OUP user menu

Effects of physical training on heart rate variability in diabetic patients with various degrees of cardiovascular autonomic neuropathy

Kinga Howorka, Jiri Pumprla, Paul Haber, Jeanette Koller-Strametz, Jerzy Mondrzyk, Alfred Schabmann
DOI: http://dx.doi.org/10.1016/S0008-6363(97)00040-0 206-214 First published online: 1 April 1997


Objective: To investigate the effects of regularly performed endurance training on heart rate variability in diabetic patients with different degrees of cardiovascular autonomic neuropathy (CAN). Methods: Bicycle ergometer training (12 weeks, 2×30 min/week, with 65% of maximal performance) was performed by 22 insulin-requiring diabetic patients (age 49.5±8.7 years; diabetes duration 18.6±10.6 years; BMI 25.1±3.4 kg/m2): i.e., by 8 subjects with no CAN, 8 with early CAN and by 6 patients with definite/severe CAN. A standard battery of cardiovascular reflex tests was used for grading of CAN, a short-term spectral analysis of heart rate variability for follow-up monitoring of training-induced effects. Results: While the training-free interval induced no changes in spectral indices, the 12-week training period increased the cumulative spectral power of the total frequency band (P=0.04) but to a different extent (P=0.039) in different degrees of neuropathy. In patients with no CAN the spectral power in the high-frequency (HF) band (0.15–0.50 Hz) increased from 6.2±0.3 to 6.6±0.4 ln [ms2]; P=0.016, and in the low-frequency (LF) band (0.06–0.15 Hz) from 7.1±0.1 to 7.6±0.3 ln [ms2]; P=0.08, which resulted in an increase of total spectral power (0.06–0.50 Hz) from 7.5±0.1 to 8.0±0.3 ln [ms2] (P=0.05). Patients with the early form of CAN showed an increase of spectral power in HF (5.1±0.2 to 5.8±0.1 ln [ms2], P=0.05) and LF bands (5.6±0.1 to 6.3±0.1 ln [ms2], P=0.008), resulting in an increase of total power from 6.1±0.1 to 6.6±0.1 ln [ms2] (P=0.04), whereas those with definite/severe CAN showed no changes after the training period. Training improved fitness in the whole patient cohort. The increased autonomic tone as assessed by spectral indices disappeared after a training withdrawal period of 6 weeks. Conclusions: In diabetic patients with no or early CAN, regularly performed endurance training increased heart rate variability due to improved sympathetic and parasympathetic supply, whereas in subjects with definite/severe CAN no effect on heart rate variability could be demonstrated after this kind of training.

  • Autonomic nervous system
  • Heart rate, variability
  • Spectral analysis
  • Diabetes
  • Endurance training
  • Human

Time for primary review 28 days.

1 Introduction

Diabetic autonomic neuropathy is associated with a reduced autonomic supply to the heart [1]which can be estimated by a decrease of heart rate variability (HRV) [2]. Although cardiovascular autonomic neuropathy (CAN) and decreased HRV are associated with an increased mortality [3–6], induced in part by sudden death, no data are available on factors which might improve the autonomic supply to the heart in diabetes. In healthy athletes [7–9], in hypertension [10], in patients with chronic congestive heart failure [11]and/or coronary artery disease [12], and in patients after recent myocardial infarction [13, 14]systematic physical training was shown to induce improvement in the autonomic balance with a restoration to normal of the reflex activity of the system. Although the mechanisms of training-induced regulatory changes have still not been completely delineated, regularly performed physical training is generally thought to improve cardiovascular performance and to increase HRV [15]. Moreover, in diabetes physical exercise is considered to be essential in reducing insulin resistance [16], overweight [17], and enhancing the change in life style [18].

The prevalence of cardiovascular autonomic dysfunction in diabetes is high [2]and in part responsible for the diminished life expectancy in this chronic disease [4]. The primary goal of our study was therefore to investigate the effects of regularly performed physical training on HRV in insulin-requiring diabetic patients with different degrees of CAN as assessed by variables of short-term analysis of HRV in frequency and time domains. The trial was designed to investigate the potential reversibility of at least the early stages of autonomic neuropathy by those means which are available to every sedentary diabetic subject.

2 Methods

The investigation conforms with the principles outlined in the Declaration of Helsinki [19].

2.1 Study design

Intraindividual comparison of baseline values (mean of two measurements before and after a run-in period of 4–6 weeks with sedentary, unchanged life style) with values after a training intervention of 12 weeks endurance training and with final values after a consecutive period of 4–6 weeks of training withdrawal. Every patient in this design served as his own control.

2.2 Patients

From our clinical survey population of about 300 insulin-requiring diabetic patients we recruited 3 patient groups comparable in age, glycemic control and BMI, but differing in severity of their CAN as assessed by a standard battery of cardiovascular reflex tests [1]. Clinical data of patients who completed the study are summarized in Table 1. Patients with exercise-limiting circumstances, uncontrolled associated diseases such as hypertension, and those with clinical and/or electrocardiographic signs of coronary artery disease were excluded. Pharmacological treatment of concomitant diseases and nutrition were kept constant during all study periods. All patients had been receiving functional insulin treatment for at least 3 months before the study. As previously described [20]this program involves separate substitution of fasting requirements by twice-daily long-acting human insulin independent of prandial and correctional use of regular insulin according to individual algorithms. All patients undergo an extensive, individually tailored patient education program including group and individual sessions.

View this table:
Table 1

Clinical characteristics of insulin-requiring diabetic patients with various degrees of autonomic neuropathy

Patient groupsAllNo CANEarly CANSevere CAN
n (F/M)22 (13/9)8 (5/3)8 (5/3)6 (3/3)
Age (years)49.5±8.744.1±8.551.8±7.250.3±11.5
BMI (kg/m2)25.1±3.424.4±2.825.3±3.125.8±4.8
Diabetes duration (years)18.6±10.618.9±12.715.5±8.521.7±8.1
HbA1c (%)6.9±0.96.7±0.77.1±1.06.9±1.1

2.3 Training procedures

Patients underwent 12 weeks bicycle ergometer training controlled by individual training heart rate twice per week 30 min per training unit with 60–70% of individual maximal capacity. The target training heart rate was estimated during initial spiroergometry and was defined as heart rate at 65% of maximal exercise capacity [training heart rate = resting heart rate + (maximal heart rate − resting heart rate) × 0.65] [21]. This endurance training modality was chosen because it is known from clinical experience [22, 23]to improve cardiovascular performance.

2.4 Outcome variables and their assessment

Main outcome variables were those of short-term spectral analysis of HRV.

2.4.1 CAN assessment

CAN was assessed initially by a standard battery of cardiovascular reflex tests to standardized stimuli [1, 24]including deep breathing, Valsalva maneuver, orthostatic load and sustained handgrip. Diabetic patients with a Total Ewing Score of 0–0.5 were considered as those without CAN, patients with score of 1–2.5 as with early CAN and those with a score of 3–5 as patients with definite/severe CAN [1, 2]. For follow-up monitoring a short-term spectral analysis consisting of 3 time segments (adapted from Bellavere [25], in positions supine–standing–supine, 256 artifact-free heart beats each) was performed. This short-term option delivers well-reproducible results when recordings are performed under similar laboratory conditions as applied in our study [25]. Both CAN examinations were performed with the VariaPulse TF3 system (Sima Media Olomouc, Ltd., Czech Republic) [26]. A surface ECG was continuously monitored with a time resolution of 1 ms. R–R intervals were telemetrically transferred to a receiver connected to a PC-compatible computer and displayed on-line together with an instantaneous spectral curve on a monitor. The computational method was based on fast Fourier transform modified by algorithm of coarse-graining spectral analysis [27]. This allowed one to extract a broad-band non-harmonic ‘noise’ contaminating particularly the lower frequencies (1/f component). Each dataset was filtered automatically by excluding recorded artifacts using a recognition algorithm, as well as manually. The final results were immediately displayed on the monitor as 3-dimensional running spectra (Fig. 1), permitting a general overview of the dynamics and of the absolute energy content of the system. Parameters of the frequency domain were observed in every position within the high-frequency band (0.15–0.50 Hz), which has been attributed exclusively to parasympathetic tone [28], and within the low-frequency band (0.05–0.15 Hz), said to represent a combination of sympathetic and parasympathetic effects on cardiac autonomic tone [29, 30]. Main outcome variables were spectral power (units [ms2]) in both frequency bands. To increase the reliability of short-term measurements and to assess even small intraindividual improvements in global autonomic tone during the trial, we used cumulative indices (spectral power of the total frequency band with its high- and low-frequency components over all 3 positions) representing the total averaged area under all consecutive spectral curves within the short-term recording. Standard deviations were calculated for each parameter as this information was necessary for assessment of the stationarity of the examination. We excluded any findings having a more than 30% relative deviation in any of the positions recorded. In those cases the examinations were repeated. Time-domain analysis was also performed to calculate the mean square of the difference of successive R–R intervals (MSSD) [ms2].

Fig. 1

Examples of typical impact of regularly performed physical training on heart rate variability as assessed by short-term spectral analysis before and after (right panel) training period in patients with various degrees of CAN. Diabetic patient with no CAN (41-year-old male, Ewing score 0, BMI 27.4 kg/m2, diabetes duration 8 years, top panel), with early involvement (50-year-old male, Ewing score 2, BMI 23.5 kg/m2, diabetes duration 29 years, middle panel), and one with severe involvement (43-year-old male, Ewing score 4, BMI 21.4 kg/m2, diabetes duration 25 years, bottom panel).

2.4.2 Spiroergometry

Symptom-limited incremental bicycle spiroergometry (beginning with 25 W, increments of 25 W every 2 min) was performed (Computer Ergometer Ergoline, Ergometrics 900, Bits, Germany, and SensorMedics Metabolic Measurement Cast 2900, CA, USA) during all 4 main points of the trial before and after the training period to define maximal work capacity and maximal oxygen consumption [31], anaerobic threshold [32], and training heart rate [21].

2.4.3 Ambulatory blood pressure monitoring (ABPM)

ABPM was performed using a SpaceLabs 90207 monitor (SpaceLabs, Redmond, USA, 1993) according to the recommendations of the British Hypertension Society [33]. Additional outcome variables such as mean systolic and diastolic blood pressure, mean heart rate, and night-time dipping were assessed as previously described [34].

2.4.4 Echocardiography

Transthoracic two-dimensional echocardiography was performed to assess the presence and/or changes in left ventricular hypertrophy, end-systolic and end-diastolic left ventricular diameter and wall thickness of the interventricular septum. Continuous waved Doppler sonography was used to assess disturbances of relaxation [35]and left ventricular fractional shortening. All measurements were performed by the same physician (J.K.-S.) experienced in echocardiography, using the VingMed CFM 750 system (Diasonics-Sonotron, Zug, Switzerland).

2.4.5 Blood analysis

HbA1c was assessed by the HPLC method (Variant, Bio-Rad, Richmond, CA, USA; reference range 4.4–6.6%) and blood lipids by standard commercial kits from Boehringer Mannheim (Mannheim, Germany).

2.4.6 Depression scale

Symptoms of depression were assessed during all main investigation points by respective scores of the Beck Depression Inventory [36], a well-established self-reporting measure frequently utilized in empirical evaluation of psychiatric depression.

2.5 Statistical analysis

Statistical analysis was performed using standard statistical packages (SPSS, Statistical Package for the Social Sciences V7.0, SPSS Inc., Chicago, USA). Analysis of variance with repeated measures (General Linear Model) was used to evaluate differences between groups in the course of parameters during and after endurance training. The two-tailed paired Student's t-test was applied to estimate differences vs. baseline values. Because of the skewness of the frequency-domain data distribution, log (ln) transformation was performed to produce a normal distribution before the final results were assessed.

Demographic data are presented as means±s.d., outcome variables as means±s.e.m., unless otherwise indicated.

3 Results

While the training-free run-in interval induced no changes in spectral indices in all patient groups, the 12-week training period increased the cumulative spectral power of the total frequency band (P=0.04) but to a different extent (two-way interaction, differences in course between groups, P=0.039) in various degrees of neuropathy (Figs. 1 and 2). Whereas patients with severe CAN showed after the training period no measurable changes in cumulative total spectral power (0.06–0.50 Hz) (ln [ms2]before 4.7±0.3; ln [ms2]after 4.6±0.5; P=0.8), those with the early form of CAN and those without any detectable neuropathy showed an increase from ln [ms2]before 6.1±0.1 to ln [ms2]after 6.6±0.1; P=0.04 and from ln [ms2]before 7.5±0.1 to ln [ms2]after 8.0±0.3; P=0.05, respectively (Fig. 2, top panel).

Fig. 2

Effect of regularly performed physical training and training withdrawal period on spectral indices of HRV in groups of diabetic patients with various degrees of CAN: cumulative spectral power of total frequency band (top panel), cumulative spectral power in low-frequency band (middle panel) and in high-frequency band (bottom panel).

Spectral power in the low-frequency band (0.06–0.15 Hz) indicated significant effects of training (P=0.002), but also to a different degree between groups (two-way interaction, differences in course between groups, P=0.01), showing a highly significant rise in the group with early CAN (from ln [ms2]before 5.6±0.1 to ln [ms2]after 6.3±0.1; P=0.008), borderline amplification in the group without CAN (from ln [ms2]before 7.1±0.1 to ln [ms2]after 7.6±0.3; P=0.08), but no changes in the group with severe CAN (ln [ms2]before 4.1±0.3; ln [ms2]after 4.1±0.4; P=0.6) (Fig. 2, middle panel).

In the high-frequency band (0.15–0.5 Hz) no more statistical differences between the groups in the course of the parameters were revealed (two-way interaction, P=0.121). If compared to baseline values, however, increases were still evident in both groups without severe CAN (early CAN from ln [ms2]before 5.1±0.2 to ln [ms2]after 5.8±0.1; P=0.05; no CAN group from ln [ms2]before 6.2±0.3 to ln [ms2]after 6.6±0.4, P=0.016), whereas the group with severe CAN showed no statistical changes (ln [ms2]before 3.7±0.5; ln [ms2]after=3.4±0.7, P=0.6) (Fig. 2, bottom panel).

Time-domain analysis of HRV showed the effects of training on the MSSD (P=0.03), but the differences in course between the groups were not significant (two-way interaction, P=0.15). When values after training were compared to baseline values, significant amplification was present only in the group with no CAN (from [ms2]before 1848±375 to [ms2]after 3339±797, P=0.03; early CAN from [ms2]before 636±115 to [ms2]after 1062±372, P=0.3; severe CAN from [ms2]before 201±51 to [ms2]after 227±132, P=0.9).

After training withdrawal of 4–6 weeks, all training-induced changes in HRV could no longer be detected.

The impact of training on spectral and time-domain indices of HRV, if considered in any particular position (supine–standing–supine), as well as on their ratios, was not sufficient to reveal any statistical changes.

As assessed by incremental spiroergometry, maximal performance capacity increased over the training period in the whole patient cohort (P<0.001), although the impact of endurance training was different between the groups (two-way interaction, differences in course between the groups, P=0.02) and was higher in patients with no and early involvement (Fig. 3, top panel). The maximal oxygen consumption increased in the whole cohort (P=0.005) and no differences in its course between the groups could be detected (two-way interaction, P=0.11) (Fig. 3, lowest panel). Similarly, a significant increase in anaerobic threshold was observed in all patient groups (P<0.001 Fig. 3, middle panel), and no differences in its course between the groups were assessed (P=0.79). At different load-levels (e.g., 50 and 100 W) slight, but non-significant decreases in heart rate were recorded after endurance training in all groups, indicating a tendency to improvement of cardiac output/beat. With the exception of the group with no CAN, which still showed an increased exercise capacity even after training withdrawal (Fig. 3, top panel, P=0.002 vs. baseline values), after the period without training all initially induced beneficial effects had disappeared.

Fig. 3

Effect of regularly performed physical training and training withdrawal period on maximal performance capacity (top panel), anaerobic threshold (middle panel) and maximal Vo2 consumption (bottom panel) in groups of diabetic patients with various degrees of CAN.

Ambulatory blood pressure monitoring demonstrated a numerical decrease of a mean of 24 h heart rate in the whole patient cohort (P=0.081) and no differences between the degree of impact were detectable between the groups (P=0.58). Moreover, the heart rate remained slightly lower even after 6 weeks of training withdrawal. Endurance training induced an increase of night-time dipping only in the group with no CAN for systolic (16±3 vs. 21±2 mmHg, P=0.03) as well as for diastolic blood pressure (14±2 vs. 17±2 mmHg, P=0.04).

Echocardiographic evaluation revealed during baseline investigations no signs of left ventricular hypertrophy in patients with no and early involvement. In patients with severe CAN septum thickness (10.9±1.8 mm) indicated left ventricular hypertrophy. Training induced no significant reductions of interventricular wall thickness (no CAN from 9.5±0.6 to 9.0±0.5 mm, P=0.48; early CAN from 9.2±0.8 to 9.2±0.6 mm, P=0.85) and of left ventricular end-diastolic diameter (no CAN from 52.5±1.9 to 51.9±2.5 mm, P=0.57; early CAN from 49.2±1.6 to 47.4±1.1 mm, P=0.14). In contrast, in the group with severe involvement, training had an inverse but non-significant effect on both interventricular wall thickness (10.9±1.8 to 11.4±2.2 mm, P=0.52) and left ventricular end-diastolic diameter (from 50.9±1.6 to 53.3±1.8 mm, P=0.43). Disturbance of relaxation was present only in patients with early and severe neuropathy. Training induced minor improvements in both groups in this respect.

The training period had no relevant effects on blood lipids or HbA1c.

Depression symptoms as assessed by Beck's Depression Inventory were not affected by the training (P=0.66), but it should be stressed that the depression level revealed was quite low (level of severity 0–26), with the exception of 2 individuals in the group with severe involvement (level of severity 26–40).

4 Discussion

This study documents for the first time the beneficial effects of low-grade endurance training on heart rate variability in the early stages of diabetic cardiovascular autonomic neuropathy. As our trial was designed to investigate the training-induced reversibility of diabetic cardiovascular autonomic neuropathy and data are gathered on the HRV-related impact of endurance training in healthy subjects [7–9], we dispensed with a non-diabetic control group.

The effect of low-grade endurance training was more pronounced in the group with early CAN than in subjects with no CAN. Our study demonstrates the already known phenomenon that pathologic values (including blood lipids, blood pressure values, incipient CAN, low initial fitness levels) are more strongly influenced towards normal by endurance training than normal values. This finding is consistent with our previous observations [22]and reports of others [37]. Additionally, it is possible that the chosen training modality represented a stronger stimulus for patients with early involvement, as they displayed a much lower exercise capacity.

The design included a period of run-in with no training to ensure that the selected methods for assessment of CAN deliver reproducible staging of neuropathy. In order to reduce the standard deviation of particular indices of HRV, mean values of both pre-training investigations were used for baseline. The relevant reconditioning of autonomic balance of the heart supply was expected first after a training period of 12 weeks [22, 23]; the described training intensity and length were well accepted and tolerated by the majority of patients, although in 5 initially selected cases (3 with severe and 2 with early involvement) the training could not be completed. These early drop-outs were caused by acute severe pemphigus (hospitalization), severe neuropathic diarrhea, benign lung tumor (surgery), and non-compliance with the necessary supervised heart-rate-controlled training.

This study provided evidence that the used method of short-term spectral analysis of HRV is practicable and easily applicable in ambulatory routine. As it was found that the immediate variability of short-term spectral measures of HRV was low [38]and short, 2- to 15-min samples were excellent predictors of mortality and correlated with prognostically important data from sustained recording periods [39, 40], we have chosen to exclusively perform short-term spectral analysis for repetitive assessment of HRV. Due to this simplification, patients could accept 4 consecutive measurement series during the study; the method takes only a short time and is quite independent of the patient's compliance during the examination.

In contrast to the previous, much less sensitive standard score of the cardiovascular reflex battery, the selected cumulative indices over all 3 time periods (supine–standing–supine) proved able to assess even small changes in cardiac autonomic supply. The cumulative indices proposed here should be understood as a method for increasing the reliability of the measurements during short-term recordings in all 3 positions. Although the information given by the cumulative spectral indices corresponds to that given by indices of time-domain analysis (MSSD), it became evident that this index is more representative of the actual global state of autonomic regulation and reflects the total instantaneous sympathetic and parasympathetic effects. A similar investigation using an analogue training program in sedentary middle-aged men [41]showed apparently no effects on HRV, which—to our present knowledge—can be viewed partly as a methodological problem since the cumulative indices designed by us were not considered in this previous investigation.

The current methods most commonly used for analysis of HRV in the frequency domain are based either on the fast Fourier transform or on the autoregressive model. Under several conditions considered also in our study design both analytical approaches deliver similar results [28, 42]. The procedure of coarse-graining spectral analysis [27]was additionally chosen to increase the reliability of the measurement process. This method has the advantage that the non-harmonic ‘noise’ contaminating the spectrum is removed and the high processing speed allows on-line graphic display.

Our study documented a training-induced increase of complex sympathetic and parasympathetic supply as reflected by amplification of the spectral power of the low- and high-frequency bands. A similar finding was reported by Furlan [9], although in healthy subjects the changes induced by heavy training were mainly restricted to the low-frequency band.

The endurance training modality applied was not able to improve cardiac supply in patients with severe CAN. In 3 severely affected subjects we recorded even an on-going loss of total spectral power. Training improved maximal performance capacity and maximal oxygen consumption, and increased the anaerobic threshold in the whole diabetic cohort, but, similar to the effects on HRV, the effects of endurance training on fitness were dependent on the degree of CAN.

It is evident that neither the emotional state as assessed by Beck's Depression Inventory nor other factors such as current treatment of diabetes or hypertension were responsible for the changes assessed as the impact of training disappeared after training withdrawal.

If the demonstrated reversibility of CAN as induced by systematic endurance training in our study corresponds to the decrease of risk associated with cardiac denervation, endurance training and rehabilitative exercise programs in diabetic patients with early involvement should be recommended. Further studies are necessary to investigate the potential reversibility of nephropathy and other cardiac denervation-associated phenomena, such as increased microalbuminuria [43]and decreased glomerular filtration rate [44]. It should be kept in mind, however, that—as documented also by our investigation—apparently only a systematic, regularly performed physical training is necessary for lasting beneficial effects. Our study provides the first necessary rationale for a long-term study of the effects of endurance training upon HRV indices in diabetic patients with cardiovascular autonomic neuropathy. Although that should still be proven in longer studies, it is already known that the time period needed to inverse the training-induced effects on fitness (including its morphological correlates) is related to the duration and intensity of the training [37].

Recent data suggest that maximal oxygen consumption itself is a good predictor of mortality risk [45]. Further mortality analyses in diabetes should consider the physical state and exercise behavior as important variables which might influence the mortality risk.


Our work was supported by promotional grants for international cooperation from the University of Vienna, Austria, and the Preventa Foundation, Olomouc, Czech Republic to J.P.


View Abstract