analysis, 223 patients with T2DM were included, of whom 80 did not subsequently develop cardiovascular autonomic neuropathy (CAN- group) and 143 subsequently developed cardiovascular autonomic neuropathy (CAN + group). Clinical characteristics of participants’ baseline status based on the presence or absence of subsequent development of CAN subgroups are shown in Table 1. Compared with the CAN- group, patients in the CAN + group had a significantly higher prevalence of DPN (90.9% vs. 78.8%, P = 0.011), history of stroke (50.3% vs. 35%, P = 0.027), and percentage of smoking (46.9% vs. 26.2%, P = 0.003).In addition, the CAN + group had lower HDL-C levels (median 1.0 vs. 1.1 mmol/L, P = 0.018), higher UACR (median 19.5 vs. 13.1 mg/g, P = 0.009), and significantly higher glycemic variability (SD) (median 2.3 vs. 1.9, P = 0.004).However, there was no statistically significant difference between the two groups in terms of gender distribution, age, BMI, DR, and CKD (P > 0.05).In addition to this, the significantly higher incidence of HU in patients in the CAN + group can be clearly demonstrated in Fig. 2 (54.5% vs. 26.2%, P < 0.001). It is worth noting that although the median duration of diabetes mellitus was 10 years in the CAN + group, which was slightly higher than that of 8 years in the CAN- group, the difference did not reach statistical significance (P = 0.079), which differed from the results of some previous studies.

Table 1.

Characteristics of T2DM patients with or without cardiovascular autonomic neuropathy (CAN)

	CAN-	CAN+	P
N	80	143
Gender, n(%)			0.246
Male	44 (55)	90 (62.9)
Female	36 (45)	53 (37.1)
Age (years)	56.4 ± 12.5	58.8 ± 10.7	0.134
Diabetes duration(years)	8.0 (3.0, 14.0)	10.0 (5.0, 14.0)	0.079
BMI (kg/m2)	24.7 (23.0, 28.4)	24.9 (23.3, 26.2)	0.377
DKD, n(%)			0.389
No	72 (90)	123 (86)
Yes	8 (10)	20 (14)
DR, n(%)			0.55
No	69 (86.2)	119 (83.2)
Yes	11 (13.8)	24 (16.8)
DPN, n(%)			0.011
No	17 (21.2)	13 (9.1)
Yes	63 (78.8)	130 (90.9)
Stroke, n(%)			0.027
No	52 (65)	71 (49.7)
Yes	28 (35)	72 (50.3)
Hypertension, n(%)			0.071
No	39 (48.8)	52 (36.4)
Yes	41 (51.2)	91 (63.6)
Smoking, n(%)			0.003
No	59 (73.8)	76 (53.1)
Yes	21 (26.2)	67 (46.9)
Drinking, n(%)			0.186
No	62 (77.5)	99 (69.2)
Yes	18 (22.5)	44 (30.8)
OHA, n(%)			0.357
No	6 (7.5)	6 (4.2)
Yes	74 (92.5)	137 (95.8)
Insulin, n(%)			0.12
No	28 (35)	36 (25.2)
Yes	52 (65)	107 (74.8)
FCP(pmol/L)	1.7 (1.2, 2.5)	1.8 (1.3, 2.6)	0.237
HbA1c, n(%)	8.1 (7.0, 9.7)	8.3 (7.1, 9.7)	0.58
TG(mmol/L)	1.4 (1.0, 2.4)	1.6 (1.1, 2.5)	0.43
TCH(mmol/L)	4.2 (3.4, 5.0)	4.0 (3.1, 4.7)	0.234
HDL-C(mmol/L)	1.1 (1.0, 1.4)	1.0 (0.9, 1.3)	0.018
LDL-C(mmol/L)	2.5 (1.8, 3.3)	2.4 (1.8, 3.0)	0.5
FBG(mmol/L)	7.4 (6.0, 9.9)	8.1 (6.3, 10.8)	0.316
UACR(mg/g)	13.1 (6.4, 25.3)	19.5 (9.1, 102.6)	0.009
Vascular plaque, n(%)			0.089
No	39 (48.8)	53 (37.1)
Yes	41 (51.2)	90 (62.9)
SD	1.9 (1.6, 2.7)	2.3 (1.9, 2.7)	0.004
MG	8.6 (7.8, 9.4)	8.4 (7.5, 9.7)	0.598
CV	23.8 (19.2, 28.6)	23.3 (19.6, 27.1)	0.64

Note: Data shown are mean (standard deviation), median (interquartile range), or number (percentage) unless otherwise indicated

Abbreviations: CAN+: new-onset CAN, CAN -: no new-onset CAN, BMI: body mass index, DKD: diabetic kidney disease, DR: diabetic retinopathy, DPN: diabetic peripheral neuropathy, OHA: oral hypoglycemic agents, FCP: fasting C-peptide, HbA1c: glycated hemoglobin, TG: total triglycerides, TC: total cholesterol, HDL-C: high-density lipoprotein cholesterol, LDL-C: low high-density lipoprotein cholesterol, FBG: fasting glucose, UACR: urinary albumin excretion rate, SD: standard deviation, MG: mean glucose value, CV: coefficient of variation

Fig. 2.

Cardinality test for the presence or absence of HU versus new-onset CAN

Survival curve analysis

To assess the impact of HU, we employed Kaplan-Meier survival analysis curves to evaluate the incidence of the primary outcome across groups (Fig. 3). The HU group exhibited a significantly higher probability of new-onset CAN in the later stages, with a statistically significant difference between groups (log-rank P = 0.039). The graph reveals that the probability of new-onset CAN in the early stages did not differ significantly between the non-hypoglycemia group and the HU group. To further explore the relationship between HU and new-onset CAN, we conducted landmark analysis (Fig. 4), using the median follow-up time as the cutoff point. After 27 months, the HU group demonstrated a markedly higher probability of new-onset CAN than the non-hypoglycemia group, with statistical significance (P = 0.005). This suggests that HU may chronically impair cardiac autonomic function over time, thereby promoting the development of CAN.

Fig. 3.

Survival curve (hypoglycemic of 0 is no hypoglycemia group; 1 is HU group)

Fig. 4.

Further landmark analysis for survival curves (hypoglycemic of 0 is no hypoglycemia; 1 is HU)

Relationship between HU and CAN

During the median follow-up period of 27 months, 143 (64.1%) patients developed new-onset CAN. As shown in Table 2, in the unadjusted model, patients with HU exhibited a significantly higher risk of CAN (HR = 1.44, 95% CI: 1.03–2.00, P = 0.001). In Model 1, adjusted for age and gender, the hazard ratio remained stable (HR = 1.42, 95% CI: 1.02–1.99, P = 0.002). Further adjustments in Model 2, which included BMI, duration of diabetes, DKD, DR, DPN, stroke, hypertension, smoking history, oral hypoglycemic agent use, insulin use, FCP, HbA1c, LDL-C, and CV, continued to demonstrate a significant association between HU and new-onset CAN risk (HR = 1.75, 95% CI: 1.23–2.49, P = 0.002). These results indicate that HU is an independent risk factor for CAN in T2DM patients, and this association remains robust after multivariate adjustments.

Table 2.

Risk ratio of new CAN in T2DM patients with HU

		HR	95%CI	P
HU	Unadjusted model	1.44	1.03-2	0.031
	Model1	1.42	1.02–1.99	0.039
	Model2	1.75	1.23–2.49	0.002

Model 1; adjusted for age, gender

Model 2: In addition to adjusting for model 1 variables, additionally adjusted for BMI, duration of diabetes, DKD, DR, DPN, stroke, hypertension, Smoking, history of oral hypoglycemic agents, history of insulin use, FCP, HBA1c, LDL-c, CV

Subgroup analyses

Subgroup analyses based on clinical characteristics such as age, gender, history of stroke, hypertension, smoking, and insulin use (excluding the investigated variables, with confounding factors consistent with the fully adjusted COX regression model) revealed that the association between HU and new-onset CAN in T2DM patients was consistent with that observed in the overall population, without significant heterogeneity (Fig. 5). Notably, in subgroups without a history of stroke, non-smokers, and individuals aged < 60 years, HU remained an important risk factor for new-onset CAN in T2DM patients.

Fig. 5.

Risk ratio of CAN in T2DM patients with HU in different subgroups

Sensitivity analysis

After excluding participants taking β-blockers that may affect HRV, 209 patients were included. After adjusting for the same confounders, we found that HU remained an independent risk factor for the development of CAN in patients with T2DM (HR = 1.52, 95% CI: 1.04-2.21, P = 0.03) (as shown in Table 3).

Table 3.

COX regression for new-onset CAN in T2DM patients with HU and without beta-blockers

Group	n. total	n(%)	HR (95%CI)	P
No hypoglycemia	117	61 (52.1)	1(Ref)
HU	92	72 (78.3)	1.52 (1.04–2.21)	0.03

Discussion

Previous studies on factors associated with CAN have tended to focus on laboratory markers [16–18]and little research examined the relationship between hypoglycemic conditions and CAN. One study showed [19] that CAN was an independent prognostic factor for the development of severe hypoglycemia in patients with T2DM, which seems to indicate a link between hypoglycemic conditions and CAN. This study aimed to investigate the relationship between HU and CAN. The results of Cox regression analyses showed that HU was a risk factor for the development of CAN in patients with T2DM, and the results still held after adjusting for many confounders. Meanwhile, we obtained similar results after performing multiple subgroup analyses. This result suggests that patients with HU should be screened for CAN as early as possible.

Hypoglycemia, a common complication of diabetes mellitus, has been a challenge in blood glucose regulation. Typically, hypoglycemia triggers sympathetic excitation with palpitations and sweating. Repeated hypoglycemia can lead to “HAAF,” which is manifested by a weakened sympathetic response and compensatory parasympathetic inhibition, which in turn reduces blood glucose perception. Thus, the patient develops HU [20]. From a practical point of view, HU and CAN often coexist. Previous researchers have made studies based on the association between the two, showing that the causal relationship between the two is unclear [21]– [22].

Some scholars have offered opinions different from those of this study. A study from Qatar [23] analysed indicators of glycemic variability in 40 diabetic patients and concluded that CAN was associated with increased glycemic variability and that shorter normoglycemic times were attributable to longer periods of hyperglycemia. Another study conducted by academics from Italy [24] in a nondiabetic population with a load of 1-hour hyperglycemia found similar conclusions. Hyperglycemia [25] causes oxidative stress, and increased oxidative stress is an important factor in the development of autonomic neuropathy. The differences between such studies and the present study are not entirely clear. Several plausible explanations can be proposed. First, such studies have mostly explored the correlation between the two in a cross-sectional state and have included a small number of cases. Second, intensive glycemic control reduces the incidence of CAN, as shown in the Diabetes Control and Complications Trial and the Epidemiology of Diabetes Interventions and Complications study [26]– [27]. In other words, neither high nor low blood glucose is conducive to reducing the likelihood of CAN in patients.

The results of the present study showed that HU was an independent risk factor for the development of CAN in patients with T2DM. Further mechanistic studies suggest that HU affects autonomic function through multiple pathways, thereby promoting the development of CAN. Specifically, nocturnal hypoglycemia can lead to elevated electrocardiographic indices such as the LF/HF ratio, suggesting increased sympathetic activity [28]; also, HU causes changes in heart rate variability (HRV), such as decreased SDNN and RMSSD, suggesting a diminished vagal response [29]. In addition, a study by Mamta Jaiswal et al. further revealed that hypoglycemic stress-induced glycemic variability is associated with reduced HRV, and this association is independent of the level of glycemic control as assessed by HbA1c [30]. Together, these studies suggest that HU plays a vital role in the onset and progression of CAN by affecting the sympathetic and vagal balance. In patients with type 2 diabetes mellitus (T2DM), hypoglycemia exacerbates the progression of CAN through multiple pathophysiological mechanisms. Firstly, hypoglycemia directly activates the sympathetic-adrenal system (SAS), and repeated stimulation leads to blunted sympathetic responses, known as HAAF [20]. This manifests as β-adrenergic receptor desensitization and adaptive changes in the central nervous system. Concurrently, hypoglycemia induces a burst of reactive oxygen species (ROS), triggering oxidative stress that damages mitochondrial and lipid membranes of autonomic nerve fibers, resulting in Schwann cell demyelination and axonal degeneration [31]. Additionally, hypoglycemia inhibits endothelial nitric oxide synthase (eNOS) activity, reducing nitric oxide production and causing microvascular constriction and endothelial cell apoptosis, further aggravating ischemic injury to autonomic nerves [32]. Hypoglycemia synergizes with GV ^{33 34}. Rapid glucose fluctuations disrupt neuronal energy metabolism by decoupling glycolysis from oxidative phosphorylation. Meanwhile, advanced glycation end products (AGEs) accumulated during hyperglycemic periods activate RAGE receptors during hypoglycemia, exacerbating inflammatory responses (e.g., NF-κB and NLRP3 inflammasome pathways) and creating a “metabolic memory effect.” Hypoglycemia also inhibits γ-aminobutyric acid (GABA) synthesis and promotes glutamate release, leading to excitotoxicity and impairing signal integration in autonomic ganglia. At the cardiac level, hypoglycemia induces electrolyte imbalances (e.g., hypokalemia) and autonomic dysregulation, prolonging the QT interval and increasing the risk of ventricular arrhythmias. It also forces the myocardium to rely on inefficient fatty acid oxidation for energy, exacerbating myocardial ischemia and autonomic dysfunction. The potential role of hypoglycemia in impairing protective cardiac neural mechanisms may directly affect cardiac electrical activity [35–37]. Furthermore, hypoglycemia reduces cardiac vagal outflow and diminishes vagal control in diabetic patients^{38 39}. The interplay between the immune and neuroendocrine systems further complicates the pathological process. Hypoglycemia suppresses the function of type 2 innate lymphoid cells (ILC2), impairing glucagon release, and alters gut microbiota to reduce short-chain fatty acids (SCFAs), thereby lowering vagal tone [34]. In summary, hypoglycemia drives CAN progression through oxidative stress, inflammation, metabolic toxicity, and immune modulation. Attentive monitoring and targeted interventions are essential to break this vicious cycle and improve patient prognosis.

Given the clinical importance and applicability of this study, this study reminds clinicians to do timely screening for CAN when HU is found in patients with type 2 diabetes. This is similar to clinical practice guidelines [15] that recommend early screening for CAN [15]. Most of the known methods of screening for CAN in patients with T2DM [40] are cumbersome and their clinical dissemination is hampered. Our study provides evidence for early screening and intervention for CAN when HU is detected by clinicians and for HU itself. It is associated with impaired glucose counter-regulation, notably reduced adrenal responses, leading to severe hypoglycemia [41–43]. HU and recurrent hypoglycemia interact with each other, thus forming a vicious cycle. Therefore, HU is not only harmful in its own right but also suggests that patients may subsequently develop new-onset CAN. The HypoDE study demonstrated that hypoglycemia awareness scores in the whole cohort also improved by 40% with the use of CGM [44]which may suggest that the combination of CGM and assessment of autonomic function by clinical practitioners, early identification of high-risk patients, and optimization of glucose-lowering strategies may help to reduce the number of CAN-associated cardiovascular events. In addition, regarding the choice of glucose-lowering regimen, glucose-lowering agents with significant differences in hypoglycemic risk may influence the progression of CAN. For those at high risk of CAN (e.g., comorbid DPN or history of stroke), preference should be given to agents with lower risk of hypoglycemia (e.g., SGLT2 inhibitors, GLP-1RA) and avoiding excessively stringent glycemic control (e.g., HbA1c < 7.0%) to reduce hypoglycemic events.

Our study possesses several strengths. Firstly, CGMS enables precise glucose monitoring, addressing the limitations of traditional methods in detecting HU. Secondly, we employed multiple subgroup population analyses and sensitivity analyses to validate the relationship between HU and new-onset CAN, while adjusting for multiple confounding variables to obtain similar results.

However, it is equally important to acknowledge the limitations of this study. Firstly, the fact that this study is not a randomised study and has a relatively small sample size may affect the accuracy and generalizability of the findings. Secondly, the study spanned only three years, potentially missing long-term effects. Future studies could expand the sample size to further validate the results. Thirdly, a questionnaire assessing the degree of impaired hypoglycemic awareness was not used for the diagnosis of HU, which may have had some impact on the results of our study. Fourthly, the regression analysis used in this study was a standard COX regression, and changes in variables during the follow-up period may have affected the association between HU and CAN. Fifthly, despite adjusting for various metabolic and complication indicators, inflammatory markers (e.g., IL-6, CRP) or oxidative stress indicators (e.g., MDA) were not included, possibly omitting insights into specific pathological mechanisms. Sixthly, this study did not include information on patients’ statin use, ACEI use, sulfonylureas, GLP1, and SGLT2i, which could be confounding factors affecting the study and need to be explored in further studies. Seventhly, this single-center study may be limited by regional and population characteristics, potentially affecting the extrapolation of the results. Multicenter studies could include a broader range of participants, reducing the impact of regional and population differences and enhancing the reliability of the findings. Lastly, as a clinical study, this research does not establish a causal relationship between HU and CAN in T2DM patients. Future studies should employ multicenter designs and molecular biology approaches to further elucidate the underlying pathological mechanisms and identify potential intervention targets.

In summary, HU is an independent risk factor for CAN, and this association remains robust after multivariate adjustments. While previous studies have primarily focused on the acute effects of hypoglycemia, this study, through long-term follow-up, highlights the chronic hazards of HU, filling a critical gap in this field. Clinicians should remain vigilant for the development of CAN in patients experiencing HU, ensuring early and effective comprehensive treatment to reduce CAN-related mortality.

Conclusion

In conclusion, a series of analyses in our study suggest that HU is an independent risk factor for CAN in T2DM patients, and we recommend early screening for CAN in T2DM patients with HU, including heart rate variability testing and postural hypotension assessment. Meanwhile, we found that HU may affect new-onset CAN in patients through a mechanism that produces chronic damage in the body and superimposes it, and this potential mechanism needs to be further explored. This study provides new ideas and evidence-based clues for the comprehensive management of T2DM patients.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Abbreviations

T2DM

Type 2 diabetes mellitus

CAN

Cardiac autonomic neuropathy

HU

Hypoglycemia unawareness

CGMS

Continuous glucose monitoring system

HAAF

Hypoglycemia-associated autonomic failure

HbA1c

Glycated hemoglobin

CV

Coefficient of variation

MG

Mean glucose value

SD

Standard deviation

BMI

Body mass index

FBG

Fasting blood glucose

FCP

Fasting c-peptide

HbA1c

Glycated hemoglobin

TC

Total cholesterol

TG

Total triglycerides

LDL-C

Low high-density lipoprotein cholesterol

HDL-C

High-density lipoprotein cholesterol

UACR

Urinary albumin excretion rate

CAN +

New-onset CAN

CAN -

No new-onset CAN

DKD

Diabetic kidney disease

DR

Diabetic retinopathy

DPN

Diabetic peripheral neuropathy

OHA

Oral hypoglycemic agents

LF/HF ratio

Low-Frequency to High-Frequency Ratio

​ SDNN

Standard Deviation of All Normal to Normal RR Intervals

RMSSD

Root Mean Square of Successive Differences​

Author contributions

SQL, YHC and WD designed the study. SQL and XXS collected clinical data. SQL and XW performed the statistical analysis of the data. SQL authored the first draft. YHC examined and revised the paper. All authors read and approved the final manuscript.

Funding

Supported by Hefei Municipal Health Commission Applied Medical Research Project (Hefei Health Science and Education [2019] No. 172, Hwk2022zc048, Hwk2023zd003); Bengbu Medical College Scientific Research Project (2022byzd193); and Anhui Provincial Health Research Project (No. AHWJ2024Aa20141).

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

The Ethics Committee of Hefei Hospital of Anhui Medical University approved the study protocol by the principles of the Declaration of Helsinki, and each participant signed an informed consent form before inclusion.

Consent for publication

All authors agree to publish this work.

Competing interests

The authors declare no competing interests.