Discrimination at work is linked to high blood pressure. Icy fingers and toes: Poor circulation or Raynaud's phenomenon? When it comes to your health, it is now easy to measure and track all kinds of information. In the comfort of our homes, we can check our weight, blood pressure, number of steps, calories, heart rate, and blood sugar.

Researchers have been exploring another data point called heart rate variability HRV as a possible marker of resilience and behavioral flexibility. HRV is simply a measure of the variation in time between each heartbeat.

This variation is controlled by a primitive part of the nervous system called the autonomic nervous system ANS. It works behind the scenes, automatically regulating our heart rate, blood pressure, breathing, and digestion among other key tasks. The ANS is subdivided into two large components: the sympathetic and the parasympathetic nervous system, also known as the fight-or-flight mechanism and the relaxation response.

The brain is constantly processing information in a region called the hypothalamus. The ANS provides signals to the hypothalamus, which then instructs the rest of the body either to stimulate or to relax different functions.

It responds not only to a poor night of sleep, or that sour interaction with your boss, but also to the exciting news that you got engaged, or to that delicious healthy meal you had for lunch. Our body handles all kinds of stimuli and life goes on. However, if we have persistent instigators such as stress, poor sleep, unhealthy diet, dysfunctional relationships, isolation or solitude, and lack of exercise, this balance may be disrupted, and your fight-or-flight response can shift into overdrive.

The gold standard is to analyze a long strip of an electrocardiogram done in the doctor's office. But in recent years, companies have launched apps and wearable heart rate monitors that do something similar.

The accuracy of these methods is still under scrutiny, but the technology is improving. If you do wish to give it a try, chest strap monitors tend to provide a more accurate measure of HRV than wrist devices.

HRV may offer a noninvasive way to signal imbalances in the autonomic nervous system. Based on data gathered from many people, if the system is in more of a fight-or-flight mode, the variation between subsequent heartbeats tends to be lower.

If the system is in more relaxed state, the variation between beats may be higher. This suggests some interesting possibilities. People who have a high HRV may have greater cardiovascular fitness and may be more resilient to stress. HRV may also provide personal feedback about your lifestyle and help motivate those who are considering taking steps toward a healthier life.

You might see a connection to HRV changes as you incorporate more mindfulness, meditation, sleep, and especially physical activity into your life. For those who love data and numbers, this could be a way to track how your nervous system is reacting not only to the environment, but also to your emotions, thoughts, and feelings.

There are questions about the accuracy, reliability and overall usefulness of tracking HRV. While HRV has been linked to overall physical fitness, the correlation between changes in HRV and how your autonomic nervous system is functioning will require much more research.

Still, if you decide to use HRV as another piece of health data, do not get too confident if you have a high HRV, or too worried if your HRV is low. Think of HRV as another way you might tap into your body and mind are responding to what your daily experiences.

As a service to our readers, Harvard Health Publishing provides access to our library of archived content. Please note the date of last review or update on all articles. No content on this site, regardless of date, should ever be used as a substitute for direct medical advice from your doctor or other qualified clinician.

With this Special Health Report, Living Better, Living Longer , you will learn the protective steps doctors recommend for keeping your mind and body fit for an active and rewarding life.

Thanks for visiting. Don't miss your FREE gift. The Best Diets for Cognitive Fitness , is yours absolutely FREE when you sign up to receive Health Alerts from Harvard Medical School. This is Part Four of a five-part series on total load, the cumulative training and life stress an athlete experiences while training.

Read Part One , Part Two , and Part Three to get the full story. HRV is a valuable tool for athletes, as it allows a daily assessment of total stress which can be used to gauge current state of recovery, as well as what kind of training your body is ready for.

Three most common factors that create nutritional stress in athletes are:. Anyone who has exercised for more than a couple of hours in a warm environment may have noticed that their heart rate may continue to increase despite remaining at the same pace or power.

Any upward trend in heart rate while maintaining the same load on the body indicates a shift from parasympathetic to sympathetic dominance and an acute increase in stress. If the stress state brought on by dehydration continues, then recovery is delayed relative to an athlete who is adequately hydrated.

In , researchers from universities in Chile studied the effect of dehydration, both on HRV and on resting metabolic rate RMR.

RMR, or the rate that your body consumes energy in a resting state, is known to stay at a high level after intensive exercise, and the researchers wanted to see whether this was also impacted by dehydration.

Fourteen male college-level athletes were weighed, had their HRV and resting metabolic rate measured, and had their urine tested. The athletes were asked to exercise intensively in 32C 90F environment with circuit training exercises for one minute bursts with one minute of rest in between for 45 minutes total.

The aim of this was to dehydrate the athletes by 3. The athletes were then split into two equal groups, one of which received 1.

They found that:. This is especially true in a hot environment in order to minimize the dehydration contribution to total load. Although dehydration would also be the first, and perhaps the only reason, that athletes typically think of for not consuming alcoholic drinks, there are, in fact, several ways in which alcohol adds to the total stress load of the body:.

They showed that HRV declined with increasing weekly alcohol consumption:. This data shows a minimal impact for the recommended two small drinks per day, but by the time research subjects were consuming twice that amount, HRV was reduced by around 10 percent.

In , researchers from universities in Chile studied the effect of dehydration, both on HRV and on resting metabolic rate RMR. RMR, or the rate that your body consumes energy in a resting state, is known to stay at a high level after intensive exercise, and the researchers wanted to see whether this was also impacted by dehydration.

Fourteen male college-level athletes were weighed, had their HRV and resting metabolic rate measured, and had their urine tested. The athletes were asked to exercise intensively in 32C 90F environment with circuit training exercises for one minute bursts with one minute of rest in between for 45 minutes total.

The aim of this was to dehydrate the athletes by 3. The athletes were then split into two equal groups, one of which received 1.

They found that:. This is especially true in a hot environment in order to minimize the dehydration contribution to total load. Although dehydration would also be the first, and perhaps the only reason, that athletes typically think of for not consuming alcoholic drinks, there are, in fact, several ways in which alcohol adds to the total stress load of the body:.

They showed that HRV declined with increasing weekly alcohol consumption:. This data shows a minimal impact for the recommended two small drinks per day, but by the time research subjects were consuming twice that amount, HRV was reduced by around 10 percent.

Doubling consumption again reduces HRV by a very significant 25 percent compared to consuming no alcohol at all.

There are several adjustments that can help minimize the impact of alcohol on recovery:. For example, inflammation is necessary for muscle growth as it participates in protein breakdown, removal of damaged muscle fibres, and production of prostaglandins.

However, a lot of the time the body is struggling with low grade chronic inflammation that adds to total load. What does this have to do with nutrition you may ask?

Well, take a look at this example of ithlete HRV readings taken by leading HRV researcher Dr. Andrew Flatt during a week when he relaxed his usual healthy diet in favour of high glycaemic, highly processed, and refined foods known to promote inflammation:.

Sleep, a vital brain phenomenon, significantly affects both ANS and metabolic function. Objectives: This study investigated the relationships among sleep, ANS and metabolic function in type 2 diabetes mellitus T2DM , to support the evaluation of ANS function through heart rate variability HRV metrics, and the determination of the correlated underlying autonomic pathways, and help optimize the early prevention, post-diagnosis and management of T2DM and its complications.

Materials and methods: A total of 64 volunteered inpatients with T2DM took part in this study. Conclusions: HRV metrics during sleep period play more distinct role than during awake period in investigating ANS dysfunction and metabolism in T2DM patients, and sleep rhythm based HRV analysis should perform better in ANS and metabolic function assessment, especially for glycemic control in non-linear analysis among T2DM patients.

Type 2 diabetes mellitus T2DM is a chronic hyperglycemia that causes physiological dysfunction and failure of various organs. The dysfunction of autonomic nervous system ANS usually manifests first in vagal nerve the longest parasympathetic nerve in the body, responsible for nearly three-quarters of parasympathetic activity , and damage to vagal nerve leads to resting tachycardia and an overall decrease in parasympathetic tone Balcıoğlu and Müderrisoğlu, Dysfunction of ANS, caused by T2DM, happens in any part of ANS from early stages of diabetes, which damages the cardiovascular, gastrointestinal, genitourinary and neurovascular systems Pfeifer et al.

The function of healthy ANS has obvious circadian rhythm Cui et al. Consequently, diabetes is often followed by sleep disorders Barone and Menna-Barreto, Meanwhile, sleep disorders accelerate the development of T2DM by worsening the metabolic control, which forms into a vicious spiral Barone and Menna-Barreto, Researchers have found that poor sleep quality Martyn-Nemeth et al.

are associated with poor diabetic control. Also, sleep disorders are strongly related to ANS function of T2DM patients Jordan et al.

The activation of the sympathetic nervous system SNS and the unbalance of ANS in patients with T2DM could be mediated through sleep impairments, including nocturnal breathing disturbances Punjabi and Polotsky, ; López-Cano et al.

Heart rate variability HRV is considered to be an effective measure of heart-brain interaction and tension of ANS Montano et al. Cardiologists tend to analyze h HRV, whereas internists tend to detect whether HRV is abnormal in different physiological states, known as Ewing test Gerritsen et al.

Despite the full utilization in clinical fields, HRV analysis in current researches mainly focused on linear analysis of h or 5—15 min electrocardiogram ECG collected in different physiological states, which fails to effectively extract the information in long-term signals and neglects the space-time complexity and fractal properties of heart rate time series.

Meanwhile, it was a general problem that HRV metrics were accepted as measures of autonomic function without examination of the underlying physiological patterns Stein and Pu, and autonomic pathways. Hence, controversial research results were presented.

For example, researchers have found that low-frequency component of HRV was not predominantly influenced by SNS, but vagal nerve system Reyes del Paso et al. Therefore, based on the complex associations among sleep, ANS and T2DM, we monitored h ECG, clinical metabolic indicators, sleep quality as well as sleep staging results of T2DM patients.

Our objective was to help optimize the early prevention, post-diagnosis and management of T2DM and its complications. The study was conducted form July to January Because of the small number of female inpatients in the study only 4 , and considering that women in the same age group may be affected metabolically by perimenopausal syndrome, and decreasing the bias brought by sex Nunan et al.

Prior to the study, all subjects were informed of the experimental protocol and precautions, and signed the written informed consent. Demographic characteristics of the subjects are demonstrated in Table 1.

The Strengthening the Reporting of Observational Studies in Epidemiology STROBE guidelines for cross-sectional studies were followed for study reporting Cuschieri, TABLE 1. Characteristics of the study cohort. Data are presented as mean ± standard deviation SD if the variable is normally distributed, otherwise presented as median p25, p During hospitalization, subjects underwent clinical examinations to obtain their metabolic function and assess their health status.

Upon admission, a fasting blood draw and urinary sample were obtained the next morning for routine glucose, lipid and renal panels. Subjects were also involved in investigation for diabetic complications. Clinical indicators of metabolic function analyzed in this study and diabetic complications statistics are demonstrated in Table 1.

Subsequently, clinical indicator outlier rejection was performed where box plot was utilized. The data were rearranged from largest to smallest, with the difference between the upper quartile U and the lower quartile L defined as IQR.

Nevertheless, some data that did not appear to be gross errors, i. An HbA1c level greater than 9. ECG recordings were collected by an FDA U. Food and Drug Administration approved ambulatory electrocardiogram monitor DynaDx Corporation, Mountainview, CA, United States with a computer-based data-acquisition system.

The ECG recording equipment was a single-lead Holter device that can record ECG for over 24 h. Sampling frequency of ECG monitoring was set to Hz.

All subjects were monitored in hospital for h, starting at 10 p. on the second day of hospitalization. The h ECG data was analyzed for HRV, which will be reported elsewhere.

All ECG recordings were carefully checked with noise level, artifacts, and ectopic beats. Seventeen ECG recordings were discarded due to low data quality, too short recording time, or the fact that the inpatients had atrial fibrillation or server arrhythmia.

The band-pass Butterworth filter and zero-phase shift filter with a cut-off frequency of 5 Hz—35 Hz were first used to eliminate the baseline drift, power frequency interference, myoelectric interference, motion artifacts and equipment noise.

At present, plenty of R-peak detection methods have been proposed on ECG signal analysis. However, when applying these detectors on ECG signals collected in long-term recordings especially via wearable single-lead ECG devices, the R-peak detection accuracies were usually unsatisfying.

To guarantee the accuracy of R-peak detection, a method for extracting high-quality RR intervals proposed in our previous study Cui et al.

Outlier rejection for RR interval time series was performed, and box plot was utilized to find and reject outliers. In this study, we evaluated HRV based on the widely-used linear and non-linear HRV metrics Task Force of the European Society of Cardiology the North American Society of Pacing Electrophysiology, ; Sassi et al.

Linear HRV metrics include time-domain metrics and frequency-domain metrics. Time-domain metrics quantify the amount of variability of the time period between successive heartbeats RR intervals , and include HR-mean mean of heart rate , SDNN standard deviation of RR intervals , RMSSD root mean square of successive RR intervals differences , SDSD standard deviation of successive RR interval differences , SDANN the standard deviation of the average value every 5 min in the RR interval and pNN50 the percentage of successive RR intervals that differ by more than 50 ms.

Frequency-domain metrics estimate the distribution of absolute or relative power at different frequency bands. Generally, the whole frequency band whose absolute power is noted as TP is divided into three frequency bands: high-frequency band 0. The widely used HRV analysis model for assessing ANS homeostasis includes three core statements Pagani et al.

Non-linear analysis utilized in this study includes approximate entropy ApEn Pincus, , sample entropy SampEn Chen et al. Calculation steps are introduced in the supplementary material. ApEn is a rather conventional measure to quantify irregularity and complexity and reflects the probability of new subsequences, and the more complex time series corresponds to larger ApEn.

SampEn is a modification of ApEn, and is more accurate than ApEn in the case of less data. FuzzyEn is improved on the basis of SampEn, which is the entropy of a fuzzy set that contains vagueness and ambiguity uncertainties.

When calculating ApEn, the embedding dimension m was set to 2, and the similar tolerance r was set to 0. MSE analyzes the complexity of time series from multiple time scales.

The 25 coarse-grained time series are contracted from the original RR time series, in which we obtained 28 indices, including the SampEn value at each coarse-graining scale and 3 complexity metrics, MSEsum5, MSEsum10 and MSEsum20 defined as the area of the MSE curve at scales 1—5, 1—10 and 1—20 respectively.

The MSE curve usually rises rapidly at lower scales, peaks at scale 5, and then declines slowly, reaching a plateau after scale 20, therefore calculating the area of the MSE curve at scales 1—5 and 1—20 can portray the complexity characteristics of this curve at high and low resolutions.

Similarly, MFE were calculated based on FuzzyEn. DFA is a non-linear fractal analysis tool to discover potential self-similarity in the series and quantify the fractal scale of the time series. The α1 and α2 portray the short- and long-range correlation respectively.

Subjective sleep quality was assessed by Pittsburgh Sleep Quality Index PSQI and a brief sleep log was used to record sleep duration for the studied night.

PSQI includes multiple sleep-related variables over the preceding month, using Likert and open-ended response formats Spira et al. The PSQI yields seven component scores: subjective sleep quality, sleep latency, sleep duration, habitual sleep efficiency, sleep disturbances, sleep medication, and daytime dysfunction.

Component scores range from 0 to 3 and are summed to obtain a global score, which ranges from 0 to Higher scores suggest greater sleep disturbance Buysse et al. During h ECG monitoring, 53 subjects filled in PSQI questionnaire. Objective sleep quality and sleep staging were assessed by ECG-based cardiopulmonary coupling CPC analysis Thomas et al.

The ECG recordings during sleep at night were extracted for sleep analysis in this study. CPC analysis is based on mathematical analysis of the coupling between HRV and the respiratory modulation of QRS waveform on a beat-to-beat basis.

Major physiological sleep states derived from CPC analysis include stable sleep indicated by high-frequency coupling, or HFC , unstable sleep indicated by low frequency coupling, or LFC , and rapid eye movement REM or wakeful states indicated by very low frequency coupling, or VLFC Thomas et al.

In this study, 38 CPC sleep reports were available. Sleep quality metrics analyzed in this study are demonstrated in Table 2. TABLE 2.

Sleep quality metrics analyzed in this study. Data are presented as mean ± SD if the variable is normally distributed, otherwise presented as median p25, p The data were analyzed by SPSS version The statistical analysis included significance analysis and correlation analysis.

For two normally distributed continuous variables, the Pearson correlation analysis was utilized to estimate the level of correlation, otherwise the Spearman analysis. Partial correlation analysis was also involved in the assessment of correlations between variables.

For normally distributed variables, group differences were compared by independent sample t -test 2 groups and one-way ANOVA 3 or more groups , otherwise Mann-Whitney U test 2 groups or Kruskal-Wallis test 3 or more groups were used.

Missing data was completely at random, and pairwise deletion was conducted during the statistical analyses. There are 25 scales and 3 complexity metrics for MSE and MFE respectively, and in order to demonstrate their overall correlations with clinical indicators in a straightforward way, we define the correlations between MSE and clinical indicators as mean ± SD of the correlations between SampEn at specific scales and clinical indicators.

Similarly, the correlations between MFE and clinical indicators can be calculated. As for complexity metrics, the correlations between MSE complexity and clinical indicators are defined as mean ± SD of the correlations between specific complexity metrics MSEsum5, MSEsum10 and MSEsum20 of the same significant level and clinical indicators, and similarly the correlations between MFE complexity and clinical indicators could be calculated.

Table 3 demonstrates the correlation coefficients between sleep quality metrics and clinical indicators. TABLE 3. Correlation coefficients between sleep quality metrics and clinical metabolic indicators. Table 4 demonstrates the mean values of HRV metrics for 40 T2DM patients, which showed significant differences between sleep and awake periods, especially non-linear HRV metrics.

During sleep period, significant correlations could be observed between many HRV metrics and DBP e. However, during awake period, the correlations between HRV metrics and clinical indicators decreased observably, and HRV metrics mainly correlated with liver function indicators ALT e.

TABLE 5. Correlation coefficients between HRV metrics during different recording period and clinical indicators. It could be seen that the overall level of correlation coefficients was significantly higher during sleep period than h and awake periods.

To investigate the relationship between HRV metrics ANS function in different sleep stages and the clinical indicators metabolic function of T2DM patients, the RR interval time series during sleep period were segmented according to sleep stages, including unstable sleep, stable sleep and REM sleep, and series from the same sleep stage were stitched together as a whole.

Significant differences were observed among HRV metrics in different sleep stages. More details are shown in Supplementary Table S2. The correlation coefficients between HRV metrics in different sleep stages and clinical indicators are demonstrated in Table 6. TABLE 6.

Correlation coefficients between HRV metrics in three sleep stages and clinical indicators. The findings hinted that HRV metrics in different sleep stages were more associated with T2DM clinical indicators. During each sleep stage, the correlation coefficients between HRV metrics and clinical indicators demonstrated a higher overall level compared with those during sleep period.

Compared with linear HRV metrics, non-linear HRV metrics significantly correlated with more clinical indicators, including LDL-C and UACR. Interestingly, non-linear HRV metrics in different sleep stages were significantly associated with glycemic control, while few weak correlations were found between HRV metrics of h and awake periods and glycemic control indicators.

Table 7 demonstrates correlation coefficients between non-linear HRV metrics in three sleep stages and glycemic control indicators including admission FBG, discharge FBG and HbA1c.

Most of non-linear HRV metrics were significantly correlated with admission FBG e. TABLE 7. Correlation coefficients between non-linear HRV metrics in three sleep stages and clinical glycemic control indicators including admission FBG, discharge FBG and HbA1c.

Table 8 demonstrates between-group differences in non-linear HRV metrics. TABLE 8. Data are presented as mean ± SD. We also applied partial correlation analysis between HbA1c, admission FBG and non-linear HRV metrics in sleep stages as controlling the effect of sleep quality metrics, in order to briefly discover the interaction among glycemic control, non-linear HRV analysis and sleep.

Table 9 demonstrates partial correlations between HbA1c and non-linear HRV metrics in unstable and REM sleep.

When nullifying the effect of SST, ALSS and SSP, the correlations between HbA1c and non-linear HRV metrics significantly strengthened e. However, the correlations between HbA1c and non-linear HRV metrics decreased and even were lost when corrected for TST e. The associations were mostly retained when corrected for other sleep metrics e.

TABLE 9. Partial correlation analysis between HbA1c and non-linear HRV metrics in unstable and REM sleep. Table 10 demonstrates partial correlations between admission FBG and non-linear HRV metrics in 3 sleep stages.

When nullifying the effect of SST, the correlations between HbA1c and non-linear HRV metrics at sleep stages significantly strengthened e. When nullifying the effect of TST, the correlations between HbA1c and non-linear HRV metrics decreased and even were lost e.

The associations between admission FBG and non-linear HRV metrics, especially DFA-α2, were retained when corrected for other sleep metrics e. TABLE Partial correlation analysis between admission FBG and non-linear HRV metrics in 3 sleep stages.

This study focused on the relationships among sleep, HRV derived ANS function and metabolic function in T2DM, and the main findings were: 1 metabolic function was significantly correlated with sleep quality; 2 HRV derived ANS function was different between sleep and awake, showing strengthened and distinct distributions of correlations with metabolic function compared to h analysis; 3 HRV derived ANS function and its correlation with clinical indicators of metabolic function altered during sleep cycles; 4 HRV analysis during sleep was strongly associated with clinical indicators of metabolic function, and this also applied to the glycemic control in patients with T2DM in non-linear analysis.

These findings demonstrated that sleep rhythm based HRV analysis should be emphasized in ANS and metabolic function assessment and the rationality of non-linear HRV analysis for further application among T2DM patients and investigation of potential interactions among ANS, sleep and metabolic function Lombardi, ; Perkiömäki, In sleep quality assessment, PSQI had few and weak correlations with clinical indicators, while TST and UST correlated with more clinical indicators.

The results suggest limitations of PSQI from its objectivity and indicate the important role of sleep in regulation of metabolic function.

Researchers have found that circadian rhythms and sleep regulate hormones and lipids involved in energy metabolism VanCauter et al.

The enhanced association between ANS and metabolic function during sleep is consistent with previously published results.

Researchers have found that most of the biological functions changes during sleep compared to wake, such as heart rate, arterial blood pressure and etc. Results in Table 5 , 6 showed that HRV metrics during sleep were associated with clinical indicators of metabolic function more significantly and broadly, compared to h HRV, and there were significant differences between HRV metrics during sleep and awake.

Thus, sleep might be a good model to investigate ANS activity and the fluctuation of ANS function caused by intrinsic factors, especially circadian rhythm, without influence factors of daytime activities such as eating behaviour, physical movement, etc.

Meanwhile, studies have shown that HRV metrics are significantly influenced by routine life states and ECG recording time morning, noon, evening , and the RR values are rather stable during night sleeping Cui et al.

Significant differences observed in many HRV metrics among different sleep stages suggest the dominance of SNS in unstable sleep and the dominance of PNS in stable sleep Pagani et al. Researchers have found that ANS tone and modulation are profoundly influenced by sleep related mechanisms Somers et al.

For ANS impose regulatory control over the cardiovascular system by SNS and PNS, which leads to HRV, the changes of HRV metrics in different sleep stages and their correlations with clinical indicators may associate more closely with the variability of ANS function during sleep cycles.

Researchers have conducted HRV analysis during sleep to confirm different ANS function patterns at sleep stages Scholz et al. Studies have shown that, since most of HRV metrics applied in this study were based on the assumption of stationary series Akselrod et al. We applied Augmented Dickey-Fuller test to all the HRV sequences for stationarity test, and found sequences during h, daytime and sleep were mostly stationary.

However, during sleep cycles, the sequences were mostly non-stationary. Hence, the distinct distribution pattern presented in analysis during sleep cycles may demonstrate the shift of sympathovagal balance toward sympathetic predominance since non-stationarities of the HRV sequences will highlight the shift Magagnin et al.

With enhanced and clarified interaction between ANS function and HRV analysis in each sleep stage, the underlying autonomic pathways Stein and Pu, of different HRV metrics may be better distinguished from the remarkably strengthened correlations between clinical indicator of metabolic function and HRV metrics with distinct distribution patterns compared to traditional analysis of sleep quality and HRV, which is crucial to reveal the interaction among sleep, ANS and HRV.

Compared to linear analysis, non-linear HRV metrics in different sleep stages were associated with more T2DM clinical indicators. According to the development of chaotic theory and non-linear dynamics, it is now generally recognized that heart beat intervals RR intervals are non-linear and non-stationary time series, caused by complex interactions between physiological systems Peng et al.

Traditional methods in time- and frequency-domain may not be able to detect subtle but important changes in RR series Costa et al. Currently, HRV frequency-domain metrics are extensively used to assess the function of ANS components, leading to controversial results Stein and Pu, ; Reyes del Paso et al.

In this study, most comprehensive and strong correlations between HRV derived ANS function and metabolic function with distinct distribution patterns were observed in non-linear analysis, demonstrating the rationality of non-linear analysis in assessing ANS function.

However, a part of non-linear HRV metrics we chose are somehow influenced by some biases. ApEn is a biased statistic, and the bias arises from two separate sources.

Second, the concavity of the logarithm implies a bias in all the regularity statistics Pincus and Huang, ; Delgado-Bonal and Marshak, On some occasions, ApEn failed to capture the sympathovagal balance Porta et al. The MSE method also has shortcomings.

The course of the entropy-based complexity as a function of the time scale is partially linked to the reduction of variance inherent to the procedure for the elimination of the fast temporal scales, which confines more and more patterns inside phase space cells of constant dimension Nikulin and Brismar, As a consequence, more and more patterns become indistinguishable, thus artificially reducing complexity with the scale factor τ.

Obviously, this effect is more evident at large time scales when variance is importantly reduced, thus producing biased MSE courses especially at large time scales.

Meanwhile, the procedure devised to eliminate fast time scales is suboptimal, thus producing uncontrolled effects on the assessment of complexity at any scale Valencia et al. It is likewise not negligible that the impact of the data length on the ability to predict the health status of ANS Laborde et al.

However, increased HRV may correspond to longer recordings instead of the change of ANS function, and it is inappropriate to compare metrics like SDNN when they are calculated from epochs of different length Saul et al.

Meanwhile, since mechanisms responsible for heart period modulations of a certain frequency, especially LF and HF power components, may not remain unchanged during long-term recording Furlan et al.

In this study, surrogate tests Porta et al. The percentages of non-linear dynamics were: 1 Therefore, non-linear dynamics in HRV sequences decreased during sleep, possibly due to the suppression of regulation of respiratory rhythm Mador and Tobin, ; González et al.

And when HRV sequences were segmented based on sleep stages, the percentage of non-linear dynamics elevated, suggesting better manifestation of the non-linear components in HRV sequences during sleep cycles and demonstrating the effectiveness of non-linear analysis.

Additionally, the distribution of non-linear features during different sleep stages, that the maximum of non-linear dynamics appeared during REM sleep, was consistent with non-linear respiratory dynamics Sako et al. However, researchers have found that in the analysis of short-term, stationary and successive HRV sequences, non-linear model-free and linear model-based conditional entropy correlate closely and have similar performance in the assessment of cardiac control Porta et al.

The conflicting conclusions may be a consequence of differences in stationary and non-linear features of HRV sequences utilized in analysis Magagnin et al. Previous study has shown that short-term HRV in healthy young adults at rest is mainly linear Porta et al.

Sleep rhythm based HRV analysis presents us significant correlations between ANS and metabolic function of T2DM patients covered by limitations of traditional measures, demonstrating its efficiency on distinguishing potential dynamic characteristics of ANS function and its correlation with metabolic function in different sleep stages.

Significant correlations were found among DBP and HRV metrics during sleep see Table 5 , and non-linear HRV metrics during sleep cycles significantly correlated with DBP with distinct distribution patterns see Table 6 , suggesting strong association between cardiac baroreflex control and the fluctuation of HRV metrics, especially non-linear ones, in T2DM patients.

These findings demonstrate the effectiveness of non-linear HRV analysis and promising strategies for the early detection of autonomic dysfunction.

ANS plays an important role in the exchange of metabolic information between organs and regulation on peripheral metabolism Yamada and Katagiri, In addition to affecting the activity of ANS, sleep also exerts a significant regulatory effect on glycemic control by influencing the balance and levels of hormones including leptin, growth hormone-releasing peptide, insulin and cortisol Balbo et al.

Moreover, non-linear HRV metrics during sleep were significantly associated with FBG level, which is one of the important indicators for glycemic compliance Association, The results indicates that non-linear HRV metrics during sleep may better manifest neuronal dysfunction caused by hyperglycemia, which induces oxidative stress and toxic glycosylation products Pop-Busui, , highlighting the importance of non-linear analysis during sleep.

Interestingly, higher HbA1c level was associated with higher entropy related metrics see Table 8 , the elevation of complexity may indicate worse metabolic function due to poor diabetic control Costa et al.

Additionally, in each sleep stage, the correlations between non-linear HRV metrics and FBG changed significantly before and after hospitalization, possibly due to the immediate impact of hospitalization on blood glucose control, reflecting the sensitivity of non-linear HRV assessment on glycemic control.

Due to the complex interaction among sleep, diabetic control and ANS function that sleep have a significant effect on both ANS and metabolic function, especially glycemic control of T2DM patients Koren et al. With the nullification of sleep quality metrics, most of the correlations retained, which supported the significant correlations acquired by simple correlation analysis.

Whereas the association significantly decreased or even vanished when TST was corrected, indicating the significant role of sleep quality in adjusting ANS and glycemic function of T2DM patients Jordan et al. Additionally, when stable sleep related metrics, especially SST, were nullified, the correlation between non-linear HRV metrics in sleep stages and glycemic control indicators significantly strengthened, demonstrating the functional significance of stable sleep stage.

As well as stable sleep stage slow wave sleep is widely recognized as the most restorative of all sleep stages Van Cauter et al. In our original study, we also used RR interval time series from four time periods late night, —; dawn, —; after lunch, —; after dinner, — to perform HRV analysis and correlation analysis with metabolic indicators see Supplementary Table S4—S7.

This may be due to the different physiological states of individuals at specific time periods and is influenced by meals and daytime activities. To the best of our knowledge, few studies have applied time-, frequency-domain and non-linear analysis on HRV sequences in each sleep stage to inquire the association among ANS function, sleep cycle and metabolic function.

We compared our results with those of using traditional methods in all-night sleep analysis and short-term linear HRV analysis. Representative results are shown in Table 11 in summary. In comparison with studies of correlations between all-night sleep and metabolic function, conducted among subjects without T2DM Koren et al.

In comparison with studies of correlations between short-term linear HRV analysis and metabolic function, conducted among T2DM patients Bhati et al. However, when compared with a study with larger sample size Balikai et al. Comparison of related studies on the interactions among sleep, HRV and metabolic function: subjects, methods, results and conclusion.

Therefore, this study provides new insights into research on the interaction among sleep, ANS and metabolic function, for T2DM patients, especially for diabetic control. The effective use of HRV analysis during sleep cycles to extract dynamic features of PNS and SNS in each sleep stage provides support for early detection of autonomic neuropathy.

In addition, the extensively strong correlations between non-linear HRV metrics and glycemic control indicators including FBG and HbA1c demonstrates the validity of HRV analysis in each sleep stage, which can be another promising strategy for diabetic control and early diagnose of T2DM with less computational costs.

For further research, this study proposed to select HRV sequences based on physiological rhythms can locally reveal the dynamic features of HRV and magnify the potential biological origins and kinetic mechanisms of HRV under the influence of many external factors, such as respiration, circadian rhythm, sympathetic and vagal activity, as well as intrinsic factors, such as ion channel fluctuations and other molecular fluctuations, while reducing the computational cost and elevating local resolution of long term signals, providing new insights into long term physiological signal processing.

There are several potential limitations of this study. Mainly middle-aged male inpatients were included in the analysis, who do not represent the whole spectrum of T2DM patients, to eliminate the influence of climacterium on metabolic function.

Although we did a further outlier rejection to reduce the influences of specific situations, there are still diversities between people with different physiological factors age, gender, health status, course of disease, etc. that were not included in this study, requiring further studies.

Despite of the non-linear dynamics in HRV sequences appeared in our study, researchers have found that short-term heart period variability in healthy young adults at rest is mainly linear Porta et al.

Lastly, due to the limited sample size, we did not build more accurate statistical models of HRV, ANS, sleep and metabolic function to further investigate their potential biological origins and kinetic mechanisms. The raw data supporting the conclusion of this article will be made available by the authors, without undue reservation.

WC, LT, and XC conceived and designed the study. WC and HC performed the experiment and analyzed the data. WC and HC drafted the manuscript. XC and ZM supervised the analysis, reviewed and editing the manuscript.

All authors contributed to the article and approved the submitted version. This work was supported in part by National Key Research and Development Program of China under Grant YFC, in part by the National Natural Science Foundation of China under Grant , The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers.

Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher. Akselrod, S. Power spectrum analysis of heart rate fluctuation: A quantitative probe of beat-to-beat cardiovascular control.

Science , — PubMed Abstract CrossRef Full Text Google Scholar. Association, A. Diagnosis and classification of diabetes mellitus. Diabetes Care 37 1 , S81—S Azami, H.

Matlab codes for "refined multiscale fuzzy entropy based on standard deviation for biomedical signal analysis. University of Edinburgh, School of Engineering, Institute for Digital Communications. Google Scholar.

Balbo, M. Impact of sleep and its disturbances on hypothalamo-pituitary-adrenal Axis activity. Balcıoğlu, A. Diabetes and cardiac autonomic neuropathy: Clinical manifestations, cardiovascular consequences, diagnosis and treatment. World J. Diabetes 6 1 , 80— Balikai, F. Correlation of serum HDL level with HRV indices using multiple linear regression analysis in patients with type 2 diabetes mellitus.

Diabetes Res. Barone, M. Diabetes and sleep: A complex cause-and-effect relationship. Bhati, P. Subclinical inflammation and endothelial dysfunction are linked to cardiac autonomic neuropathy in type 2 diabetes.

Bigger, J. Comparison of time- and frequency domain-based measures of cardiac parasympathetic activity in Holter recordings after myocardial infarction.

Brandenberger, G. Inverse coupling between ultradian oscillations in delta wave activity and heart rate variability during sleep. Buysse, D. The Pittsburgh sleep quality index: A new instrument for psychiatric practice and research.

Psychiatry Res. Cerutti, S. Long-term invariant parameters obtained from h holter recordings: A comparison between different analysis techniques.

Chaos An Interdiscip. Nonlinear Sci. Chen, W. Measuring complexity using FuzzyEn, ApEn, and SampEn. Cooke, W. Human responses to upright tilt: A window on central autonomic integration. Physiology 2 , — Costa, M. Multiscale entropy analysis of biological signals.

E 71 2 , Multiscale entropy analysis of complex physiologic time series. Cui, X. On the variability of heart rate variability—evidence from prospective study of healthy young college students.

Entropy 22 11 , Cuschieri, S. The STROBE guidelines. Saudi J. Cysarz, D. Entropies of short binary sequences in heart period dynamics.

Physiology-Heart Circulatory Physiology 6 , H—H de Moura-Tonello, S. Cardiovascular variability analysis and baroreflex estimation in patients with type 2 diabetes in absence of any manifest neuropathy. PLOS ONE 11 3 , e de Zambotti, M.

Dynamic coupling between the central and autonomic nervous systems during sleep: A review. Delgado-Bonal, A. Approximate entropy and sample entropy: A comprehensive tutorial.

It responds not only to a poor night of sleep, or that sour interaction with your boss, but also to the exciting news that you got engaged, or to that delicious healthy meal you had for lunch. Our body handles all kinds of stimuli and life goes on. However, if we have persistent instigators such as stress, poor sleep, unhealthy diet, dysfunctional relationships, isolation or solitude, and lack of exercise, this balance may be disrupted, and your fight-or-flight response can shift into overdrive.

The gold standard is to analyze a long strip of an electrocardiogram done in the doctor's office. But in recent years, companies have launched apps and wearable heart rate monitors that do something similar.

The accuracy of these methods is still under scrutiny, but the technology is improving. If you do wish to give it a try, chest strap monitors tend to provide a more accurate measure of HRV than wrist devices.

HRV may offer a noninvasive way to signal imbalances in the autonomic nervous system. Based on data gathered from many people, if the system is in more of a fight-or-flight mode, the variation between subsequent heartbeats tends to be lower.

If the system is in more relaxed state, the variation between beats may be higher. This suggests some interesting possibilities. People who have a high HRV may have greater cardiovascular fitness and may be more resilient to stress. HRV may also provide personal feedback about your lifestyle and help motivate those who are considering taking steps toward a healthier life.

You might see a connection to HRV changes as you incorporate more mindfulness, meditation, sleep, and especially physical activity into your life. For those who love data and numbers, this could be a way to track how your nervous system is reacting not only to the environment, but also to your emotions, thoughts, and feelings.

There are questions about the accuracy, reliability and overall usefulness of tracking HRV. While HRV has been linked to overall physical fitness, the correlation between changes in HRV and how your autonomic nervous system is functioning will require much more research.

Still, if you decide to use HRV as another piece of health data, do not get too confident if you have a high HRV, or too worried if your HRV is low.

Think of HRV as another way you might tap into your body and mind are responding to what your daily experiences. As a service to our readers, Harvard Health Publishing provides access to our library of archived content. Please note the date of last review or update on all articles.

No content on this site, regardless of date, should ever be used as a substitute for direct medical advice from your doctor or other qualified clinician. With this Special Health Report, Living Better, Living Longer , you will learn the protective steps doctors recommend for keeping your mind and body fit for an active and rewarding life.

Thanks for visiting. Don't miss your FREE gift. The Best Diets for Cognitive Fitness , is yours absolutely FREE when you sign up to receive Health Alerts from Harvard Medical School.

Sign up to get tips for living a healthy lifestyle, with ways to fight inflammation and improve cognitive health , plus the latest advances in preventative medicine, diet and exercise , pain relief, blood pressure and cholesterol management, and more.

Get helpful tips and guidance for everything from fighting inflammation to finding the best diets for weight loss from exercises to build a stronger core to advice on treating cataracts.

PLUS, the latest news on medical advances and breakthroughs from Harvard Medical School experts. Sign up now and get a FREE copy of the Best Diets for Cognitive Fitness. Stay on top of latest health news from Harvard Medical School.

Recent Blog Articles. Flowers, chocolates, organ donation — are you in? What is a tongue-tie? What parents need to know. Which migraine medications are most helpful? More studies are needed to further validate our findings. Cardiac autonomic dysfunction due to hyperglycemia has been suggested as a mechanism of cardiovascular complications in type 2 diabetes mellitus T2D 1.

Major organs responsible for insulin secretion and sensitivity, glucose production, and metabolism, including the pancreas, liver, and skeletal muscle, are innervated by the autonomic nervous system 2.

Yet, there are numerous pathways whereby autonomic dysfunction could, in turn, affect glucose metabolism. Notably, an autonomic imbalance was found to be already present in persons with prediabetes 3 and was associated with incident diabetes 4.

Autonomic dysfunction has also been related to reduced insulin sensitivity and beta-cell function in people without diabetes 5 , 6. Heart rate variability HRV is a marker of cardiac autonomic dysfunction, and a single assessment of HRV has been associated with changes in fasting glucose level 7 and incident T2D 8.

Taken together, alterations in autonomic function may contribute to the pathogenesis of T2D. However, HRV has a strong and inverse relationship with heart rate, so HRV parameters should be corrected for heart rate during analysis. Therefore, the results from previous studies using uncorrected HRV might be confounded.

In addition, given the considerable impact of age on HRV 9 and the possible bidirectional association between hyperglycemia and autonomic dysfunction 3 , 4 , studies using only single HRV measurements, cross-sectional designs, and short follow-up periods are all prone to confounding and reverse causation.

Joint modeling is a novel method that can perform simultaneous analyses of repeated exposure measurements and survival data, and its principal advantage is the proper treatment of noisy and incompletely observed time-varying exposure information.

Thus, this approach is appropriate to estimate the hazard of incident T2D for the HRV metrics as time-varying covariates, which enables unbiased estimation of the relationship between the exposure and the outcome.

In the large prospective population-based Rotterdam Study with repeated measurements of HRV, we investigated the prospective association of evolution of HRV, as a proxy for autonomic function, with the incidence of T2D.

In addition, we conducted a bidirectional Mendelian randomization MR analysis using summary-level data to explore the causality of the association between HRV and T2D.

This study was embedded within the Rotterdam Study RS , a prospective cohort study of community-dwelling persons in Ommoord, Rotterdam, The Netherlands.

The detailed study design has been described elsewhere There were no eligibility criteria to enter the Rotterdam Study apart from the minimum age and residential area based on postal codes. Participants attended follow-up examinations every 3 to 5 years. The Rotterdam Study has been approved by the Medical Ethics Committee of the Erasmus MC registration number MEC The Rotterdam Study Personal Registration Data collection is filed with the Erasmus MC Data Protection Officer under registration number EMC nl and into the WHO International Clinical Trials Registry Platform ICTRP; www.

All participants provided written informed consent to participate in the study and to have their information obtained from their treating physicians. The current study was based on the third visit of RS-I RS-I-3 and the first visit of RS-II RS-II-1 and RS-III RS-III Therefore, participants were included in the study Supplementary Figure S1 A standard s, lead resting ECG was recorded during each follow-up examination with an ACTA Gnosis electrocardiograph Esaote Biomedica, Italy at a sampling frequency of Hz and stored digitally.

The ECGs were processed by the Modular ECG Analysis System MEANS , an ECG computer program that has been validated extensively 12 , The HRV was calculated based on RR intervals between normal heartbeats; RR intervals were excluded if they immediately preceded or followed premature atrial complexes or premature ventricular complexes.

The following HRV indices were used for the analyses: the heart rate—corrected standard deviation of the normal-to-normal RR intervals SDNNc and the heart rate—corrected root mean square of successive RR-interval differences RMSSDc 9.

Information on covariates was collected at baseline using a structured questionnaire. Fasting blood glucose and insulin levels, and total and high-density lipoprotein cholesterol were measured using standard laboratory techniques. Homeostatic model assessment of insulin resistance HOMA-IR and beta-cell function HOMA-β were calculated based on fasting blood glucose and serum insulin concentration to assess insulin resistance and β-cell function separately.

Smoking status was categorized into never, former, and current smoking. Blood pressure was measured in the right upper arm with the participant in a sitting position, of which the mean of 2 consecutive measurements was used.

Physical activity levels were assessed using validated questionnaires the Zutphen Physical Activity Questionnaire for RS-I and RS-II 14 , the LASA Physical Activity Questionnaire for RS-III 15 and further quantified into metabolic equivalent task MET values per week doing moderate and vigorous-intensity activities classified according to the Dutch Physical Activity Guideline Medication use blood pressure— and lipid-lowering drugs was derived from baseline questionnaires and pharmacy data and was categorized and defined according to the World Health Organization Anatomical Therapeutic Chemical WHO ATC classifications.

Specifically, antihypertensive medication, use of beta blockers, use of calcium blockers, and lipid-lowering medication were defined according to the WHO ATC categories c02, c07, c08, and c10, respectively.

In addition, information about prevalent cardiovascular disease including coronary heart disease, heart failure, and stroke was also collected at baseline. Participants were followed up from the date of attending the baseline visit onward. At baseline and during follow-up, cases of T2D were ascertained by the use of general practitioners records, hospital discharge letters, and serum glucose measurements collected from center visits, which took place roughly every 4 years.

T2D was defined as a fasting blood glucose concentration equal to or above 7. Information about blood glucose-lowering medications was obtained from both structured home interviews and pharmacy dispensing records.

Two study physicians independently adjudicated all potential events of T2D. In the case of disagreement, a consensus was sought from a diabetologist. Participants were followed until incident T2D, death, or the end of the study period January 1, Descriptive statistics were performed by reporting mean SD or median interquartile range, IQR for continuous variables and number percentages for categorical variables.

Two HRV metrics SDNNc and RMSSDc were log-transformed to fulfill the normality assumption. Heart rate and log-transformed HRV metrics were further standardized to allow for direct comparisons of effect sizes, per 1-SD increase.

Linearity was explored with restricted cubic splines for each exposure, with no evidence of deviation from linearity P for nonlinear:. For the longitudinal analysis, joint models for longitudinal and time-to-event data were performed The joint model estimates continuous profiles of each HRV metric based on the repeated measured data collected during the whole follow-up period for each individual; therefore, it would consider individual variations and reduce the bias associated with missing data.

In addition, joint models are more appropriate for estimating the hazard of incident T2D for the HRV metrics as time-varying covariates because they account for their endogenous nature.

For the HRV metrics, we used linear mixed-effect models. When appropriate and judged by residuals plots, transformed HRV metrics were used as dependent variables. We included age the time scale variable and sex in the fixed-effects, with both the intercept and the slope fitted as random effects.

Next, a joint model was implemented by combining the joint distribution of HRV metrics in the linear mixed-effects model with the Cox model. For the crude model, we included baseline age, sex, and cohort in the survival part of the models. The full model was fitted by further adjusting for BMI, smoking status, systolic blood pressure, total and high-density lipoprotein cholesterol, use of blood pressure—lowering or lipid-lowering medication, and prevalent cardiovascular disease.

We also used Spearmen correlation to examine the cross-sectional associations between heart rate and different HRV metrics and glycemic traits fasting blood glucose and insulin levels, HOMA-IR, and HOMA-β at baseline.

To check for any possible effect modification by age, sex, BMI, or use of blood pressure—lowering medication, we separately added an interaction term between each variable; age [continuous], sex [dichotomous], BMI [continuous], use of blood pressure—lowering medication [dichotomous] , and HRV metrics to the joint model and then further explored these by stratification.

To ensure a sufficient sample size for subgroup analyses, age stratification was based on the median age 62 years , and BMI stratification was based on the cutoff point for overweight Additionally, we conducted two-sample bidirectional MR analyses to examine the association between heart rate—uncorrected HRV SDNN and RMSSD and T2D.

The inverse variance weighted IVW method was the main method used in our analyses. More details on the rationale, assumptions, and sensitivity analyses of the MR analyses are shown in the Supplementary Methods S1 Information on covariables was missing for up to 2.

To deal with missing values, we used single imputation with the expectation-maximization method. Data were handled and analyzed with SPSS Statistics version Among included participants, the median age was During a median follow-up time of 8.

Values are mean SD or median interquartile range for continuous variables and number percentages for categorical variables. Abbreviations: HOMA-β, homeostatic model assessment of beta-cell function; HOMA-IR, homeostatic model assessment of insulin resistance; RMSSDc, heart rate—corrected root mean square of successive RR-interval differences; SDNNc, heart rate—corrected standard deviation of normal-to-normal RR intervals.

In the joint model analysis, heart rate and HRV metrics were positively associated with incident T2D Table 2. For heart rate, a 1-SD increment was associated with the risk of developing T2D in the crude model hazard ratio [HR], 1.

After adjustments, the association remained significant HR 1. For HRV, both metrics showed positive associations with T2D development with statistically significant associations only found for RMSSDc. The association of SDNNc with incident T2D was not statistically significant HR 1.

Joint model results for the association between longitudinal measures of heart rate and heart rate variability with incident type 2 diabetes.

Model 1 was adjusted for baseline age, sex, and cohort for relative risk model. Model 2 was further adjusted for body mass index, smoking status, systolic blood pressure, use of blood pressure—lowering medications, total cholesterol, high-density lipoprotein, use of lipid-lowering medications, and history of cardiovascular disease coronary heart disease, heart failure, and stroke at baseline.

The hazard ratio for incident diabetes was calculated per 1-SD increase in heart rate or the log of HRV indices SDNNc and RMSSDc. Abbreviations: HR, hazard ratio; RMSSDc, heart rate—corrected root mean square of successive RR-interval differences; SDNNc, heart rate—corrected standard deviation of normal-to-normal RR intervals.

We stratified participants based on the median age 62 years and found that the association between heart rate and incident T2D was relatively stronger among younger participants Fig.

Although significant associations were restricted to men Fig. We also did not find a significant interaction for BMI or use of blood pressure—lowering medication Supplementary Table S1 11 , although statistically significant associations were restricted to participants who were overweight or without use of blood pressure—lowering drugs, respectively.

Forest plot summarizing the joint model results for the associations of heart rate and heart rate variability with incident type 2 diabetes. Plot is based on the results of the model adjusted for baseline age, sex, cohort, body mass index, smoking status, systolic blood pressure, use of blood pressure—lowering medications, total cholesterol, high-density lipoprotein, use of lipid-lowering medications, and history of cardiovascular disease coronary heart disease, heart failure, and stroke at baseline for relative risk model.

The hazard ratio for incident diabetes was calculated per 1-SD increase in heart rate or in the log of HRV indices SDNNc and RMSSDc. Abbreviations: RMSSDc, heart rate—corrected root mean square of successive RR-interval differences; SDNNc, heart rate—corrected standard deviation of normal-to-normal RR intervals.

In further analyses with additional adjustments for baseline measurement of fasting blood glucose and physical activity and in a complete case analysis, results remained consistent with our main results Supplementary Table S2 The Spearmen correlation analyses indicated that heart rate was significantly associated with all glycemic traits, including fasting blood glucose, insulin, HOMA-IR, and HOMA-β, while RMSSDc was only related to fasting blood glucose Fig.

After excluding individuals with baseline prediabetes, the associations between HRV metrics and glycemic traits were not statistically significant Supplementary Figure S2 Correlation plot between heart rate and heart rate variability and glycemic traits.

Abbreviations: FBG, fasting blood glucose; HOMA-β, homeostatic model assessment of beta-cell function; HOMA-IR, homeostatic model assessment of insulin resistance; RMSSDc, heart rate—corrected root mean square of successive RR-interval differences; SDNNc, heart rate—corrected standard deviation of normal-to-normal RR intervals.

A total of 9 single nucleotide polymorphisms SNPs for SDNN and 9 SNPs for RMSSD were available in the T2D genome-wide association study GWAS and were used for the MR analyses after removal of potential outliers Supplementary Table S3 In addition, SNPs for T2D were available in the HRV GWAS Supplementary Table S5 11 , and the results from the MR analyses showed that genetically predicted T2D was not significantly associated with log SDNN or log RMSSD Supplementary Table S6 The WME and MR-Egger slope estimates were also insignificant, consistent with the inverse variance weighted method after correcting for outliers using MR-PRESSO during the bidirectional MR analysis, and we found no evidence for violation of the MR assumptions.

In this large prospective population-based cohort study, longitudinal evolutions of both heart rate and different HRV metrics were significantly associated with new-onset T2D, independent of a vast number of other contributing factors.

However, the effects were largely restricted to younger individuals. MR analyses suggested no causal association between HRV and incident T2D. At first sight, our findings suggest an association between elevated heart rate and increased risk of developing T2D, consistent with previous studies 18 , However, the effects were restricted to younger individuals.

As a surrogate marker for autonomic activity, a high heart rate usually indicates increased sympathetic activity, potentially inducing insulin resistance. On the one hand, a more straightforward relationship between a fast heart rate and sympathetic predominance at a young age may explain the relatively strong association among the young participants 18 , On the other hand, older participants tend to have worse health status and use more medications such as beta blockers that reduce heart rate 20 and hyperglycemia Hence, the associations might be diluted at old age.

Although diabetes is the leading cause of primary autonomic dysfunction, limited evidence exists regarding the relationship between autonomic dysfunction and incident diabetes Prior studies assessing the risk of developing T2D associated with HRV have mostly shown an association of autonomic dysfunction with incident T2D.

However, the direction between various HRV metrics and glycemic traits is still inconsistent , 8. For example, the Atherosclerosis Risk In Communities ARIC study found no significant association between SDNN and incident T2D 8 , while the Kangbuk Samsung Health Cohort reported that as SDNN and RMSSD tertiles increased, the risk of diabetes decreased 4.

Our findings support an association between increased SDNNc and RMSSDc with incident T2D. Notably, these 2 heart rate—corrected HRV metrics have not been studied before concerning incident T2D, which limits the comparability with former studies.

The participants in our study were also considerably older than Kangbuk Samsung Health study 4. Age has an impact on both SDNNc and RMSSDc. The upper limit of their normal values decreases until the age of 60 and increases markedly afterward 9.

Besides the autonomic nervous system dysfunction, increased HRV may also be affected by the sinus node dysfunction With growing age, pathologic changes occur in the sinoatrial node, including increasing collagen and elastic fibers Intrinsic sinus node function tends to deteriorate with age, resulting in prolonged RR intervals and increased, irregular HRV 25 , which was also found in our study Therefore, the association between increased HRV and incident T2D could, at least partly, be explained by sinus node dysfunction.

Unlike results from our longitudinal analyses, only heart rate and not HRV metrics was associated with different glycemic traits at the baseline of our study. The HRV effect could be difficult to observe due to compensatory mechanisms preserving glucose homeostasis among a substantial number of nondiabetic participants.

In line, we found that the effect of HRV disappeared after excluding persons with prediabetes. A prior study also reported that only heart rate, not HRV, is associated with changes in insulin sensitivity.

This could imply that pathways other than autonomic dysfunction mediate the associations with diabetes or that heart rate is just a marker of other mechanisms 5. These results regarding the correlations with glycemic traits should, however, be interpreted with caution since they were based on cross-sectional analyses and cannot address the temporal relationship of heart rate and HRV with glycemic traits.

More studies are needed to further delineate the underlying mechanisms. However, inconsistent with the longitudinal findings, our bidirectional MR analysis showed no causal association between HRV and T2D. This may be due to limited power since only a few instrumental variables for SDNN and RMSSD were available to be used for the MR analyses.

Furthermore, unlike longitudinal analysis using these novel heart rate—corrected HRV parameters SDNNc and RMSSDc , the MR analysis could only use heart rate—uncorrected HRV SDNN and RMSSD due to the lack of available SNPs, which may partly explain the heterogeneity we observed.

A previous study reported substantial overlap of loci between HRV and heart rate, with SNPs in 5 of the 21 heart rate loci being associated with HRV at genome-wide significance level and 6 more attaining nominal significance This suggests that part of the HRV SNPs exert their effect on heart rate through oscillatory modulation of pacemaker activity by the vagal nerves.

Therefore, the insignificant association between RMSSD and T2D in our MR analysis might be biased by heart rate. Future GWAS with a larger sample size and individual-level data could identify more genetic variants that could be used to assess the association between the heart rate—corrected HRV and T2D.

The strengths of this population-based study include the prospective cohort design, long follow-up time, and meticulous assessment of incident T2D.

We also had detailed information regarding possible confounders. Another strength is using joint models, which enables the analysis of individual heart rate and HRV values, including those with missing data.

It generates the most likely continuous exposure profile for each individual while simultaneously accounting for exposure and survival processes. Also, we are the first study to report the health effect of heart rate—corrected HRV metrics, which are more appropriate to allow meaningful comparison of different HRV measurements and their association with adverse outcomes.

However, our study mainly included older individuals of European ancestry, limiting our findings generalizability to younger populations and other ethnicities. In addition, although the moderately nonlinear change of HRV was reported by former studies, we found no evidence of deviation from linearity, which might be due to the different outcomes we used.

The additional MR analysis we used also assumes linearity. Given that the more novel MR approaches can check the potential nonlinear association between exposure and outcomes using individual-level data, future studies with more detailed data and using comprehensive methods are needed to validate our findings.

Our results suggest that high heart rate and HRV were significantly associated with an increased risk of developing T2D, especially among younger individuals.

To our knowledge, this is the only prospective investigation using repeated measurements of heart rate and HRV to investigate the role of autonomic dysfunction in the development of T2D.

More studies are needed to validate our findings and to elucidate further the underlying mechanisms. The authors are grateful for the dedication, commitment, and contribution of the study participants and the general practitioners, pharmacists, and the staff from the Rotterdam Study. Furthermore, the authors would like to thank the Genetic Variance in Heart Rate Variability VgHRV , DIAbetes Genetics Replication And Meta-analysis DIAGRAM , Genetic Epidemiology Research on Adult Health and Aging GERA , UK Biobank UKB , and individual studies for sharing their summary statistics in GWAS.

The Rotterdam Study is funded by Erasmus MC and Erasmus University Rotterdam; Netherlands Organization for Scientific Research; Netherlands Organization for Health Research and Development ZonMw ; Research Institute for Diseases in the Elderly; Netherlands Genomics Initiative; Netherlands Ministry of Education, Culture and Science; Netherlands Ministry of Health, Welfare and Sports; European Commission; and Municipality of Rotterdam.

We would like to thank the China Scholarship Council for the scholarship to K. is responsible for the study concept and design; K. and S. composed the statistical dataset and performed the statistical analyses; K.

wrote the manuscript; F. Data generated by the authors or analyzed during the study are available upon request. Requests should be directed toward the management team of the Rotterdam Study secretariat. epi erasmusmc. nl , which has a protocol for approving data requests.

Because of restrictions based on privacy regulations and informed consent of the participants, data cannot be made freely available in a public repository. Sorensen BM , Houben AJ , Berendschot TT , et al. Prediabetes and type 2 diabetes are associated with generalized microvascular dysfunction: the Maastricht study.

Google Scholar. Faber CL , Deem JD , Campos CA , Taborsky GJ , Morton GJ. CNS control of the endocrine pancreas. Coopmans C , Zhou TL , Henry RMA , et al. Both prediabetes and type 2 diabetes are associated with lower heart rate variability: the Maastricht study.

The STROBE guidelines. Saudi J. Cysarz, D. Entropies of short binary sequences in heart period dynamics. Physiology-Heart Circulatory Physiology 6 , H—H de Moura-Tonello, S. Cardiovascular variability analysis and baroreflex estimation in patients with type 2 diabetes in absence of any manifest neuropathy.

PLOS ONE 11 3 , e de Zambotti, M. Dynamic coupling between the central and autonomic nervous systems during sleep: A review. Delgado-Bonal, A. Approximate entropy and sample entropy: A comprehensive tutorial.

Entropy 21 6 , Dohare, A. An efficient new method for the detection of QRS in electrocardiogram. CrossRef Full Text Google Scholar. Faust, O. Linear and non-linear analysis of cardiac health in diabetic subjects. Signal Process. Control 7 3 , — Fei, L.

Short- and long-term assessment of heart rate variability for risk stratification after acute myocardial infarction. Feng, N.

The associations between sleep architecture and metabolic parameters in patients with obstructive sleep apnea: A hospital-based cohort study.

Neurology 12, Francis, D. Physiological basis of fractal complexity properties of heart rate variability in man. Frattola, A. Time and frequency domain estimates of spontaneous baroreflex sensitivity provide early detection of autonomic dysfunction in diabetes mellitus.

Diabetologia 40 12 , — Furlan, R. Continuous hour assessment of the neural regulation of systemic arterial pressure and RR variabilities in ambulant subjects.

Circulation 81 2 , — Garber, A. Consensus statement by the American association of clinical endocrinologists and American college of endocrinology on the comprehensive type 2 diabetes management algorithm — executive summary.

Gerritsen, J. Impaired autonomic function is associated with increased mortality, especially in subjects with diabetes, hypertension, or a history of cardiovascular disease: The hoorn study.

Diabetes Care 24 10 , — Glos, M. Cardiac autonomic modulation and sleepiness: Physiological consequences of sleep deprivation due to 40h of prolonged wakefulness. Physiology Behav. Goit, R. Cardiovascular autonomic function and vibration perception threshold in type 2 diabetes mellitus.

Diabetes its Complicat. González, J. Detection and sources of nonlinearity in the variability of cardiac R-R intervals and blood pressure in rats. Hamilton, P. Ieee Trans. Hur, M. Deterioration of sleep quality according to glycemic status.

Inzucchi, S. Management of hyperglycemia in type 2 diabetes, A patient-centered approach: Update to a position statement of the American diabetes association and the European association for the study of diabetes.

Diabetes Care 38 1 , — Jo, J. Determinants of heart rate variability in obstructive sleep apnea syndrome during wakefulness and sleep. Physiology-Heart Circulatory Physiology 3 , H—H Johnson, A. Multimodal heart beat detection using signal quality indices. Jordan, A. Adult obstructive sleep apnoea.

Lancet , — Kent, B. Diabetes mellitus prevalence and control in sleep-disordered breathing: The European sleep apnea cohort ESADA study.

Chest 4 , — Kleiger, R. Heart rate variability: Measurement and clinical utility. Noninvasive Electrocardiol. Koren, D. Sleep architecture and glucose and insulin homeostasis in obese adolescents. Diabetes Care 34 11 , — Laborde, S. Heart rate variability and cardiac vagal tone in psychophysiological research — recommendations for experiment planning, data analysis, and data reporting.

Leproult, R. Sleep loss results in an elevation of cortisol levels the next evening. Sleep 20 10 , — Li, F. A real-time QRS complex detection algorithm based on composite filter and curve feature point extraction.

Liu, S. Circadian rhythm dynamics on multiscale entropy identifies autonomic dysfunction associated with risk of ventricular arrhythmias and near syncope in chronic kidney disease.

Lombardi, F. Origin of heart rate variability and turbulence: An appraisal of autonomic modulation of cardiovascular function. Physiology 2, López-Cano, C. Sympathetic hyperactivity and sleep disorders in individuals with type 2 diabetes.

Mador, M. Effect of alterations in mental activity on the breathing pattern in healthy subjects. Magagnin, V. Non-stationarities significantly distort short-term spectral, symbolic and entropy heart rate variability indices.

Malliani, A. Cardiovascular neural regulation explored in the frequency domain. Circulation 84 2 , — Marcheva, B. Disruption of the clock components CLOCK and BMAL1 leads to hypoinsulinaemia and diabetes. Nature , — Markwald, R. Impact of insufficient sleep on total daily energy expenditure, food intake, and weight gain.

Editors P. Shiromani, T. Horvath, S. Redline, and E. Van Cauter New York, NY: Springer New York , — Martelli, D. The low frequency power of heart rate variability is neither a measure of cardiac sympathetic tone nor of baroreflex sensitivity.

Physiology-Heart Circulatory Physiology 7 , H—H Martyn-Nemeth, P. Poor sleep quality is associated with nocturnal glycaemic variability and fear of hypoglycaemia in adults with type 1 diabetes. Maser, R. The association between cardiovascular autonomic neuropathy and mortality in individuals with diabetes: A meta-analysis.

Diabetes Care 26 6 , — Montano, N. Heart rate variability explored in the frequency domain: A tool to investigate the link between heart and behavior. Mosenzon, O.

Treatment of type 2 diabetes: From "guidelines" to "position statements" and back: Recommendations of the Israel national diabetes council. Diabetes Care 39 2 , S—S National Center for Cardiovascular Diseases, Chinese medical doctor association, hypertension committee of Chinese medical doctor association, Chinese society of Cardiology, Chinese medical association,, and hypertension committee of cross-straits medical exchanges association clinical practice guidelines for the management of hypertension in China.

Nguyen, J. Influence of weeks of circadian misalignment on leptin levels. Nikulin, V. Comment on "Multiscale entropy analysis of complex physiologic time series". author reply Nolan, J. Prospective study of heart rate variability and mortality in chronic heart failure: Results of the United Kingdom heart failure evaluation and assessment of risk trial UK-heart.

Circulation 98 15 , — Nunan, D. A quantitative systematic review of normal values for short-term heart rate variability in healthy adults. Pacing Clin. Pagani, M. Power spectral analysis of heart rate and arterial pressure variabilities as a marker of sympatho-vagal interaction in man and conscious dog.

Circulation Res. Pan, J. A real-time QRS detection algorithm. Peng, C. Mosaic organization of DNA nucleotides. E 49 2 , — Quantification of scaling exponents and crossover phenomena in nonstationary heartbeat time series. Perkiömäki, J. Heart rate variability and non-linear dynamics in risk stratification.

Pfeifer, M. Quantitative evaluation of cardiac parasympathetic activity in normal and diabetic man. Diabetes 31 4 , — Pincus, S. Approximate entropy as a measure of system complexity.

Heart rate control in normal and aborted-SIDS infants. Physiology-Regulatory, Integr. Physiology 3 , R—R Approximate entropy: Statistical properties and applications.

Statistics - Theory Methods 21 11 , — Pop-Busui, R. Cardiac autonomic neuropathy in diabetes: A clinical perspective. Diabetes Care 33 2 , — Porta, A. Prediction of short cardiovascular variability signals based on conditional distribution.

IEEE Trans. Progressive decrease of heart period variability entropy-based complexity during graded head-up tilt. Physiology 4 , — Complexity and nonlinearity in short-term heart period variability: Comparison of methods based on local nonlinear prediction.

Entropy, entropy rate, and pattern classification as tools to typify complexity in short heart period variability series. Information domain analysis of cardiovascular variability signals: Evaluation of regularity, synchronisation and co-ordination. Are nonlinear model-free conditional entropy approaches for the assessment of cardiac control complexity superior to the linear model-based one?

Punjabi, N. Disorders of glucose metabolism in sleep apnea. Physiology 99 5 , — Reyes del Paso, G. The utility of low frequency heart rate variability as an index of sympathetic cardiac tone: A review with emphasis on a reanalysis of previous studies.

Psychophysiology 50 5 , — Richman, J. Physiological time-series analysis using approximate entropy and sample entropy.

Ruiz, J. Diabetic neuropathy is a more important determinant of baroreflex sensitivity than carotid elasticity in type 2 diabetes. Hypertension 46 1 , — Sako, T. Nonlinear behavior of human respiratory movement during different sleep stages. Chronobiology Int.

Sassi, R. Advances in heart rate variability signal analysis: Joint position statement by the e-cardiology ESC working group and the European heart rhythm association co-endorsed by the Asia pacific heart rhythm society.

EP Eur. Saul, J. PubMed Abstract Google Scholar. Scheer, F. Adverse metabolic and cardiovascular consequences of circadian misalignment. Scholz, U. Vegetative background of sleep: Spectral analysis of the heart rate variability. Schramm, P. Cardio-pulmonary coupling CPC measures correlate to standard sleep variables in a random clinical sample of patients suspected with sleep disordered breathing SDB.

Sleep 32, A Shaffer, F. A healthy heart is not a metronome: An integrative review of the heart's anatomy and heart rate variability. Silvani, A. Central control of cardiovascular function during sleep. Physiology-Heart Circulatory Physiology 12 , H—H Singh, J.

Association of hyperglycemia with reduced heart rate variability The Framingham Heart Study. Solnik, S. Teager-Kaiser Operator improves the accuracy of EMG onset detection independent of signal-to-noise ratio.

Acta Bioeng. Somers, V. Sympathetic-nerve activity during sleep in normal subjects. Spiegel, K. Sleep loss: A novel risk factor for insulin resistance and type 2 diabetes.

Impact of sleep debt on metabolic and endocrine function. Twenty-four-Hour profiles of acylated and total ghrelin: Relationship with glucose levels and impact of time of day and sleep.

Metabolism 96 2 , — Spira, A. Reliability and validity of the Pittsburgh sleep quality index and the epworth sleepiness scale in older men. Journals Gerontology Ser. A 67A 4 , — Stein, P. Heart rate variability, sleep and sleep disorders.

Tank, J. Spontaneous baroreflex sensitivity and heart rate variability are not superior to classic autonomic testing in older patients with type 2 diabetes. Tasali, E. Slow-wave sleep and the risk of type 2 diabetes in humans. Task Force of the European Society of Cardiology the North American Society of Pacing Electrophysiology Heart rate variability.

Circulation 93 5 , — Thomas, R. An electrocardiogram-based technique to assess cardiopulmonary coupling during sleep. Sleep 28 9 , — Impaired sleep quality in fibromyalgia: Detection and quantification with ECG-based cardiopulmonary coupling spectrograms.

Tobaldini, E. Sleep, sleep deprivation, autonomic nervous system and cardiovascular diseases. Effects of acute and chronic sleep deprivation on cardiovascular regulation. Trinder, J. Autonomic activity during human sleep as a function of time and sleep stage.

Sleep Res. Turek, F. Obesity and metabolic syndrome in circadian clock mutant mice. Valencia, J. Refined multiscale entropy: Application to h holter recordings of heart period variability in healthy and aortic stenosis subjects.

Van Cauter, E. Metabolic consequences of sleep and sleep loss. VanCauter, E. Roles of circadian rhythmicity and sleep in human glucose regulation. Vinik, A. Diabetic autonomic neuropathy. Diabetes Care 26 5 , — Diabetic cardiovascular autonomic neuropathy. Circulation 3 , — Voss, A.

The application of methods of non-linear dynamics for the improved and predictive recognition of patients threatened by sudden cardiac death.

Cardiovasc Res. Wessel, N. Short-term forecasting of life-threatening cardiac arrhythmias based on symbolic dynamics and finite-time growth rates. E 61 1 , — Yamada, T. Ye, S.

The impact of the HbA1c level of type 2 diabetics on the structure of haemoglobin. Yeh, G. Enhancement of sleep stability with Tai Chi exercise in chronic heart failure: Preliminary findings using an ECG-based spectrogram method. Yu, Y. Impact of blood glucose control on sympathetic and vagus nerve functional status in patients with type 2 diabetes mellitus.

Acta Diabetol. Zhang, L. Sleep heart rate variability assists the automatic prediction of long-term cardiovascular outcomes. Ziegler, D. Diabetes Care 31 3 , — Zoccoli, G. The cerebral circulation during sleep: Regulation mechanisms and functional implications. Keywords: T2DM, heart rate variability HRV , sleep, autonomic nervous system ANS , metabolic function, nonlinear analysis, glycemic control.

Citation: Cheng W, Chen H, Tian L, Ma Z and Cui X Heart rate variability in different sleep stages is associated with metabolic function and glycemic control in type 2 diabetes mellitus. doi: Received: 02 February ; Accepted: 24 March ; Published: 14 April Copyright © Cheng, Chen, Tian, Ma and Cui.

This is an open-access article distributed under the terms of the Creative Commons Attribution License CC BY. The use, distribution or reproduction in other forums is permitted, provided the original author s and the copyright owner s are credited and that the original publication in this journal is cited, in accordance with accepted academic practice.

No use, distribution or reproduction is permitted which does not comply with these terms. com ; Xingran Cui, cuixr seu. Disclaimer: All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers.

Any product that may be evaluated in this article or claim that may be made by its manufacturer is not guaranteed or endorsed by the publisher. Top bar navigation. About us About us. Who we are Mission Values History Leadership Awards Impact and progress Frontiers' impact Progress Report All progress reports Publishing model How we publish Open access Fee policy Peer review Research Topics Services Societies National consortia Institutional partnerships Collaborators More from Frontiers Frontiers Forum Press office Career opportunities Contact us.

Sections Sections. About journal About journal. Article types Author guidelines Editor guidelines Publishing fees Submission checklist Contact editorial office.

ORIGINAL RESEARCH article Front. This article is part of the Research Topic Understanding the Role of the Autonomic Nervous System in Health and Disease View all 10 articles.

Heart rate variability in different sleep stages is associated with metabolic function and glycemic control in type 2 diabetes mellitus. TABLE 4. It shows a negative trend for age, and higher PRV for men than women.

Table 1 shows characteristics for healthy controls, MetS and DM subgroups, respectively all variables used in the study are listed in Table 2. Mean SDNN ms as a function of age for men and women from the period before the cold pressure test.

The Pearson correlation between measurements taken pre and post CPT was 0. The correlation between baseline heart rate and pre-CPT SDNN was In the presentation of HRV versus number of MetS components, all Figs. After the centering, the intercept represents the HRV levels for women at age 55 with no MetS components.

Mean SDNN from the pre-CPT period adjusted for sex and age was The effect of sex and age was 2. As shown in Fig. There were no other differences between the groups with 3—5 MetS levels and diabetes, consistent with a non-linear leveling off effect.

In other words, there is a significant reduction in SDNN for every added MetS component up to three components, but not beyond that point.

Results were similar for the other outcome variables, as shown in supplementary figures S1 and S2. The PRV was higher for SDNN outcomes compared to RMSSD, and in the post-CPT period as compared to pre-CPT see supplementary figure S1.

This analysis was repeated with participants on beta blockers excluded, see supplementary figure S3 for results. A breakdown of the percentage of fulfilled specific MetS components within each MetS level is shown in Table 3. Green area represents healthy subjects and yellow area represents subjects with metabolic syndrome according to current definitions.

Only Tukey tests from adjacent groups are shown in this plot, for a complete overview, see supplementary figure S3. In the models with the specific components and diabetes as dichotomized variables, all components contributed significantly to the alteration in PRV when SDNN from the post-CPT was used as outcome.

For pre-CPT SDNN, low HDL was not a significant component. For both RMSSD outcomes, diabetes and central obesity were not significant. In the models where low HDL was significant, it was associated with a higher PRV value. No three-way interactions were statistically significant for any outcome.

The relation between pre-CPT SDNN and HbA1c is presented in Fig. A Point density plot for the association between pre-CPT SDNN and HbA1c. A lighter color means that there is a higher density of observations in that area. Although HbA1c was included as a non-linear term in the GAM, the prediction line had a linear shape.

The untransformed association between pre-CPT SDNN and different ranges of HbA1c with and without moderately or severely increased albuminuria ACR above 3. The analyses of the main hypothesis were repeated, stratified by CVD status. The healthy group had similar results as in the main results, with the same contrasts being significant, and similar results for the Tukey tests except for the pairwise difference between 3 components and diabetes which was no longer significant.

The group with CVD, consisting of participants, did not seem to follow the same pattern as the healthy group. The same contrast analysis was performed, but no contrasts were significant with any of the outcomes.

The Tukey tests only showed that the level with 2 components was significantly different from most of the other levels. The result for the pre-CPT SDNN outcome is presented in Fig. This analysis was repeated with participants on beta blockers excluded, see supplementary figure S6 for results.

Only participants had known CVD, resulting in wider confidence intervals compared to the group without CVD. The models were also run on unimputed data, indicating that imputation did not have a large effect on the overall conclusions results not presented here, but available on request.

The results show that PRV was significantly reduced in subjects having one or more metabolic syndrome components or diabetes compared to healthy subjects. We found a significant decrease in PRV with increasing HbA1c up to the defined range for diabetes. The results regarding albuminuria were ambiguous.

The pattern from the main results was not present in the group with known CVD. Results were generally stronger with SDNN as the outcome as compared to RMSSD.

The decrease in PRV with increasing levels of MetS was non-linear and plateaued after the third component. This leveling at a low PRV for three MetS levels or more is of interest to the definition of the metabolic syndrome and support the ATP III definition 23 , where a diagnosis of metabolic syndrome can be made when three of these components are present.

We expected to find a further reduction in PRV going from five levels to the diabetes group, but these groups had equal mean PRV. The age distributions in the two groups are similar. In addition, some of the diabetes patients are well controlled, e.

As shown in supplementary Fig. S2 , the PRV was higher during the post-CPT period compared to pre-CPT. This could be explained as an effect of the CPT, as the test induces a significant increase in blood pressure and a sympathetic activation, with a possible parasympathetic compensation after the test.

In addition, the longer recording period in the post-CPT phase could include more slow-varying heart rate patterns and a possible effect of withdrawing the hand from the cold water, increasing the HRV estimate in particular for the SDNN variable. The components that were most influential on PRV differed depending on the outcome.

Many earlier studies have presented either the correlation between MetS components and HRV based on ECG recording , or the coefficient values from multiple regression using the continuous variables behind the MetS components. The strongest component with highest correlation or the largest regression coefficient is inconsistent across studies, sex, analysis method and type of HRV variable, and is therefore not easily comparable to this study.

In previous studies, participants with diabetes are also generally either excluded completely or included in groups of MetS, as opposed to this study where they are presented as a separate group. With respect to frequency domain HRV, every component has in some way been reported as the most influential, depending on the study and method of analysis.

For time domain measures, HDL is the only component consistently not listed among the most influential predictors according to our knowledge. Some studies have found this as the weakest contributor 24 , 25 , 26 , especially for SDNN.

In one study HDL was the second most influential for SDNN, but also the only component not significant for RMSSD All these four studies were based on correlations. In two studies that reported regression coefficients for SDNN and RMSSD, one study found that systolic blood pressure was the only component that was significant for both measures The other study did not find any significance for RMSSD, but triglycerides and systolic blood pressure were significant for SDNN, with triglycerides having the strongest association.

Two studies used dichotomized MetS components in accordance with this study, but none of those are directly comparable as both used frequency domain measures 28 , All significant associations between HRV and HDL in the mentioned studies were positive, while our study showed an inverse relation.

The interaction between central obesity and blood pressure was only significant in pre-CPT SDNN, and therefore we do not consider it to be a rigorous finding. No other two-way or any three-way interactions were detected given the statistical power of the sample. We did not find studies looking for interactions among the components in a comparable way.

A systematic review by Stuckey et al. Two studies tested the difference in SDNN between 0 and 1 components, one of them found a significant difference. Many grouped several MetS levels together.

We used different analysis methods to answer more specific questions related to MetS, diabetes and CVD. In addition, our study was an opportunity to see if the association holds with ultra-short PRV data as a surrogate for longer ECG-based HRV recordings, which was confirmed.

Since the review by Stucky et al. Bhagyashree et al. There were no differences between 0—2 and 3 components. Yoo et al. Véber et al. All nine studies found on this topic used ECG recordings, in contrast to the plethysmography-derived PRV used in our study.

Figure 3 A shows a relationship between SDNN and HbA1c that is visually not linear, but binning of HbA1c Fig. The GAMs also reported nearly linear relations, especially in the normal HbA1c range, even though it was included as a non-linear term.

From Fig. According to a meta-analysis by Hillebrand et al. With this assumption, the decrease in average SDNN seen in Fig. This is consistent with data from the EPIC-Norfolk study 31 that showed a positive continuous relationship between HbA1c concentration and rates of CVD in men, starting well within normal ranges of HbA1c.

Among people with diabetes, data from the UKPDS 35 32 shows that there is no definite lower HbA1c threshold for microvascular or macrovascular events, although the event rate is much lower with lower HbA1c levels Stratton et al.

BMJ Even though plasma glucose and HbA1c are not directly comparable, this finding supports the results of Ziegler et al. Jarczok et al. They used locally weighted regression lowess to fit regression lines, which were nearly linear as well in the normal glucose range , and in agreement with our results for the GAMs.

According to hypothesis 3, we investigated whether the association between ACR and PRV would hold regardless of whether HbA1c was included in the model or not, and the primary interest was not in an interaction term between the two.

We found that the assumption underlying the hypothesis, that ACR was associated with PRV at all, was weak. This relationship was only significant in the models that used SDNN as outcome. Furthermore, statistical significance for ACR log transformed vs PRV remained only for one of the two SDNN outcomes after inclusion of HbA1c, and we are therefore unable to conclude whether HRV is associated with ACR independently of HbA1c based on our data.

In three of the four models, we found a statistically significant interaction between log ACR and HbA1c, in which there was a negative association between log ACR and HRV for low values of HbA1c, but a positive association for higher values of HbA1c.

This number means that if the ACR is doubled, the expected decrease in SDNN is 0. The corresponding number in the pre-CPT model was 0. For participants with albuminuria, there was no visible trend between SDNN and HbA1c supplementary figure S4 , possibly due to a low sample size for this group.

For the participants without albuminuria, the SDNN-HbA1c trend was negative, in accordance with the overall association shown in Fig. When the main analysis is done separately in the CVD and healthy groups, we see that the confidence intervals for the CVD group are much wider, making it hard to detect any pattern of PRV over MetS levels for the participants with CVD.

One obvious reason for the wider intervals is that there are much fewer participants with known CVD than without, as the interval width decreases with sample size.

However, it is also possible that the CVD group has a higher natural variance because different types of CVD were included in this analysis, ranging from angina without damage to small troponin infarctions to large infarctions with heart failure.

As we do not see a big difference between the estimated standard deviations for the CVD and healthy group, the wide confidence intervals are more likely due to the relatively low number of subjects with CVD.

This study has some limitations. First, the Tromsø 6 sample does not correspond to the age distribution of the general population, but is skewed towards older participants, in accordance with the Tromsø 6 inclusion strategy. Second, we did not have access to fasting glucose for determining the MetS glucose criteria.

This could potentially lead to bias if healthy participants were misclassified as fulfilling the glucose criteria because they had elevated glucose levels following a meal.

However, this limitation does not seem to have affected the results notably, as the glucose component contributed significantly to all four HRV outcomes Table 4 , and there were clear differences between the healthy controls and MetS groups in most analyses.

Initially, we attempted to incorporate the self-reported time since the last meal variable in the definition, but that likely caused some participants with elevated glucose to be misclassified as controls. The criterion was consequently defined without that variable. This process is explained in detail in the supplementary information.

Third, some of the variables in the study are fully or partially based on self-report, such as diabetes and age at diagnosis, time since last meal, and CVD information that was not registered in the end-point registry. Self-reported data can be affected by biases, such as recall bias or social desirability bias.

However, since only the age at diabetes diagnosis goes far back in time, and the questions are not particularly sensitive, any bias from self-reporting for these variables is expected to be modest. Tromsø 6 has a good gender balance and a high response rate.

It also has the advantage of having a large number of participants, giving enough statistical power to answer detailed questions in the main hypothesis. Still, when looking into subgroups such as those with albuminuria or known CVD, it is possible that the lack of detected associations is due to an insufficient statistical power.

In this study we have used ultra-short-term PRV 30 or 60 s based on plethysmographic measurement as a surrogate for HRV that is conventionally derived from longer recordings of ECG.

As it is known that PRV may deviate from ECG-based HRV 15 and that ultra-short-term HRV is not completely validated as a surrogate 17 , our PRV data possibly contains noise components of either technical or physiological origin that increase variation and could have reduced the statistical power of the dataset.

In order to reduce noise in the dataset, the data had previously been cleaned for artifacts using automated procedures 33 and recordings with the highest PRV most likely to contain artifacts were visually inspected in this study and excluded if errors were suspected.

Some studies have found ultra-short-term HRV to be good surrogates for five-minute recordings depending on the recording length and the type of HRV metric.

With respect to the metrics used in this study RMSSD and SDNN , Baek et al. Shaffer et al. Munoz et al. With respect to diabetes patients, Nussinovitch et al. In our study, PRV was derived from 30 s pre-CPT and 60 s post-CPT recordings. Although RMSSD seems to require shorter recordings, this metric also seems to be more sensitive to recording artifacts from photoplethysmography-based PRV compared to SDNN Thus, the accuracy of our SDNN data may be reduced by the short recording length, while the RMSSD accuracy is more likely reduced as a consequence of plethysmography-based measurement.

The results of this study are of interest for applications based on PRV measurement for the purpose of predicting risk of complications associated with metabolic syndrome such as cardiovascular disease. The technology behind PRV measurement is simple to miniaturize and is already available in smart-watches 36 , It is important to note that although there is a significant difference on a population level, this does not guarantee that the data has the precision needed for a clinically useful test.

The ability to predict outcomes of interest based on this short test must be investigated in a further study. PRV may also be valuable as an addition to the metabolic syndrome criteria for risk prediction at the individual level, but this needs to be confirmed by further studies.

The large variance in the PRV data also indicates that experimental or technical factors could have influenced the measurement and that the reproducibility may be low. For predictions at the individual level, the measurement reproducibility must be optimized by consideration of the testing situation, measurement instrumentation and the signal processing used to derive PRV variables.

More reliable PRV data may be acquired by, for instance, smart watches, during sleep and other contexts where individual changes can be tracked over time. Having metabolic syndrome is defined as meeting at least three of the five criteria 3.

This definition corresponds well with our findings on PRV from the Tromsø 6 population, where the PRV drops from none to three MetS levels, but plateaus at the lowest level from three levels and above Fig. The PRV had significantly decreased already at the first MetS level, indicating that changes in cardiac or autonomous regulation may be detected at the earliest stages of MetS in individuals considered healthy.

This could be important, as symptoms of autonomic deficits often appear late in CAN, when reversibility is limited. Screening for CAN at an early stage could help manage or reverse the progression The MetS and DM populations are different from healthy controls with respect to PRV, indicating an impaired autonomic nervous system in both conditions, but we could not find a stronger alteration in the DM versus the MetS population in our sample.

This study supports the notion that both the MetS components and manifest diabetes affect the autonomic nervous system in this population. The Tromsø Study is a prospective epidemiological study of health problems, symptoms, and chronic diseases initiated in So far, 7 surveys have been carried out.

This study is based on data from the sixth wave, Tromsø 6, which was performed in —; 19, participants aged 30—87 and of both genders were invited and 12, The age distribution was similar for both sexes. The Tromsø study maintains a cardiovascular end-point registry that contains verified CVD diagnoses.

The Norwegian Data Protection Authority and the Regional Committee of Medical and Health Research Ethics, North Norway have approved the Tromsø 6 study.

The study complies with the Declaration of Helsinki, and each participant gave written informed consent prior to participation. Tromsø 6 included several questionnaires and interviews with questions on a wide variety of demographic, social and health-related topics, anthropometric measurements, clinical examinations and the sampling of blood and other biological materials.

Sampling procedures can be found in In addition, there were data from 10, participants who underwent a cold pressor test CPT 39 , 40 , 41 to evaluate pain sensitivity and cardiovascular reactivity.

They were continuously monitored with a non-invasive beat-to-beat blood pressure monitor Finometer Pro; Finapres Medical Systems, Amsterdam, The Netherlands before, during and after the test. The data from the Finometer Pro was used to calculate estimates of HRV as described by Bruehl et al. Briefly, short recordings of inter-beat intervals before approximately 30 s and after the CPT approximately 60 s were cleaned for artifacts and used to calculate different HRV parameters using the RHRV package in R.

The recordings during the test were not used in this study. After exclusion of pregnant women, recordings with technical errors and participants with diabetes mellitus type 1 DMT1 or atrial fibrillation, there were subjects with readable HRV measurements that were included in the study Fig.

Three urine specimens from 7, subjects were collected. Flowchart describing the sample of HRV data for healthy controls, subjects with metabolic syndrome and diabetes type 2 derived from the Tromsø 6 study. DMT1 diabetes mellitus type 1, DMT2 diabetes mellitus type 2, MetS metabolic syndrome.

In order to answer the research questions, the variables presented in Table 2 were selected for analysis. From these, additional variables were derived, such as the components of the metabolic syndrome. Missing values were therefore imputed employing the multivariate multiple imputation method 42 using the MICE package in R 43 with the default imputation models for continuous and categorical variables.

Derived variables such as the MetS components , were not included in the imputations, following the impute, then transform approach As not all missing values needed imputation planned missing, e. Variables with planned missing were not allowed to predict other variables controlled through the predictorMatrix argument in the mice function as this would make the missing values propagate through the dataset.

However, the variables from day 2 were allowed to predict each other. To supply the imputation model with more data, rows with missing outcome variables were also included. The missing outcome was imputed to not let them propagate , but all rows with imputed outcome variables were removed before the analysis.

The dataset was imputed 30 times, analyzed and pooled to include the uncertainty from the imputation process in the estimates. The imputation was done separately for each hypothesis.

We assumed that the missing data was missing at random. Some participants had abnormally high SDNN or RMSSD measures that were regarded as outliers. In order to determine whether or not to include these in the analysis, visual inspection of the inter-beat intervals was done for all SDNN values at or above , both pre-CPT recordings and post-CPT recordings.

This resulted in the exclusion of 32 and 59 measurements from each time period respectively, based on likely technical measurement errors. The participants were not removed from the dataset, but the HRV measure from the relevant time period was deleted.

Participants were excluded as having diabetes type 1 if they reported being diagnosed before or at the age of We used the harmonized MetS criteria 3 , which stated that metabolic syndrome should be diagnosed if three or more of the following five criteria were met:.

We also considered criteria 2, 3, and 4 to be met if the participant was on drug treatment for the condition at the time. The glucose criterion in the definition of MetS reflects fasting glucose, but the participants in this study were not asked to fast before attendance.

We did not use the self-reported time since last meal variable in the definition of the criterion, the justification for this can be found in the supplementary information. Instead, the criterion was defined as follows:. CVD was defined as myocardial infarction, angina and stroke.

For myocardial infarction and stroke, information from the cardiovascular end-point registry was combined with the information from the questionnaires. As angina is not in the registry it is based on the questionnaire alone. Albuminuria was measured as the albumin-creatinine ratio ACR calculated as the average ratio from three different measurements.

This was only available from the subgroup of participants that attended the second visit, and participants had both ACR data and readable PRV data. Using the Tromsø 6 dataset, our aim was to investigate how the metabolic syndrome and diabetes type 2 is associated with changes in HRV.

Our primary hypothesis was that altered HRV is related to the number of metabolic syndrome components, with the strongest alteration in manifest diabetes. To answer this, a variable counting the number of fulfilled MetS criteria per individual was constructed, with a separate category for participants with diabetes independent of their number of MetS criteria.

This will from now on be referred to as the MetS levels. Not assuming a linear relationship, this variable was used as a categorical independent variable in an ANCOVA model. To further investigate differences between specific groups, contrast analysis and Tukey tests were used. In order to estimate how each specific MetS component and diabetes contributed to alterations in PRV, a model with the MetS components as binary variables and diabetes was employed.

Main effects and interactions up to the three-way level were assessed. Alteration in HRV increases with increasing number of metabolic syndrome components, with the strongest alteration in manifest diabetes in subjects with and without known CVD.

Linear regression and generalized additive models GAM were used to explore the relationship between PRV and HbA1c.

As the GAM package mgcv was not compatible with imputation, results from unimputed data are presented. To test whether the relationship between PRV and albuminuria is independent of the level of HbA1c, linear regression with log transformed ACR was used.

To answer the last hypothesis, the analysis of the main hypothesis was repeated with data grouped on whether the participants had known CVD or not. The analyses were repeated with SDNN and RMSSD, both before and after the CPT, as outcome variables.

All models were adjusted for sex and age centered and all statistical tests were two-sided and done with a fixed significance of 0. All analyses were done in R version 4. To respect terms of use and protect human privacy, the authors cannot directly share the dataset. James, S. et al. Global, regional, and national incidence, prevalence, and years lived with disability for diseases and injuries for countries and territories, — A systematic analysis for the Global Burden of Disease Study The Lancet , — Article Google Scholar.

Afshin, A. Health effects of overweight and obesity in countries over 25 years. Article PubMed Google Scholar. Alberti, K. Harmonizing the metabolic syndrome. Circulation , — Article CAS PubMed Google Scholar.

Thanopoulou, A. Epidemiological evidence for the non-random clustering of the components of the metabolic syndrome: Multicentre study of the Mediterranean Group for the Study of Diabetes. Stuckey, M.

Heart rate variability and the metabolic syndrome: A systematic review of the literature. Diabetes Metab. Liao, D. Multiple metabolic syndrome is associated with lower heart rate variability: The atherosclerosis risk in communities study. Diabetes Care 21 , — Vinik, A. Cardiac autonomic neuropathy in diabetes: A predictor of cardiometabolic events.

Article PubMed PubMed Central Google Scholar. Maser, R. The association between cardiovascular autonomic neuropathy and mortality in individuals with diabetes a meta-analysis. Diabetes Care 26 , — Spallone, V. Cardiovascular autonomic neuropathy in diabetes: Clinical impact, assessment, diagnosis, and management.

Williams, S. Cardiac autonomic neuropathy in obesity, the metabolic syndrome and prediabetes: A narrative review. Diabetes Therapy 10 , — Ziegler, D. Increased prevalence of cardiac autonomic dysfunction at different degrees of glucose intolerance in the general population: The KORA S4 survey. Diabetologia 58 , — Yoo, H.

Clinical implication of body size phenotype on heart rate variability. Metabolism 65 , — Véber, O. Obstructive sleep apnea and heart rate variability in male patients with metabolic syndrome: Cross-sectional study. Bhagyashree, N.

Is autonomic function test helps to assess the severity of metabolic syndrome: A study on comparison of frequency-domain recordings of Heart rate variability HRV with the severity of metabolic syndrome.