Abstract
Objective
To explore the correlation between sleep quality (SQ) during the perioperative period of patients undergoing cardiac surgery and heart rate variability (HRV).
Methods
This prospective observational cohort study included 95 patients undergoing cardiac surgery. SD was defined as a Pittsburgh Sleep Quality Index (PSQI) total score >7, and all participants were categorized into an SD group and a non-SD group. HRV parameters were collected and compared at four predefined time points: 1 day preoperatively (T0), 1 day postoperatively (T1), 2 days postoperatively (T2), and 3 days postoperatively (T3). Repeated-measures ANOVA analyzed dynamic changes in HRV parameters within the SD group. Lasso regression screened the main HRV parameters associated with SD, followed by logistic regression to identify significantly related to SD. Point-biserial correlation analysis was used to test the positive or negative correlation between HRV parameters and SD.
Results
No significant differences in clinical data were found between the two groups (P>0.05). Compared with the non-SD group, the SD group showed significantly lower HF, SDNN, SDANN, PNN50, and RMSSD at all time points (P<0.05), while LF, LF/HF, and SDNNI showed no differences (P>0.05). In the SD group, all HRV parameters exhibited a significant time effect (P<0.05); LF, HF, LF/HF, SDNN, SDNNI, PNN50, and RMSSD showed significant group effect (P<0.05); SDANN and RMSSD demonstrated significant time×group interaction (P<0.05). Lasso and logistic regression identified HF (T2), SDANN (T1), SDANN (/T2), PNN50 (T0/T1/T3), RMSSD (T0/T1/T2) as related to SD (OR<1, P<0.001). Point-biserial correlation revealed that all included HRV parameters were significantly negatively correlated with SD (correlation coefficient range: −0.40 to −0.78); PNN50 measured at T1 had the strongest negative correlation with SD (r= −0.78).
Conclusion
Perioperative SQ in cardiac surgery patients is closely related to HRV. HF, SDANN, PNN50, and RMSSD are HRV parameters significantly correlated with SD, suggesting that HRV monitoring may aid in early identification of autonomic dysfunction and guide individualized sleep interventions.
Keywords: cardiac surgery, SD, HRV, correlation
Introduction
Patients undergoing cardiac surgery often experience respiratory difficulties, loss of appetite, and sleep quality (SQ) during the perioperative period due to trauma, anesthesia, pain, and psychological stress.1,2 Unique cardiac-surgery-specific aggravating factors further disrupt sleep and alter autonomic function, including cardiopulmonary bypass-related inflammatory stress, postoperative chest wound pain, indwelling invasive catheters, frequent nocturnal vital sign monitoring in the ICU, mechanical ventilation support, and ward noise interference. Among these, the incidence of sleep disorders (SD) is alarmingly high, ranging from 35% to 68%, and has been recognized as a common yet critical complication that hampers postoperative recovery.3 Poor SQ not only impairs immune function and heightens pain perception but also exacerbates psychological distress, potentially compromising cardiac function recovery, increasing the risk of postoperative adverse events, prolonging hospital stays, and imposing substantial burdens on both patients and their families.4 Physiologically, there is a bidirectional regulatory link between sleep disturbance and autonomic imbalance: insufficient deep sleep persistently suppresses vagal parasympathetic activity and elevates sympathetic excitability, while sustained sympathetic overactivity in turn fragments sleep architecture, forming a vicious cycle that impairs cardiac perioperative recovery. Therefore, timely and accurate assessment of perioperative SD in this population is of paramount clinical significance for enabling early intervention and optimizing postoperative management strategies.
In recent years, heart rate variability (HRV) has emerged as a robust, non-invasive physiological indicator for assessing autonomic nervous system (ANS) function. HRV parameters, derived from electrocardiographic signals, provide valuable insights into the dynamic balance between sympathetic and parasympathetic activities.5,6 Accumulating evidence has demonstrated that patients with abnormal sleep architecture frequently exhibit distinctive alterations in HRV characteristics, suggesting that HRV parameters may serve as potential surrogate biomarkers for SD.7 Moreover, HRV has been extensively utilized in predicting prognosis and guiding management in various cardiovascular conditions.8
Serial perioperative HRV monitoring carries distinctive clinical value: it provides objective, quantitative physiological indicators independent of subjective sleep questionnaires, helps early screen patients at high risk of autonomic dysfunction secondary to sleep loss, and supports clinicians to design targeted vagus nerve activation and sleep improvement interventions. Although HRV parameters have been widely used in the assessment of cardiovascular disease prognosis and other fields.8 However, relevant research on their application in SD patients during the perioperative period of cardiac surgery still needs to be supplemented. The existing research gap remains prominent: most existing studies only observe single time-point HRV indicators in general insomnia or sleep apnea populations, few longitudinal studies track continuous dynamic changes of multi-dimensional HRV indices across preoperative and multiple postoperative days in cardiac surgery patients, and limited evidence quantifies the correlation between PSQI-defined perioperative SD and time/frequency domain HRV parameters.
Based on this, this study analyzed the changes in HRV parameters of patients of undergoing cardiac surgery during the perioperative period, and the relationship between SQ during the perioperative period of patients undergoing cardiac surgery and the parameters of heart rate variability. The clear objective of this study is to explore the dynamic changing rules of multi-timepoint HRV in cardiac surgical patients with perioperative SD and identify core HRV indicators correlated with sleep disturbance; we hypothesize that patients with perioperative SD will have persistently lower parasympathetic HRV indicators (HF, SDANN, PNN50, RMSSD) at T0, T1, T2 and T3 compared with patients without SD. It provided objective physiological evidence for the early identification of patients at high risk of autonomic nerve dysfunction in clinical practice, the formulation of individualized sleep intervention strategies, and the improvement of postoperative cardiac rehabilitation outcomes.
Subjects and Methods
Participants
This study was a prospective observational cohort study. A total of 116 patients who underwent cardiac surgery at Nanjing First Hospital from January 2023 to December 2024. Among them, 21 patients were excluded based on the following criteria: 5 patients had other malignant tumors, 3 patients had infectious diseases or malignant hematological disorders, 4 patients had a surgical history within the past month or rheumatism/immune system disorders, 6 patients had severely abnormal liver and kidney functions, and 3 patients had incomplete clinical or pathological data or refused to participate. Ultimately, 95 patients were enrolled in this study. All enrolled patients were consecutively screened and divided into the SD group (PSQI total score > 7) and the non-SD group (PSQI total score ≤ 7) based on whether they met the Pittsburgh Sleep Quality Index (PSQI)-based diagnostic criteria for sleep disorders during the perioperative period. This study has received approval from the Ethics Committee of Nanjing First Hospital (Approval No.: KY20251110-KS-01; First approval date: October 3, 2022). This study has obtained the informed consent of all the research subjects.
SD diagnostic criteria: The PSQI scale9 was used as the diagnostic tool for SD. This scale consists of 7 dimensions with a total of 18 items, and the total score ranges from 0 to 21. A SD diagnosis is made when the total PSQI score is greater than 7.
Inclusion criteria: (1) Age>18 years; (2) Patients undergoing cardiac surgery with extracorporeal circulation; (3) Normal mental state; (4) Complete clinical data and follow-up information of the patients; (5) Voluntary participation in this study and signing of the informed consent form.
Exclusion criteria: (1) Patients with other malignant tumors; (2) Patients with infectious diseases or malignant hematological disorders; (3) Patients with a surgical history in the past month; Patients with rheumatism or immune system disorders; (4) Patients with severely abnormal liver and kidney functions; (5) Patients whose clinical and pathological data are incomplete or who refuse to participate in this study.
Recruitment of Patients
Patient recruitment was carried out consecutively from January 1, 2023 to December 31, 2024 in the Department of Cardiac Surgery, Nanjing First Hospital. All patients scheduled for cardiac surgery with cardiopulmonary bypass (CPB) were preliminarily screened by trained cardiac ICU nurses within 24 h before admission to the ward. Standardized one-on-one face-to-face interviews were conducted to collect basic demographic information, medical history and perioperative subjective sleep status. First, researchers reviewed electronic medical records to exclude subjects who met any exclusion criteria, including malignant tumors, active infectious diseases, severe hepatorenal dysfunction, autoimmune rheumatic diseases and incomplete clinical data. Eligible patients were fully informed of the study purpose, data collection procedures, potential risks and benefits, and were guaranteed that all personal clinical data would be anonymized and only used for academic research. Patients who voluntarily agreed to participate signed written informed consent before formal enrollment. During the perioperative period, researchers conducted repeated follow-ups at four fixed time nodes: 1 day before the operation (T0), 1 day after the operation (T1), 2 days after the operation (T2), and 3 days after the operation (T3). Patients with unstable vital signs, unexpected secondary surgery or who requested withdrawal during observation were eliminated from the final statistical analysis. According to PSQI total score >7 as the diagnostic threshold for sleep disorders (SD), enrolled participants were divided into SD group and non-SD group. The whole recruitment flow was recorded in a standardized screening log to avoid selection bias as much as possible. A total of 116 patients were initially screened, among whom 21 were excluded due to ineligible conditions or refusal to participate, and the final valid sample included 95 subjects.
Sample Size Calculation
Sample size calculation was performed based on the primary endpoint: intergroup difference in PNN50 at T1, which was the HRV parameter with the strongest correlation with perioperative sleep disorders. According to preliminary clinical observation and published similar perioperative HRV research, we assumed the mean difference of PNN50 between SD and non-SD groups was 3.8%, pooled standard deviation = 1.7%, α=0.05 (two-sided), statistical power (1-β)=0.90, allocation ratio of non-SD:SD ≈ 2:1.
Using two independent samples t-test sample size formula, the minimal required sample size was calculated as 29 patients in SD group and 58 patients in non-SD group, with a total of 87 participants. Considering potential loss to follow-up rate of 10% during 3-day postoperative continuous monitoring, we expanded the target enrollment number to at least 96 patients. In the actual recruitment process, 95 valid complete cases were finally collected, which reached the calculated required sample size and satisfied the statistical power demand for subsequent repeated-measures ANOVA, Lasso regression and point-biserial correlation analysis.
Data Collection Methods
Clinical data, including gender, age, body mass index (BMI), history of diabetes, history of hypertension, smoking history, and drinking history (If the daily intake of pure alcohol exceeds 10 grams, this amount is approximately equivalent to one standard drink, which corresponds to 250 milliliters of beer, 100 milliliters of red wine or 25 milliliters of spirits; or the total weekly intake of pure alcohol exceeds 70 grams), were collected by querying the electronic medical record system of the hospital. For sleep disorder assessment, this study used the PSQI scale,9 which is a validated and widely used questionnaire for evaluating sleep quality. The PSQI scale consists of 7 dimensions with a total of 18 items, and the total score ranges from 0 to 21. The SD diagnosis was made when the total PSQI score was greater than 7. All PSQI assessments were administered by trained research staff during the perioperative period. The HRV parameters at T0, T1, T2, and T3 were mainly obtained through electrocardiogram (ECG) or photoplethysmography (PPG) signals. The detection instruments included medical multi-lead electrocardiograph, dynamic electrocardiogram (Holter), and single-lead electrocardiogram monitoring equipment. After the instruments recorded the electrocardiogram signals, they identified the R-wave peaks and calculated the inter-R-wave intervals (RR intervals) using algorithms (such as software like Kubios, LabChart, etc). Then, using time-domain, frequency-domain, or nonlinear analysis methods, they conducted statistical analysis on the RR interval sequence, and finally calculated the HRV indicators reflecting the autonomic nerve regulatory function. The HRV parameter indicators include frequency domain indicators and time domain indicators. Frequency domain indicators include low-frequency power (LF), high-frequency power (HF) and LF/HF; time domain indicators include standard deviation of NN intervals (SDNN), average standard deviation of normal R-R intervals (SDNNI), standard deviation of all 5-minute RR intervals (SDANN), percentage of successive R-R intervals that differ by more than 50 ms (PNN50), and root mean square of successive RR interval differences (RMSSD).
Electrocardiography to Find Out HRV
HRV measurements were performed consistently at four predefined time points (T0 to T3) following unified standard protocols to minimize confounding interference during signal capture. Before each ECG recording, participants rested supine for 15 minutes in an isolated, temperature-stabilized private ward; they were instructed to refrain from eating, drinking, communicating, and experiencing mood swings to stabilize autonomic activity.
Three validated clinical ECG devices were used to acquire raw cardiac signals, including hospital multi-lead electrocardiographs, 24-hour Holter recorders, and single-lead bedside ECG monitors. A continuous 5-minute recording was obtained for every patient to generate intact RR interval waveforms. All raw signal files were transferred to Kubios HRV Premium 3.5 software for offline centralized processing. The program automatically detected R-wave peaks and applied internal artifact correction algorithms to eliminate disruptive signals such as ectopic beats, motion noise, cough interference, and distortions from loose electrodes, yielding refined RR interval sequences free of noise.
The software autonomously calculated a full set of HRV metrics covering frequency-domain and time-domain categories. Frequency-domain metrics contained low-frequency power (LF, 0.04–0.15 Hz), high-frequency power (HF, 0.15–0.4 Hz), and the LF/HF ratio. Time-domain indicators consisted of SDNN, SDNNI, SDANN, PNN50, and RMSSD. All parameters were exported with uniform units: milliseconds (ms) for time-domain measurements and squared milliseconds (ms2) for frequency-domain power values. Two independent ICU researchers double-verified all datasets before entry into the SPSS 27.0 database to rule out recording mistakes. If over 10% of RR interval signals were uncorrectable artifacts, the entire HRV acquisition was repeated at a later appropriate time.
Sleep Study
Perioperative sleep quality assessment relied on the PSQI, a validated 18-item self-rating scale covering seven dimensions: sleep latency, sleep duration, habitual sleep efficiency, sleep disturbance, use of sleep medication, daytime dysfunction and overall sleep quality. The total PSQI score ranges from 0 to 21; a total score >7 was defined as clinical SD, which served as the grouping standard of this study.
Sleep evaluation was completed at postoperative day 3 (T3) after finishing all four HRV detections. Researchers provided standardized guidance to patients with low literacy or postoperative physical discomfort, reading each item aloud and recording answers on behalf of subjects to ensure scale integrity. Patients with postoperative cognitive dysfunction, delirium or severe visual impairment were excluded from scale assessment at screening stage.
Auxiliary objective sleep observation was supplemented by nursing shift records throughout the perioperative period, including nocturnal awakening frequency, duration of continuous sleep, noise and pain interference factors, to assist PSQI score verification. Patients with inconsistent PSQI scores and actual nursing sleep records were re-interviewed to correct subjective bias. All PSQI questionnaires were coded anonymously and stored separately from HRV data to ensure blind grouping during statistical analysis.
The sleep assessment workflow strictly avoided interference from HRV measurement results; researchers responsible for sleep scoring did not access patients’ HRV parameter data before grouping, reducing assessment bias caused by knowing autonomic nervous function status.
Statistical Methods
Data processing and analysis were conducted using SPSS 27.0. The Shapiro–Wilk (SW) test was used to check the normality of the data. The measurement data met the normal distribution, which were expressed as mean±standard deviation (
). ANOVA was used for comparisons among multiple groups, and t-tests were used for comparisons between two groups. Non-normal distribution data were expressed as M (Q25, Q75), and non-parametric tests were used for comparisons between groups. Count data were expressed as n (%), and χ2 tests were used for comparisons between groups. Repeated measurement data were analyzed using repeated-measures analysis of variance to examine the time effect, group effect, and time×group interaction. Lasso regression was employed to screen the main HRV parameters affecting SD, with optimal lambda values determined by 10-fold cross-validation. Binary logistic regression analysis was conducted to identify protective factors for SD, with results presented as odds ratios (OR) and 95% confidence intervals (CI). Point-biserial correlation analysis was used to examine the correlation between continuous HRV parameters and dichotomous SD status. A difference was considered statistically significant if P < 0.05.
Result
Comparison of Clinical Data Between the SD Group and the Non-SD Group
The comparison of clinical data between the SD group and the non-SD group is shown in Table 1. There was no statistically significant difference in the clinical data between the SD group and the non-SD group (P > 0.05).
Table 1.
Comparison of Clinical Data Between the SD Group and the Non-SD Group
| Indicators | SD Group (n=33) |
Non-SD Group (n=62) |
t/χ2 | P |
|---|---|---|---|---|
| Gender (n, %) | 0.020 | 0.888 | ||
| Male | 17 (51.52) | 31 (50.00) | ||
| Female | 16 (48.48) | 31 (50.00) | ||
Age (year,  ) |
70.57±5.29 | 69.79±5.32 | 0.681 | 0.497 |
BMI (kg/m2,  ) |
25.15±2.73 | 24.96±2.36 | 0.354 | 0.724 |
| History of diabetes (n, %) | 0.039 | 0.843 | ||
| Yes | 14 (42.42) | 25 (40.32) | ||
| No | 19 (57.58) | 37 (59.68) | ||
| History of hypertension (n, %) | 0.072 | 0.788 | ||
| Yes | 22 (66.67) | 43 (69.35) | ||
| No | 11 (33.33) | 19 (30.65) | ||
| Smoking history (n, %) | 0.212 | 0.645 | ||
| Yes | 16 (48.48) | 27 (43.55) | ||
| No | 17 (51.52) | 35 (56.45) | ||
| Drinking history (n, %) | 1.434 | 0.231 | ||
| Yes | 9 (27.27) | 16 (25.81) | ||
| No | 14 (72.73) | 46 (74.19) |
Comparison of HRV Parameters Between the SD Group and the Non-SD Group
The comparison of HRV parameters between the SD group and the non-SD group is shown in Table 2 and Figure 1. There was no statistically significant difference in the LF, LF/HF and SDNNI values between the two groups at T0, T1, T2 and T3 (P > 0.05), while there were statistically significant differences in the HF, SDNN, SDANN, PNN50 and RMSSD values (P < 0.05).
Table 2.
Comparison of HRV Parameters Between the SD Group and the Non-SD Group, (
)
| Indicators | SD Group (n=33) |
Non-SD Group (n=62) |
t | P |
|---|---|---|---|---|
| T0 | ||||
| LF (ms2) | 59.21±9.28 | 60.05±9.31 | −0.419 | 0.676 |
| HF (ms2) | 38.57±7.95 | 43.03±7.96 | −2.600 | 0.010 |
| LF/HF | 3.87±1.03 | 3.84±1.01 | 0.137 | 0.891 |
| SDNN (ms) | 121.73±24.61 | 132.84±24.74 | −2.156 | 0.034 |
| SDNNI (ms) | 38.26±6.91 | 40.96±6.87 | −1.820 | 0.072 |
| SDANN (ms) | 79.27±11.58 | 87.34±10.25 | −3.476 | 0.001 |
| PNN50 (%) | 10.82±2.31 | 13.87±2.35 | −6.134 | <0.001 |
| RMSSD (ms) | 27.35±5.71 | 33.54±5.94 | −4.985 | <0.001 |
| T1 | ||||
| LF (ms2) | 43.61±6.54 | 45.05±7.36 | −0.967 | 0.336 |
| HF (ms2) | 26.38±5.21 | 31.27±6.75 | −3.710 | <0.001 |
| LF/HF | 4.71±1.24 | 4.40±1.16 | 1.227 | 0.223 |
| SDNN (ms) | 91.25±17.36 | 99.27±18.42 | −2.120 | 0.037 |
| SDNNI (ms) | 21.39±4.82 | 23.02±4.62 | −1.586 | 0.116 |
| SDANN (ms) | 62.37±8.67 | 73.68±8.91 | −5.933 | <0.001 |
| PNN50 (%) | 6.31±1.34 | 10.34±1.67 | −12.753 | <0.001 |
| RMSSD (ms) | 16.93±4.93 | 24.73±5.26 | −7.193 | <0.001 |
| T2 | ||||
| LF (ms2) | 44.03±6.63 | 46.57±6.94 | −1.753 | 0.083 |
| HF (ms2) | 27.81±5.68 | 33.54±6.84 | −4.378 | <0.001 |
| LF/HF | 4.64±1.19 | 4.39±1.06 | 1.039 | 0.301 |
| SDNN (ms) | 93.17±19.27 | 101.68±20.25 | −2.027 | 0.046 |
| SDNNI (ms) | 24.41±4.97 | 25.65±5.03 | −1.172 | 0.244 |
| SDANN (ms) | 63.58±8.91 | 75.87±9.06 | −6.314 | <0.001 |
| PNN50 (%) | 6.88±2.62 | 10.07±1.95 | −6.442 | <0.001 |
| RMSSD (ms) | 17.61±5.07 | 25.76±5.42 | −7.279 | <0.001 |
| T3 | ||||
| LF (ms2) | 47.83±7.51 | 49.28±8.25 | −0.862 | 0.391 |
| HF (ms2) | 33.83±6.58 | 38.54±7.04 | −3.249 | 0.002 |
| LF/HF | 4.30±1.05 | 4.26±1.01 | 0.192 | 0.848 |
| SDNN (ms) | 107.26±21.14 | 119.38±21.57 | −2.626 | 0.010 |
| SDNNI (ms) | 29.82±5.16 | 31.43±5.38 | −1.408 | 0.162 |
| SDANN (ms) | 67.94±9.06 | 73.67±10.24 | −2.762 | 0.007 |
| PNN50 (%) | 8.04±1.98 | 12.41±2.07 | −10.353 | <0.001 |
| RMSSD (ms) | 20.78±5.29 | 24.56±5.81 | −3.163 | 0.002 |
Figure 1.
The heart rate variability of patients in the SD group and the non-SD group at each stage.
Repeated-Measure Variance Analysis of HRV Parameters
The results of the repeated-measure variance analysis of HRV parameters in the SD group are shown in Table 3. The study found that all HRV parameters exhibited a time effect (P < 0.05), while LF, HF, LF/HF, SDNNI, PNN50 and RMSSD showed a group effect (P < 0.05). Additionally, SDANN and RMSSD exhibited an interaction effect of time×group (P < 0.05).
Table 3.
Repeated-Measure Variance Analysis of HRV Parameters
| Indicators | Time Effect | Group Effect | Time×Group |
|---|---|---|---|
| LF | F=63.35 P<0.001 | F=4.09, P=0.046 | F=0.16, P=0.922 |
| HF | F=53.82, P<0.001 | F=39.36, P<0.001 | F=0.14, P=0.937 |
| LF/HF | F=42.53, P<0.001 | F=18.31 P<0.001 | F=0.19, P=0.905 |
| SDNN | F=7.31, P<0.001 | F=1.84, P=0.178 | F=0.37, P=0.776 |
| SDNNI | F=159.16, P<0.001 | F=9.36, P=0.002 | F=0.27, P=0.848 |
| SDANN | F=44.82, P<0.001 | F=2.08, P=0.102 | F=77.17, P<0.001 |
| PNN50 | F=74.67, P<0.001 | F=227.03 P<0.001 | F=2.18, P=0.09 |
| RMSSD | F=59.36, P<0.001 | F=94.84, P<0.001 | F=3.01, P=0.031 |
Lasso Regression Model of the HRV Parameters Related to SD
The HRV parameters related to SD were selected through the Lasso regression model. The results are shown in Figure 2. The results indicate that among the HRV parameters, HF (T2) SDANN (T1), SDANN (T2), PNN50 (T0), PNN50 (T1), PNN50 (T3), RMSSD (T0), RMSSD (T1), and RMSSD (T2) were related to SD in patients.
Figure 2.
The HRV parameters related to SD were selected using the Lasso regression model. (A) Cross-validation plot of Lasso regression; (B) Path diagram of variable selection in Lasso regression; (C) Predictive factors selected by Lasso and their corresponding coefficients.
Logistic Regression Analysis of the Main HRV Parameters Related to SD
The main HRV parameters obtained by Lasso regression were included in the Logistic regression analysis as the independent variable, while whether SD occurred was taken as the dependent variable for the Logistic regression analysis. The results are shown in Table 4. The study indicates that HF (T2) SDANN (T1), SDANN (T2), PNN50 (T0), PNN50 (T1), PNN50 (T3), RMSSD (T0), RMSSD (T1), and RMSSD (T2) were negatively correlated with for SD.
Table 4.
Logistic Regression Analysis of the Main HRV Parameters Related to SD
| Indicators | B | SE | Wald χ2 | OR | 95% CI | p |
|---|---|---|---|---|---|---|
| HF (T2) | −0.138 | 0.039 | 12.428 | 0.870 | 0.805~0.940 | <0.001 |
| SDANN (T1) | −0.143 | 0.032 | 19.379 | 0.866 | 0.813~0.923 | <0.001 |
| SDANN (T2) | −0.142 | 0.031 | 20.416 | 0.867 | 0.815~0.922 | <0.001 |
| PNN50 (T0) | −0.532 | 0.118 | 20.321 | 0.586 | 0.465~0.739 | <0.001 |
| PNN50 (T1) | −1.849 | 0.416 | 19.673 | 0.157 | 0.069~0.356 | <0.001 |
| PNN50 (T3) | −0.927 | 0.185 | 25.016 | 0.395 | 0.274~0.568 | <0.001 |
| RMSSD (T0) | −0.183 | 0.046 | 15.567 | 0.831 | 0.759~0.911 | <0.001 |
| RMSSD (T1) | −0.313 | 0.067 | 21.548 | 0.731 | 0.640~0.834 | <0.001 |
| RMSSD (T2) | −0.270 | 0.055 | 23.450 | 0.762 | 0.683~0.851 | <0.001 |
Point-Biserial Correlation Analysis of HRV Parameters and SD
The point-biserial correlation analysis between heart rate variability parameters and SD is shown in Figure 3. The results indicate that the heart rate variability indicators included in the correlation analysis were all significantly negatively correlated with SD (correlation coefficient range: −0.40 to −0.78). Among them, PNN50 (T1) had the strongest negative correlation with SD (r=−0.78).
Figure 3.
Point-biserial correlation analysis of HRV parameters and SD. * indicates p<=0.05, ** indicates p<=0.01, *** indicates p<=0.001.
Discussion
This study found that the Patients with SD exhibited significantly lower HRV parameters (HF, SDNN, SDANN, PNN50, RMSSD) across multiple postoperative time points compared to non-SD patients. Moreover, HRV parameters in the SD group showed dynamic changes over time, with a sharp decrease at T1 followed by gradual recovery at T2 and T3. Additionally, The research also found that HF, SDANN, PNN50, and RMSSD were significant negative correlations with SD occurrence. These were similar to the research results reported by current study,10 indicating that sleep disorders are associated with altered the autonomic nerve balance status of patients after cardiac surgery. This is mainly attributed to the imbalance in sympathetic and parasympathetic nerve regulation caused by surgical trauma, pain stress and abnormal sleep structure.11,12
A cross-sectional survey conducted by Lee et al13 in South Korea on pediatric obstructive sleep apnea patients confirmed that parasympathetic HRV indicators HF, RMSSD and PNN50 declined markedly in subjects with poor sleep quality, which matched our finding that parasympathetic activity markers were persistently suppressed in cardiac surgery patients with perioperative sleep disturbance. A Japanese cohort study by Suzuki et al12 on patients with sleep behavior disorders also reported sustained reduction of time-domain HRV indices under sympathetic overactivation, supporting our conclusion that SD is closely linked to weakened vagal tone. A clinical trial from Taiwan region of China10 further noted that patients with preoperative poor sleep presented lower HF and RMSSD under intraoperative monitoring, consistent with our T0 baseline difference between SD and non-SD groups, indicating sleep dysfunction correlates with impaired autonomic regulation preoperatively and aggravates postoperative nerve imbalance cumulatively. Collectively, evidence from these neighboring regions echoes our core result that parasympathetic HRV metrics are sensitive markers for sleep disruption, while most overseas studies focused on non-cardiac surgical populations; our work innovatively describes perioperative dynamic HRV trajectories specifically for cardiac surgery patients, filling regional research gaps in cardiovascular perioperative sleep-autonomic interaction.
This study found that in the SD group, parameters such as LF, HF, and SDNN all showed significant decreases at the T1 time point, while the LF/HF ratio significantly increased. This is consistent with the research conclusion of the previous study.14 On the dynamic changes of autonomic nerve function after surgery, indicating that there is a significant inhibitory phenomenon of autonomic nerve function in the acute phase after cardiac surgery. This may be caused by the relative hyperactivity of the sympathetic nerve and the impaired activity of the vagus nerve caused by the surgical stress response.15 Additionally, this study also found that the HRV parameters of the SD group showed a slow recovery trend at the T2 and T3 time points. This shows that postoperative autonomic nerve function has certain plasticity, which is primarily due to the relief of the body’s stress state and the gradual reconstruction of the normal sleep-wake cycle.16 Moreover, the study found that parameters reflecting parasympathetic nerve activity, such as PNN50 and RMSSD, remained at a relatively low level in the SD group. This reveal that impaired parasympathetic nerve function may co-occur alongside sleep disorders and together hinder cardiac rehabilitation, which might be the result of the continuous inhibition of vagus nerve tension caused by insufficient sleep depth.17 Studies from Southeast Asian nations further supplement this mechanism: a Malaysian systematic review3 summarized that cardiac surgery patients with sleep disorders sustained elevated LF/HF within 72 hours postoperatively, and vagal dysfunction prolonged postoperative recovery duration, aligning with our T1-T3 slow recovery trend of HRV indicators in the SD group. Meanwhile, this study observed that parasympathetic indicators PNN50 and RMSSD remained suppressed throughout the perioperative period in SD patients. This reveals that persistent parasympathetic damage is closely associated with sleep disorders hinder cardiac rehabilitation, likely caused by chronically diminished vagal tension secondary to insufficient deep sleep, a viewpoint also validated by a Singapore-based sleep HRV review14 highlighting PNN50 as the most responsive marker for sleep deprivation-related autonomic impairment.
The Lasso regression and logistic regression analyses identified HF (T2), SDANN (T1/T2), PNN50 (T0/T1/T3), and RMSSD (T0/T1/T2) were closely associated to SD. This discovery suggests that the assessment of SQ could serve as an effective indicator linked to the autonomic nerve function status of patients, thereby providing a basis for early clinical intervention. The point-biserial correlation analysis further confirmed the negative association between these parameters and SD, with PNN50 (T1) showing the strongest correlation (r = −0.78), highlighting its potential as a sensitive biomarker.
From a physiological perspective, the protective effect of these HRV parameters can be explained by their reflection of parasympathetic activity. HF, PNN50, and RMSSD are well-established indicators of vagal modulation,5,6 while SDANN reflects long-term circadian influences on autonomic function. Higher values of these parameters correspond to preserved parasympathetic function, which is essential for maintaining normal sleep architecture and promoting postoperative cardiac recovery. Importantly, given the observational case-control design of this study, we cannot establish causal relationships between sleep disturbance and altered HRV; the identified correlations do not confirm that elevated parasympathetic HRV parameters act as protective causal factors against sleep disorders, only that the two sets of variables are statistically related.
This study has several clinical implications. First, dynamic monitoring of HRV parameters, particularly parasympathetic indicators, can provide objective physiological evidence for early identification of patients at risk for SD-related autonomic dysfunction, based on observed correlational patterns rather than proven causal effects. Second, the identification of specific protective factors (HF, SDANN, PNN50, RMSSD) may guide the development of targeted interventions aimed at enhancing vagal activity, such as relaxation training, controlled breathing exercises, or pharmacological approaches. Third, the time-dependent changes observed in this study underscore the importance of continuous rather than single-point HRV assessment in postoperative care. However, the study has the following limitations: Firstly, the sample size is relatively small and it is a single-center study, which may have selection bias; Secondly, HRV parameters are susceptible to interference from drugs, emotions, and environmental factors. Although confounding factors have been controlled, there may still be unmeasured confounding variables affecting the accuracy of the results; Thirdly, all comorbid disease data including hypertension and diabetes were collected based on historical medical records and patient self-report rather than self-conducted biochemical or blood pressure screening in this study; Fourthly, the prospective observational cohort study only permits conclusions about statistical association, and cannot clarify causal pathways between sleep disorders and autonomic dysfunction; Finally, the study did not deeply explore the specific molecular mechanism between sleep disorders and autonomic nerve dysfunction. Further verification can be conducted through multi-center large-sample studies combined with biomarker detection.
Conclusion
The results of this study indicate a statistically significant association between the sleep quality during the perioperative period of cardiac surgery patients and heart rate variability. Specifically, compared to patients without sleep disorders, patients with sleep disorders exhibited significant reductions in several HRV parameters (including HF, SDNN, SDANN, PNN50, and RMSSD), and these parameters remained persistently low at multiple postoperative time points. Lasso regression and Logistic regression analyses further identified HF (T2), SDANN (T1/T2), PNN50 (T0/T1/T3), and RMSSD (T0/T1/T2) as HRV indicators significantly associated with sleep disorders. Point-biserial correlation analysis also confirmed that these parameters were negatively correlated with sleep disorders (correlation coefficient range: −0.40 to −0.78). However, this study is an observational case-control design, and can only reveal the statistical association between variables, unable to establish the causal direction between sleep disorders and HRV changes. These HRV parameters, as objective physiological indicators reflecting the state of autonomic nerve function, may serve as an auxiliary reference for early clinical identification of autonomic nerve dysfunction during the perioperative period. However, the claim that these parameters can serve as an “effective window” to clearly reflect the state of autonomic nerve function in patients requires more prospective, interventional studies, and large-sample multicenter evidence for verification. Future research can combine dynamic monitoring and biological markers to further explore its clinical application value and potential mechanisms.
Funding Statement
This research was supported by Science and Technology Development Fund of Nanjing Medical University (NMUB20210211).
Data Sharing Statement
The data that support the findings of this study are available from the corresponding author upon reasonable request.
Ethical Statement
This study complies with the requirements of the World Medical Association’s “Helsinki Declaration” and relevant laws and regulations in China. It has been approved by the Ethics Committee of Nanjing First Hospital [Approval Number: KY20251110-KS-01].
Author Contributions
All authors made a significant contribution to the work reported, whether that is in the conception, study design, execution, acquisition of data, analysis and interpretation, or in all these areas; took part in drafting, revising or critically reviewing the article; gave final approval of the version to be published; have agreed on the journal to which the article has been submitted; and agree to be accountable for all aspects of the work.
Disclosure
The authors declare that they have no conflicts of interest.
References
- 1.Wang BB, Liu Q, Dou D, et al. Association between sleep disorder and delirium in adult patients undergoing cardiac surgery: a retrospective cohort study. Bmc Anesthesiology. 2025;25(1). doi: 10.1186/s12871-025-03243-3 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Fan P, Zhang X, Zhai Q, et al. Perioperative sleep quality in pediatric cardiac surgery: a scoping review. Sleep Medicine. 2026;138. [Google Scholar]
- 3.Soh PQP, Wong WHT, Roy T, Tam WWS. Effectiveness of non-pharmacological interventions in improving sleep quality after cardiac surgery: a systematic review and meta-analysis. J Clin Nurs. 2024;33(6):2084–12. doi: 10.1111/jocn.17115 [DOI] [PubMed] [Google Scholar]
- 4.Akar ZA, Yildirim D, Karakoyun Ö, et al. The mediating role of sleep quality in the association between inflammatory disease activity and health-related quality of life in rheumatoid arthritis. Front Med. 2026;13. doi: 10.3389/fmed.2026.1797652 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Noronha Osório D, Viana-Soares R, Marto JP, et al. Autonomic nervous system response to remote ischemic conditioning: heart rate variability assessment. BMC Cardiovasc Disord. 2019;19(1):211. doi: 10.1186/s12872-019-1181-5 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Keim OC, Bolwin L, Feldmann RE, Thiel M, Benrath J. Heart rate variability as a predictor of intraoperative autonomic nervous system homeostasis. J Clin Monit Comput. 2024;38(6):1305–1313. doi: 10.1007/s10877-024-01190-x [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Ma Y, Mullington JM, Wayne PM, Yeh GY. Heart rate variability during sleep onset in patients with insomnia with or without comorbid sleep apnea. Sleep Med. 2024;122:92–98. doi: 10.1016/j.sleep.2024.07.034 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Gu X, Wang P, He X, Guo F, Wang YC, Chen Y. Latent profile analysis of quality of life in patients undergoing on-pump cardiac surgery and its perioperative risk factors. Bmc Cardiovascul Disord. 2026;26(1). doi: 10.1186/s12872-026-05827-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Hinz A, Friedrich M, Schulte T, Petrowski K, Tibubos AN, Hartung TJ. The Pittsburgh Sleep Quality Index (PSQI) applied to cancer patients: psychometric properties and factors affecting sleep quality. Cancer Invest. 2025;43(1):103–113. doi: 10.1080/07357907.2024.2446941 [DOI] [PubMed] [Google Scholar]
- 10.Ho CN, Fu PH, Chen JY, et al. Heart rate variability and surgical pleth index under anesthesia in poor and normal sleepers. J Clin Monit Comput. 2020;34(6):1311–1319. doi: 10.1007/s10877-019-00450-5 [DOI] [PubMed] [Google Scholar]
- 11.Kong SDX, Hoyos CM, Phillips CL, et al. Altered heart rate variability during sleep in mild cognitive impairment. Sleep. 2021;44(4). doi: 10.1093/sleep/zsaa232 [DOI] [PubMed] [Google Scholar]
- 12.Suzuki M, Nakamura T, Ohba C, et al. Decreased heart rate variability in sympathetic dominant states in Parkinson’s disease and isolated REM sleep behavior disorder. Parkinsonism Relat Disord. 2024;124:107020. doi: 10.1016/j.parkreldis.2024.107020 [DOI] [PubMed] [Google Scholar]
- 13.Lee LA, Chuang HH, Hsieh HS, et al. Using sleep heart rate variability to investigate the sleep quality in children with obstructive sleep apnea. Front Public Health. 2023:11. doi: 10.3389/fpubh.2023.1103085 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Yu R, Li Y, Zhao K, Fan F. A review of automatic sleep stage classification using machine learning algorithms based on heart rate variability. Sleep Biol Rhythms. 2025;23(2):113–125. doi: 10.1007/s41105-024-00563-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Masuda Y, Kawase M, Kitano R, et al. Peripheral cholecystokinin promotes thermogenesis via vagal afferent signalling, hypothalamic oxytocin and sympathetic outflow to brown adipose tissue. J Physiol-London. 2026;604(12):4651–4677. doi: 10.1113/jp290717 [DOI] [PubMed] [Google Scholar]
- 16.Carpenter JS, Crouse JJ, Park SH, et al. Actigraphy-derived circadian rhythms, sleep-wake patterns, and physical activity across clinical stages and pathophysiological subgroups in young people presenting for mental health care. J Psychiatr Res. 2025;186:396–406. doi: 10.1016/j.jpsychires.2025.03.003 [DOI] [PubMed] [Google Scholar]
- 17.Li Q, Shi M, Steward CJ, Che K, Zhou Y. A comparison between pre-sleep heart rate variability biofeedback and electroencephalographic biofeedback training on sleep in national level athletes with sleep disturbances. Appl Psychophysiol Biofeedback. 2024;49(1):115–124. doi: 10.1007/s10484-023-09604-3 [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Data Availability Statement
The data that support the findings of this study are available from the corresponding author upon reasonable request.