Background
Cardiovascular autonomic neuropathy (CAN) is a common and clinically significant diabetic complication that causes cardiovascular disease (CVD), mortality, and morbidity in diabetic patients [
1]. Both chronic hyperglycemia and glucose fluctuation are important causes of the initiation of the pathogenic process of CAN, as they result in increased oxidative stress, which induces direct neural damage and indirect neuronal ischemia from endothelial dysfunction [
2,
3].
Although glycated hemoglobin (HbA1c) is the gold-standard parameter for glycemic control, emerging evidence has shown that glycated albumin (GA) is a useful indicator of glycemic control in subjects with hematologic disorders such as anemia, chronic kidney disease, or hemorrhage, which can affect HbA1c level [
4]. Serum GA provides a more accurate assessment of recent glycemic change due to the short half-life (2-3 weeks) of albumin [
5], and it is more strongly associated with glucose excursions, whereas HbA1c reflects average glucose level rather than glycemic variability (GV) [
6,
7]. GA actually correlates with diabetic complications such as retinopathy progression [
8], chronic kidney disease [
9], peripheral neuropathy [
10], and even CVD [
11,
12].
Most of the studies about GA and diabetic complications used GA measured at one time point, which does not reflect the long-term glycemic state. Although HbA1c variability as a long-term GV parameter has been associated with diabetic complications independent of mean HbA1c [
13‐
15], there have been few studies of GA variability and diabetic complications. This study aims to investigate CAN and its relationship to both mean GA level and GA fluctuation in type 2 diabetic patients compared to mean HbA1c and HbA1c fluctuation.
Methods
Study subjects
A total of 1019 adult type 2 diabetes mellitus (T2DM) patients (age ≥ 18 years) who underwent CAN testing in the outpatient diabetes clinic of Samsung Medical Center in Seoul, Republic of Korea, at least twice at 2-year intervals were screened from September 2011 to March 2017. T2DM was diagnosed based on the 2017 American Diabetes Association guidelines. The exclusion criteria were as follows (Additional file
1: Figure S1): clinical diagnosis of type 1 diabetes (
n = 111); history of CVD including any form of coronary artery disease, congestive heart failure, arrhythmia, valvular disease, or ischemic stroke (
n = 81); history of liver disease or abnormal liver function test at baseline (
n = 38); any malignancy upon treatment (
n = 56); pancreatitis or pancreatectomy (
n = 13); any transplantation (
n = 10); hyper- or hypothyroidism (
n = 17); history of chronic kidney disease or estimated glomerular filtration rate (eGFR) calculated by the CPD-EPI equation ≤ 60 ml/min/1.73 m
2 (
n = 54); CAN diagnosis at the time of enrollment (
n = 393); follow-up of less than 2 years (
n = 153); missing serial measurements of GA or HbA1c for 2 years (
n = 82); other missing clinical variables (
n = 11).
Finally, 498 subjects (299 men and 199 women) who were never diagnosed with CAN were enrolled in the study. The study protocol was approved by the Institutional Review Board of Samsung Medical Center.
Demographic and clinical assessment
Patient medical history, medication, smoking, and alcohol drinking status data were extracted from electronic medical records for the same day that the first CAN test was administered. HbA1c level was measured by high-performance liquid chromatography, using a VARIANT II TURBO analyzer (Bio-Rad Laboratories, Hercules, CA, USA). Serum GA level was measured by an enzymatic method using a Lucica GA-L kit (Asahi Kasei Pharma Corporation, Tokyo, Japan). The reference interval for HbA1c was 4.0–6.0%, and that for GA was 11.0–16.0%. Serum GA (median number of measurements 7, interquartile range [IQR] 6–8) and HbA1c levels (median number of measurements 8, IQR 7–8) were consecutively measured every three to 6 months for 2 years. Mean values were calculated for GA and HbA1c levels measured during the 2 years after recruitment to the day of the CAN test. The variabilities of GA and HbA1c were evaluated using the intrapersonal standard deviation (SD) and coefficient of variance (CV) of the serial measurements of GA and HbA1c. Because few visits result in apparently larger SD than many visits, the adjusted SD was defined as the SD divided by [n/(n − 1)]
0.5, where n is the number of GA or HbA1c measurements [
16], and was used in the analyses instead of the unadjusted SD.
After an 8-h overnight fast, plasma glucose level was measured using the glucose oxidase method, and serum C-peptide level was measured in duplicate with immunoradiometric assays (Beckman Coulter, Fullerton, CA, USA). Lipid profiles were assayed using a Hitachi 7600 auto analyzer (Hitachi Instruments Service, Tokyo, Japan).
Assessment of CAN
Cardiovascular tests based on heart rate response and orthostatic hypotension are an essential and irreplaceable component of CAN diagnosis [
17‐
19]. Patients were advised to avoid strenuous physical exercise, tobacco, and alcohol in the 24-h preceding the test and to avoid coffee and food for at least 3-h prior to the test. Medications such as anti-histamines, anti-depressants, and β-blockers were withheld for 12-h prior to the test.
Three tests mainly assess parasympathetic function: heart rate variability (HRV) to deep breathing (exhalation: inhalation ratio), to standing (30:15 ratio), and to the Valsalva maneuver (Valsalva ratio); blood pressure response to standing is used to assess sympathetic function. The heart rates responses were assessed automatically from electrocardiography recordings using the DICAN evaluation system (Medicore Co., Ltd., Seoul, Korea). Change in systolic blood pressure was obtained from the difference between the value measured in the supine position and the value measured 60 s after standing.
Values below the lower limit of an age-specific reference range suggested in our previous study were considered abnormal [
13]. CAN was defined as the presence of two or more abnormal results among the three parasympathetic tests (definite CAN) or presence of orthostatic hypotension (severe CAN) [
20].
Statistical analysis
All statistical analyses were performed using SPSS version 24.0 software (Chicago, IL, USA) and STATA/SE version 14.0 (StataCorp LP, College Station, TX, USA). Data are expressed as mean ± SD, median with IQR, and number (%). Student’s t test and the Mann–Whitney U test were used to assess the differences in continuous variables between two groups, while the Chi square test was used for categorical variables. Scatter plots and linear analyses were used to evaluate the correlations between mean GA and mean HbA1c as well as their mean values and GV parameters. Binary logistic regression analyses assessed the associations between CAN and various glycemic parameters. Multivariate logistic regression analysis was adjusted for covariates that were significant in univariate analysis and those that were conventional risk factor of CAN with the enter method, along with each glycemic parameter as an independent variable. A variance inflation factor > 5.0 was used as an indicator of multicollinearity. The area under the curve (AUC) was obtained from receiver operating characteristic (ROC) curve analyses to compare the predictive values between GA- and HbA1c-related parameters. ROC analyses were also adjusted using the enter method. The level of significance was set at P < 0.05 in two-tailed tests.
Discussion
This study demonstrated for the first time that 2-year consecutively measured GA level as a mean value and variability was significantly associated with CAN development, regardless of other conventional CVD risk factors. As in our previous study [
13], this study consistently showed that mean glucose level and GV parameters were associated with CAN. However, we further evaluated a novel idea about the clinical utility of GA to predict CAN development compared to HbA1c; GA variability (SD and %CV) could more accurately predict CAN development than HbA1c variability, although the predictive value of mean GA did not differ from that of mean HbA1c. Each of mean GA, adjusted SD of GA, and %CV of GA level had distinct predictive value for CAN development and concurrently showed an additive effect for the prediction of CAN with mean HbA1c level. Furthermore, the predictive value of mean GA with GA variability was superior to that of mean HbA1c with HbA1c variability.
CAN is a common chronic diabetic complication that confers high mortality and morbidity in diabetic patients. Hypoglycemia unawareness promotes a reduced threshold for uncontrolled arrhythmia and sudden cardiac death [
25]. Insensitivity to ischemic pain impairs the early recognition of myocardial infarction [
26]. Imbalance between sympathetic and parasympathetic functions can deteriorate the autoregulation of cerebral blood flow and produce susceptibility to stroke [
27]. CAN is closely associated with other diabetic microvascular complications such as retinopathy, nephropathy, and peripheral neuropathy, through changes in the vasomotor control of small vessels [
28,
29]. Therefore, early detection and screening of patients with high risk for CAN development are clinically important to prevent potential progression to overt CVD and to identify concomitant diabetic complications.
The specific pathogenesis of CAN remains unclear. However, neuronal injury due to oxidative stress from poor glycemic control is the leading cause of the initiation of CAN development [
18]. Sustained hyperglycemia induces the over-production of reactive oxygen species in the mitochondria, mitochondrial antioxidant enzymes such as superoxide dismutase 2 are depleted, and neuronal injury results [
30]. Wide fluctuations in blood glucose level can trigger the same levels of oxidative stress as prolonged hyperglycemia. Postprandial hyperglycemia increased oxidative stress markers [
31], and it was significantly higher in both the fasting state and the postprandial state in diabetic patients, whereas there was no postprandial elevation of oxidative stress markers in healthy subjects [
32]. In our study, mean glucose level measures (GA and HbA1c) and their variability did not have a significant difference for predicting CAN development (Additional file
3: Table S2). Even the CVs of GA or HbA1c level, which adjusted the effect of mean glucose exposure, remained a significant risk factor for CAN. Therefore, the finding that the predictive power of GA variability for CAN development outperformed that of HbA1c variability was partially explained by GA’s shorter half-life, which enables it to reflect glucose excursions more accurately than does HbA1c [
6,
33,
34]. The 2-year fluctuations in GA level were actually greater than those in HbA1c level in this study.
Together with being a marker of glucose control, GA might be a direct cause of diabetic complications including atherosclerosis [
35]. GA is regarded as a precursor of advanced glycation end products (AGEs), which are thought to contribute to glucose toxicity [
36]. AGEs result in increasing plasminogen activator inhibitor-1 and endothelin-1 production and impairment of endothelial nitric oxide (NO) synthase and NO actions, which can lead to reduction of neurovascular perfusion and cellular apoptosis [
37]. GA accelerates the development and progression of atherosclerosis and vascular complications of diabetes via the proliferation and migration of vascular smooth muscle cells [
38].
Although there has been no research evaluating the association between GA and CAN, growing evidence supports the crucial relationship between GA and subclinical or overt CVD; higher serum GA, rather than HbA1c, was significantly associated with more severe coronary artery stenosis [
11,
39,
40]. Higher GA increased the risk for the presence of carotid plaque [
12] and was positively related with carotid intima-media thickness [
41,
42]. GA was associated with long-term cardiovascular outcomes including myocardial infarction, ischemic stroke, heart failure, and even death, independent of traditional CVD risk factors [
43].
This study has several limitations. First, it has a retrospective design; thus, we cannot exclude the possibility of selection bias. Second, study subjects were recruited from a tertiary hospital and thus are not representative of the general Korean population. Third, concurrent hypoglycemia events were not assessed via continuous glucose monitoring, although hypoglycemia itself changes heart rate and blood pressure. However, there were no subjects who experienced a hypoglycemia event (FPG ≤ 70 mg/dL) during 2 years of observation, based on their consecutively measured FPG level in the outpatient clinic.
In conclusion, higher serum GA is significantly associated with higher risk of developing CAN. Mean GA and GA variability showed additive effects with mean HbA1c to improve the predictions for CAN, while each of them had a distinct predictive power to detect CAN. Although HbA1c has been accepted as a gold standard marker for monitoring glycemic status and diabetic complications, GA and its variability might be useful indicators for diabetic complications, and they also enhanced HbA1c’s modest clinical prediction of CAN development.
Authors’ contributions
JJE collected data, performed statistical analysis, and wrote the initial draft of the manuscript. SEL, YBL, and JYA assisted with data collection and data management. GK, SMJ, KYH, and MKL assisted with the design of the study and reviewed the manuscript. JHK takes responsibility for the integrity of the data and the accuracy of the data analysis and reviewed the manuscript. All authors read and approved final manuscript.