2.1. Participants
Participants were recruited from 2013 to 2024 and comprised a healthy control group (HC group) and patients with MDD (MDD group).
In the HC group, participants were recruited from the Health Management Center of Kaohsiung Medical University Hospital, Kaohsiung Medical University campus, the community of Kaohsiung City, and through online flyers. The inclusion criteria included as follows: (1) participants’ health examination reports showed no severe physical disorder (e.g., cancer, kidney disease, stroke, etc.) or psychiatric disorders (e.g., depressive disorder, anxiety disorder, bipolar disorder, schizophrenia, etc.), and they were not taking prescribed medications; (2) total scores on the Beck Depression Inventory-II (BDI–II) and Beck Anxiety Inventory (BAI) were lower than 14 and 8, respectively; and (3) aged between 20 and 70 years old.
In the MDD group, participants were recruited for cross-validation by the Department of Psychiatry of three medical centers of Kaohsiung Medical University Hospital. The inclusion criteria included as follows: (1) diagnosed with MDD by psychiatrists based on the Diagnostic and Statistical Manual of Mental Disorders, 5th edition (DSM−5; American Psychiatry Association, 2013); (2) total scores on the BDI-II and BAI exceeding 14 and 8, respectively; (3) aged between 20 and 70 years; and (4) without severe physical illnesses or psychiatric disorders.
The HC group was divided into five age groups: 20–29, 30–39, 40–49, 50–49, and 60–69 years old, with each group comprising a minimum of 15 males and 15 females to establish a normative database. However, only 11 males were enrolled in the 61–70 age group of the HC cohort due to the challenge of finding participants without physical illnesses or mental disorders in this age range. A total of 356 healthy participants were recruited for the study, with 26 participants excluded (age < 20 years [n = 1], age > 70 years [n = 4], BDI-II or BAI higher than 14 or 8 [n = 14], presence of a physical illness [n = 7]). Of the remaining 330 healthy participants who completed the HRV measurement, 19 participants were removed from the statistical analysis (issues during ECG measurement [n = 14], arrhythmia, or ECG artifacts [n = 5]). Ultimately, 311 healthy participants were included in the HC group, comprising 137 males and 174 females.
In the MDD group, 422 patients were referred by psychiatrists, and 118 were excluded (age < 20 years [n = 7], age > 70 years [n = 3], BDI-II and BAI scores not exceeding 14 and 8 [n = 97], primary diagnosis not being MDD [n = 2], comorbid with severe physical illnesses [n = 8], or daytime sleepiness [n = 1]). A total of 304 participants in the MDD group completed the ECG measurement. Thirty-two participants had their HRV data removed from the statistical analysis (issues during ECG measurement [n = 12], missing HRV data or questionnaires [n = 14], arrhythmia [n = 6]). Ultimately, 272 participants (58 males and 163 females) were included in the MDD group.
This study received approval from the Institutional Review Board of Kaohsiung Medical University Hospital, Taiwan (KMUH IRB-20120209 II, KMUH IRB-FI20160027, and KMUH IRB-F(I) 20200117), Kaohsiung Chang Gung Memorial Hospital, Taiwan (CGMH IRB1604250002), and Kaohsiung Tsyr Huey Hospital, Taiwan (THMH-RES-21051401). Participants provided informed consent before the study and received TWD 500 (approximately USD 16) for their participation.
2.2. Material
Demographic data (age and sex), medications used, and scores of the BDI-II and BAI were collected. The BDI-II comprises a 21-item Likert scale designed to assess depressive symptoms, with a total score ranging from 0 to 63 and further divided into cognitive and somatic depression subscales. BDI-II scores below 13 indicate a normal range of depression, 14–19 indicate mild depression, 20–28 indicated moderate depression, and scores above 29 indicate severe depression [
21]. Its Chinese version, translated by Chen [
22], demonstrated good psychometric properties, including a Cronbach’s α of 0.94, split-half reliability of 0.91, and a correlation of 0.69 with the Chinese Health Questionnaire of 0.69 [
23]. Similarly, the BAI consists of a 21-item Likert scale used to assess anxiety symptoms with a total score ranging from 0 to 63 and divided into cognitive and somatic anxiety subscales. BAI scores below 7 indicate a normal range of anxiety, 8–15 indicate mild anxiety, 16–25 indicate moderate anxiety, and scores above 26 indicate severe anxiety [
24]. Its Chinese version, translated by Lin [
25], exhibited good psychometric qualities, with a Cronbach’s α of 0.95, a split-half reliability of 0.91, and a correlation of 0.72 with the Hamilton Anxiety Scale of 0.72 [
26].
ECG recording: An ECG sensor with a sampling rate of 2048 Hz (Thought Technology Ltd., Montreal, QC, Canada) was positioned 1cm below the left and right sides of the clavicle, as well as on the skin surface of the fifth rib of the left chest. Lead II ECG was recorded, and ECG raw signals (PQRST wave) were acquired using ProComp Infiniti™ version 6.1.1 (Thought Technology Ltd., Montreal, QC, Canada).
Participants were instructed to refrain from consuming alcohol, coffee, and tea beverages for three hours before the ECG measurement, and recording sessions were scheduled between 9:00 am and 5:00 pm [
12,
27]. All participants were instructed to sit and rest in a temperature-controlled room (between 24 and 28 °C). The indoor lighting was set to a fixed brightness using 28 W, 95 lm, 6500 K LED white light (TOA, FH28D-EX), and blackout curtains were installed to prevent variations in indoor brightness due to weather changes. ECG signals of all participants were recorded for five minutes in a resting state with their eyes closed in the same laboratory.
2.3. Data Reduction, ECG Processing, and Statistical Analysis
The ECG data were analyzed using the CardioPro HRV Analysis Module (Thought Technology Ltd., Montreal, QC, Canada). Researchers checked the ECG waveform (R-spike) and removed arrhythmia and movement artifacts. Subsequently, interbeat intervals were converted to time and frequency domains of HRV, including the standard deviation of normal-to-normal intervals (SDNN, indicating total HRV), root mean square of the successive normal-to-normal interval differences (RMSSD, indicating vagal activation), low-frequency power (LF; 0.04–0.15 Hz, representing sympathetic and parasympathetic nervous systems coregulation or baroreceptor gain), high-frequency power (HF; 0.15–0.40 Hz, representing parasympathetic nervous system activity), total power (TP; 0.0033–0.4 Hz, representing total HRV), and LF/HF ratio (representing the sympathetic nervous system activity). Due to skewness and kurtosis of the HRV distributions, the LF, HF, LF/HF ratio, and TP were transformed using natural logarithms into lnLF, lnHF, lnLF/HF ratio, and lnTP [
8,
28].
Statistical analyses were performed using SPSS Statistics 21.0 (IBM Corporation, Armonk, NY, USA). Firstly, this study calculated the mean and standard deviation (SD) of the HRV indices in five age groups for 316 participants in the HC group, and the total Z-scores was 9480 (316 × 6 HRV indices × 5 discrete aged group). Secondly, the Z-scores and absolute value Z-scores (|Z score|) for the MDD group were calculated using the equation (z = (x − μ)/σ) for each discrete HRV indices, and the total number of Z-scores was 8070 (269 × 6 HRV indices × 5 discrete aged group). The Z-scores and |Z score| of the MDD group served as a benchmark for cross-validation. Thirdly, the Student’s t-test and chi-square (χ2) were used to examine group differences in demographic characteristics. Fourthly, analysis of variance (ANOVA) and analysis of covariance (ANCOVA) for controlling the covariables were used to examine the group differences in HRV indices. Fifthly, Pearson correlations were analyzed for Z-scores, depression, and anxiety in the MDD group. Finally, the R software version 4.2.0 (Free Software Foundation, Boston, MA, USA) with the ggplot2 package was utilized to visualize the HRV indices using the locally weighted scatterplot smoothing method and scatterplots.