675 T2DM patients who visited the Renmin Hospital of Wuhan University for evaluation or treatment from January 2019 to June 2021 with normocalcemia (serum calcium in the reference range of 8.46–10.10 mg/dl) were recruited. The exclusion criteria for patients were (1) dilated cardiomyopathy or hypertrophic cardiomyopathy; (2) pacemaker implantation; (3) degree II and above atrioventricular block, atrial fibrillation, sick sinus syndrome, and frequent premature contraction; (4) glomerular filtration rate <60 ml/min or urine albumin per gram urine creatinine (Alb/Cr) >300 mg/g; (5) alanine aminotransferase >120 U/l; (6) parathyroid disease or vitamin D-related disorders; (7) medication history including bisphosphonate, vitamin D, and diuretics which may influence calcium within the past 1 month; (8) serum calcium out of reference range (8.46–10.10 mg/dl); and (9) malignant tumor.

The patients’ smoking history, drinking history, and current and past medical history were collected. T2DM was diagnosed according to the World Health Organization criteria (19). This study was approved by the ethical review board of Renmin Hospital of Wuhan University and complied with the Helsinki declaration.

A 12-h overnight fasting venous blood sample was collected in all subjects. The serum electrolytes (calcium, sodium, potassium, chlorine, phosphate, and magnesium), uric acid, creatinine, albumin (ALB), total cholesterol (TC), triglycerides (TG), low-density lipoprotein cholesterol (LDL-C), high-density lipoprotein cholesterol (HDL-C), high-sensitivity C-reactive protein (hs-CRP), NT-proBNP, and fasting plasma glucose (FPG) were measured by a biochemical autoanalyzer (Abbott C8000, Chicago, IL, USA). HbA1c and homocysteine (HCY) were measured by high-performance liquid chromatography (HPLC; Bio-Rad, Hercules, CA, USA).

Serum calcium CV_w (within-subject coefficient of variation) = 2.1 mmol/l; CV_b (between-subject coefficient of variation) =2.5 mmol/l; imprecision = 1.1%; bias = 0.8%; TEa (total allowable error, p < 0.05) = 2.5%. Albumin-adjusted serum calcium levels (CaA) (mg/dl) were derived from formula (serum total calcium concentration (mg/dl) + 0.8 × [4 − serum albumin concentration (g/dl)] (17) to adjust total calcium for hypoalbuminemia to permit approximation of the ionized calcium, which is physiologically active and under homeostatic control. The fasting triglyceride-glucose index (TyG), which was sensitive for recognizing insulin resistance, was calculated using the formula: Ln [TG (mg/dl) × FPG (mg/dl)/2] (20).

24-hour ambulatory electrocardiogram (ECG) was recorded using a wearable 12-lead digital Holter device (JincoMed, Beijing, China) with a data acquisition speed of 4,000 Hz. Subjects were told to follow their daily routines but to avoid intense physical activities or shower. ECG data were processed using a professional Holter analysis system developed by JincoMed, including algorithms for QRS labeling, arrhythmia detection, artifact identification, and data correction, followed by manual review by Holter technicians in Renmin Hospital of Wuhan University. Both time- and frequency-domain HRV parameters were derived from ambulatory ECG data. Time-domain HRV parameters quantify the variability of successive heartbeat interval, including standard deviation of normal R–R intervals (SDNN), mean standard deviation of normal R–R intervals for 5-min segments within 24 h (SDNN index), root mean square of successive RR interval differences (rMSSD), and the integral of the density of the RR interval differences (triangular index). Frequency-domain parameters estimate the distribution of power into different frequency bands, including high-frequency (HF, 0.15–0.40 Hz), low-frequency (LF, 0.04–0.15 Hz), and LF/HF ratio. 24-hour and hourly averages of time-domain HRV measures and hourly averages of frequency domain measures were obtained.

SDNN is widely used to evaluate autonomic function and is considered to be a sensitive indicator in T2DM patients (21). In this study, T2DM subjects were divided into two groups, declined HRV (SDNN <100 ms) and normal HRV (100 ms ≤ SDNN ≤180 ms), according to heart rate variability standards recommended by the European Society of Cardiology and the North American Society of Pacing and Electrophysiology Task Force (22).

Echocardiographic examinations were performed using commercially available ultrasound diagnostic instruments (GE Vingmed Ultrasound, Horten, Norway) in accordance with the guidelines issued by the American Society of Echocardiography (23). We measured cardiac structure indicators, which included left atrial diameter (LAD), aortic root dimension (AOD), left ventricular end diastolic dimension (LVDd), diastolic interventricular septum thickness (IVSd), main pulmonary artery diameter (MPAD), right ventricular end diastolic diameter (RVDd), right atrium diastolic transverse diameter (RADd), and diastolic left ventricular posterior wall thickness (LVPWd).

The Doppler spectrum of the mitral valve pulse was recorded in an apical four-chamber view. The peak velocity of the filling peak in the early diastolic period (E) and late diastolic period (A) and the early diastolic mitral annulus velocity (e′) were measured. Cardiac diastolic function was assessed by E/A ratio and E/e′ ratio. Three cardiac cycles were measured, and the average value was used. Diastolic dysfunction was defined as E/A ratio <1.0 or E/e′ ratio >15 (24). Left ventricular systolic function was assessed by left ventricular ejection fraction (LVEF).

Continuous variables were presented as mean ± standard deviation (SD), as well as frequencies and percentages for categorical variables. Continuous variables with non-normal distribution are represented by median (M) and quartile range (QR). Differences in normally distributed variables were determined by independent-sample T test or one-way ANOVA. Chi-square tests were applied for categorical variables. Bivariate linear correlation (Pearson correlation) analysis was carried out to evaluate the associations between CaA and HRV parameters.

Logistic regression analysis was performed using HRV as the dependent variable to analyze the association between CaA and HRV after adjusting for potential confounders. Odds ratios (OR) with 95% confidence intervals (CI) were calculated for the relative risk of lower serum calcium level with declined HRV. The ability to predict declined HRV of CaA was evaluated using the area under the curve (AUC) in the receiver operating characteristic (ROC) curve. All statistical analysis were performed using SPSS version 22.0. All tests were two-sided, and p <0.05 was considered statistically significant.

In this study, 675 T2DM subjects (399 men and 276 women) were included, with a mean age of 62.49 ± 11.36 years (from 23 to 95 years old). Declined HRV (SDNN <100 ms), hypertension, coronary heart disease, and carotid atherosclerosis were present in 250 (37.0%), 501 (74.2%), 370 (54.8%), and 132 (71.4%) patients, respectively. There were no individuals of SDNN >180 ms in the enrolled T2DM patients.

Significant differences in CaA (8.86 ± 0.27 vs. 9.13 ± 0.39 mg/dl, p <0.001) were observed between declined HRV and normal HRV groups ( Table 1 ). The patients with declined HRV were older and had higher levels of HbA1c, NT-proBNP, and hs-CRP, as well as lower levels of HDL-C and LDL-C than those with normal HRV (p < 0.05, respectively). There was no significant difference in prevalence of smoking or drinking history, hypertension, coronary heart disease, and carotid atherosclerosis between the two groups ( Table 1 ).

We analyzed echocardiographic indicators between the declined HRV and normal HRV groups. The data showed that echocardiographic cardiac structure indicator LAD (35.40 ± 4.56 mm vs. 34.47 ± 4.56 mm, p = 0.015) was significantly increased, and cardiac function index E/A (0.78 ± 0.22 vs. 0.83 ± 0.26, p = 0.029) was significantly decreased in the declined HRV group ( Table 2 ). However, LVEF, E/e′, and most cardiac structure indicators did not differ between groups.

Furthermore, bivariate linear correlation analysis showed that the E/A level was significantly and positively correlated with HRV parameters including SDNN (r = 0.148, p < 0.001), SDNN index (r = 0.164, p < 0.001), and Triangle index (r = 0.088, p = 0.030), but not with rMSSD and LF/HF ( Table 3 ).

To determine the variables associated with SDNN, logistic regression analysis was developed to include CaA, age, HbA1c, HDL-C, LDL-C, Ln(NT-proBNP), and hs-CRP. Declined HRV (SDNN <100 ms) was significantly associated with CaA [OR (95% CI), 0.056 (0.028–0.110), p < 0.001], age [OR (95% CI), 1.018 (1.001–1.035), p = 0.037], HbA1c [OR (95% CI), 1.253 (1.109–1.416), p < 0.001], Ln(NT-proBNP) [OR (95% CI), 1.230 (1.017–1.488), p = 0.033], and hs-CRP [OR (95% CI), 1.031 (1.013–1.049), p = 0.001] ( Table 4 ). Bivariate linear correlation analysis showed that the CaA level was significantly and positively correlated with HRV parameters including SDNN (r = 0.413, p < 0.001), SDNN index (r = 0.209, p < 0.001), rMSSD (r = 0.121, p = 0.002), and Triangle index (r = 0.218, p < 0.001), but not with the LF/HF ( Table 3 ).

To evaluate the predictive performance of CaA on HRV, the AUC in the ROC curve was calculated, which was 0.730 [95% CI (0.750–0.815), p < 0.001]. The cutoff value of CaA was 8.87 mg/dl (sensitivity 0.644, specificity 0.814) ( Figure 1 ). We divided these patients into two groups (CaA level <8.87 mg/dl and ≥8.87 mg/dl) based on the cutoff value. HRV parameters, including SDNN (p < 0.001), SDNN index (p < 0.001), rMSSD (p = 0.006), and Triangle index (p < 0.001), were significantly decreased in the group of lower calcium level ( Table 5 ). However, LF/HF did not differ between the two groups.

The multivariate logistic regression analysis shows the ORs (95% CI) for declined HRV according to the two groups of CaA levels. In contrast to subjects with CaA level ≥8.87 mg/dl, there was a significantly increased risk of declined HRV in subjects with CaA level <8.87 mg/dl [OR (95% CI), 0.049 (0.024–0.099), p < 0.001] after adjusting for possible confounding factors including creatinine, serum phosphate, age, gender, smoking, drinking, HbA1c, Ln(NT-proBNP), hs-CRP, TyG index, dyslipidemia, hypertension, coronary heart disease, carotid atherosclerosis, and the use of hypertension medication.

We analyzed echocardiographic indicators and cardiac function between the two groups (CaA level <8.87 and ≥8.87 mg/dl). LVDd (45.22 ± 4.04 mm vs. 44.44 ± 4.21 mm, p = 0.026) was significantly increased in the group of lower calcium level. However, cardiac function indexes (LVEF, E/e′, E/A, and NT-proBNP) and most cardiac structure indicators (AOD, LAD, IVSDd, MPAD, RADd, RVDd, and LVPWd) did not differ between the two groups ( Table 5 ). There were no significant differences in FBG, HbA1c, and TyG index between the two groups.