CRVASA-499; No.

of Pages 10

cor et vasa xxx (2017) e1e10

Available online at www.sciencedirect.com

ScienceDirect

journal homepage: http://www.elsevier.com/locate/crvasa

Original research article

Heart rate variability evaluation in the assessment

of cardiac autonomic neuropathy in patients with
type 2 diabetes

Rudolf Metelka *, Lubica Cibikov, Jaromra Gajdov, Ondej Krystynk

IIIrd Internal Medicine Department Nephrology, Rheumatology, and Endocrinology, Palacky University Faculty of
Medicine and University Hospital Olomouc, Czech Republic

article info abstract

Article history: Introduction: Heart rate variability (HRV) is a respected measure used in the assessment of
Received 25 January 2017 cardiac autonomic neuropathy (CAN) and it can serve as an independent prognostic
Accepted 1 May 2017 indicator of sudden arrhythmic death risk. Despite the importance of early detection, the
Available online xxx diagnosis of CAN is often made too late, especially in diabetics. Besides the long subclinical
phase of CAN, reasons for this include great diversication of employed diagnostic methods
Keywords: and absence of universally accepted normal values; the latter applies mostly in HRV
Cardiac autonomic neuropathy evaluated using short-term spectral analysis (SAHRV).
Spectral analysis of heart rate Aim: The aim of this cross-sectional study involving patients with type 2 diabetes was to
variability summarize the real potential of using a testing method for CAN diagnosis by short-term
SupineStandingSupine test SAHRV, including an autonomic load imposed during an orthoclinostatic test (Supine1
Functional age of ANS StandingSupine2, short 5-min recordings). Three different normative approaches to the
Diabetes mellitus postprocessing analysis of acquired data described by different authors were employed.
Secondary aim of the study was to assess the benet of rate-controlled breathing. The
next aim was to compare the HRV data measured with the selected clinical and laboratory
indices in patient examined.
Materials and methods: The study included 43 patients with type 2 diabetes (12 women, 31
men, mean age 51.1  10.7 y) and no history of manifest CAN or serious cardiovascular
illness, except uncomplicated hypertension. Using a diagnostic system DiANS PF8 with
telemetric transfer of ECG and respiratory rate, series of reex tests according to Ewing and
SAHRV (Fourier tachogram analysis, window 256) during autonomic load imposed by
Supine1StandingSupine2 test (SSS test) and during 5 min of rate-controlled, non-deep-
ened breathing (PB, 12 cycles/min) were performed. Acquired spectral indices were analyzed
and compared with normatives of 3 authors using the same recording algorithm, SSS test,
but different data postprocessing analysis. These were (1) so called functional age of
autonomic nervous system (ANS), (2) assessment of CAN severity according to age-stratied
medians and percentiles, (3) assessment of CAN severity according to cumulative spectral
power during the entire test (cumLFHF).

* Corresponding author at: IIIrd Internal Medicine Department Nephrology, Rheumatology, and Endocrinology, Palacky University Faculty
of Medicine and University Hospital Olomouc, I. P. Pavlova 6, 77900 Olomouc, Czech Republic.
E-mail address: rudolf.metelka@fnol.cz (R. Metelka).
http://dx.doi.org/10.1016/j.crvasa.2017.05.001
0010-8650/ 2017 The Czech Society of Cardiology. Published by Elsevier Sp. z o.o. All rights reserved.

Please cite this article in press as: R. Metelka et al., Heart rate variability evaluation in the assessment of cardiac autonomic neuropathy in
patients with type 2 diabetes, Cor et Vasa (2017), http://dx.doi.org/10.1016/j.crvasa.2017.05.001
CRVASA-499; No. of Pages 10

e2 cor et vasa xxx (2017) e1e10

Results: According to the total Ewing score (ETS), 11.6% patients were categorized as CAN-
free (ETS = 0), 32.6% were diagnosed with possible CAN (ETS = 1), and 55.8% labeled with
manifest CAN (ETS = 23). Moderate correlation between ETS and individual SAHRV param-
eters following orthoclinostasis (test SSS) in Supine2 position was described [ms2]: TP (total
power, f = 0.020.5 Hz): r = 0.4, p < 0.006; LF component (low frequency, 0.050.15 Hz):
r= 0.31, p < 0.04; HF component (high frequency, 0.150.5 Hz): r = 0.45, p < 0.003) and
the same applied to rate-controlled breathing PB (TP, [ms2]: r = 0.56, p < 0.0001; LF:
r= 0.38, p < 0.018; HF: r = 0.52, p < 0.001). Moderate correlation was also found between
ETS and HRV assessment using a complex indicator functional age of ANS (r = 0.37,
p < 0.015), ETS and cumLFHF [ms2, ln ms2]: r = 0.46, p < 0.002). In most patients, signicant
difference between functional age of ANS and calendar age was conrmed (mean 21.8  12.9
y, median 23.5 years, p < 0.0001). An attempt to assess the severity CAN using age-stratied
medians and percentiles of TP, LF, HF, and LF/HF was not successful.
As for SAHRV and clinical indices (anthropometric, echocardiographic, QTc, laboratory),
moderate correlation between the glycated hemoglobin on one side and basic SAHRV indices
(TP, LF, HF, LF/HF), functional age of ANS and cumLFHF on the other side was prominent
(r = 0.360.53, p < 0.0004 to p < 0.02).
Conclusion: Assessment of CAN using evaluation of HRV can optimally be performed (and
simply realized in clinical practice) using SAHRV based on short ECG recordings during
autonomic load imposed by orthoclinostatic test (Supine1StandingSupine2) and on post-
processing data analysis using complex indicator called functional age of ANS. In the
detailed evaluation of sympathovagal balance, it complements the screening assessment
with cardiovascular reex tests (Ewing's battery). Besides the orthoclinostatic load, pro-
nounced vagal provocation using rate-controlled, non-deepened breathing (12 cycles/min)
represents a recommended facultative load option increasing the yield of the SAHRV method.
The detection and assessment of CAN severity while applying the cumulative indicator of
HRV (cumLFHF) showed a good discrimination power in the frontline screening for CAN, albeit
without the possibility to distinguish between the sympathetic and vagal branch of ANS.
Presented cross-sectional study in type 2 diabetes mellitus demonstrated a signicant
autonomic dysfunction in majority patients examined, independently of diabetes duration.
It supports the recommendation to assess the ANS integrity in type 2 diabetes already at
diagnosis, within the initial staging of the illness. The severity of CAN correlates well with
metabolic control of diabetes as evaluated by glycated hemoglobin.
2017 The Czech Society of Cardiology. Published by Elsevier Sp. z o.o. All rights reserved.

orthostatic test, Valsalva maneuver or a handgrip. Methods

Introduction
According to available data, the prevalence of diabetic
autonomic neuropathy (DAN) is between 1.6% and 90% while
the prevalence of cardiac autonomic neuropathy (CAN) varies
from 1% to 90% in patients with type 1 diabetes and from 20%
to 73% in patients with type 2 diabetes [13]. The reasons for
this immense variability of DAN prevalence include its diverse
symptomatology, long subclinical phase leading to late
diagnosis of organ-specic neuropathy, absence of uniformity
during selection of a population sample of diabetics, diversity
of diagnostic methods, and absence of uniform normative
data. It is aimed to dene norms (age-stratied normative
data) in this context to enable CAN diagnosis during its
subclinical phase, prevent irreversible changes, and widen the
therapeutic window for early intervention.
Reex tests of autonomic cardiovascular functions accord-
ing to Ewing are considered the basis of these aims and the
gold standard [4,5]. In these reex tests, autonomic
cardiovascular functions are evaluated by simple indices
describing changes in heart rate during deep breathing,
analyzing sympathovagal balance in more details using heart
rate variability (HRV) dened as physiological oscillations of
R-R interval on the electrocardiogram (ECG) most often using
short-term spectral analysis (SAHRV), developed historically
from the Ewing's tests. The basic consensual methodological
material for HRV assessment is a TASK Force from 1996 [5].
To name the largest studies dealing with normal values of
outcome short-term SAHRV data, Agelink et al. (2001) [6] tested
309 healthy volunteers and Nunan et al. (2010) published a
meta-analysis on this topic [7] summarizing the results of
other 44 studies published from 1996 to 2008 (21 438 probands).
The main feature of these trials is the usage of short ECG
recordings and analysis of R-R tachograms in the supine
position. Despite an immense effort, uniform norms enabling
valid assessment and monitoring of cardiovascular autonomic
functions were not dened.
In the Czech Republic, trials dealing with normative short-
term SAHRV data in the entire age spectrum include works of
Stejskal et al. [8] from 2002 (216 probands aged 1270 years) and
Vlkov et al. [9] from 2010 (167 healthy probands aged 2080
years). Methodological work of Howorka et al. (1998) [10],

Please cite this article in press as: R. Metelka et al., Heart rate variability evaluation in the assessment of cardiac autonomic neuropathy in
patients with type 2 diabetes, Cor et Vasa (2017), http://dx.doi.org/10.1016/j.crvasa.2017.05.001
CRVASA-499; No. of Pages 10

cor et vasa xxx (2017) e1e10 e3

concentrating on the stratication of CAN severity in 119

diabetics aged 52  9 years, should also be mentioned. Unlike
foreign authors assessing HRV using SAHRV analyzed within a
resting recording of a supine R-R tachogram following a 10 min
of relaxation only, the authors of the above 3 trials respect the
methodology of Salinger et al. (1998) [11] using the modied
orthostatic test (orthoclinostatic test Supine1StandingSu-
pine2, SSS test) according to the Ewing battery of reex tests of
autonomic cardiovascular functions as a standard autonomic
load, enabling the assessment of reactivity of both autonomic
nervous system (ANS) branches sympathetic and vagal.
Orthoclinostatic load is involved during Supine1Standing
Supine2, rst as resetting (Standing) and second as the
vagally provoking maneuver (Supine2), testing sympatho-
vagal and baroreceptor interplay mirrored in the sympathetic
and vagal modulation of heart rate variability (Fig. 1).
Aims
Respecting the guidelines of Czech Diabetes Society concern-
ing the evaluation of DAN (last version from 2016) [12] and the
need to ensure a uniform methodology for diagnostics and
staging of CAN in the Diabetes Center in Olomouc, we used our
previous experience with HRV assessment using SAHRV the
methodology of Salinger et al. [11]. We performed a pilot study
with the following primary endpoint to summarize the real
potential of the above three methods [810], using their
approach and normative data for SAHRV in terms of
orthoclinostatic test as autonomic load. Secondary aim of
the study was to assess the benet of rate-controlled breathing
as another autonomic load. The next aim was to compare the
HRV data measured with the selected clinical and laboratory
indices in patients examined.
Sample, methods
Of the numerous population of diabetics followed at the IIIrd
Internal Medicine Department, University Hospital Olomouc,
we selected 43 patients with type 2 diabetes (12 women, 31
men, mean age 51.1  10.7 years) with no manifest CAN, no
history of severe cardiovascular disease (no manifest heart
disease, atrial brillation or utter, stroke, intermittent
claudication or trophic defects in the lower limbs), no severe
pulmonary disease, systemic autoimmune disorders or severe
manifest psychiatric illness.
Obesity was the most frequent comorbidity (affecting 96%
of patients; BMI < 24.9 = 4.6%, BMI 25.029.9 = 27.9%, BMI 30.0
34.9 = 37.2%, BMI 35.039.9 = 25.6%, BMI > 40.0 = 4.6%), fol-
lowed by hypertension (58%), hyperlipidemia (44%, triglycer-
ide waist being found in 39%). Peripheral diabetic neuropathy
was conrmed in 9 (20%), retinopathy in 4 (9%), and
nephropathy in 4 (9%) patients. Smokers formed 32% of
patients in the series. The basic clinical and laboratory data are
shown in Table 1. Pharmacotherapy consisted of oral
antidiabetics (51%), insulin therapy (44%), combination of
both (20%), diabetic diet only (14%), beta blockers (23%), ACEi
and sartans (53%).
The secondary aim was also to nd within the dataset
some functional dependence of SAHRV indices on certain
clinical and laboratory indicators. In this sense, it was not
necessary to create any control dataset.
Heart rate variability and reex cardiovascular tests-testing
procedure and analysis
Respecting the primary study endpoint, i.e. the evaluation of 3
methods [810] differing in postprocessing data analysis, we

Fig. 1 Spectral analysis of heart rate variability in a healthy person (50 years) during the Supine1StandingSupine2 test.
Legend: X axis . . . frequency (Hz). Analyzed range 0.020.5 Hz according to Salinger et al., 1998 [11]. VLF (very low frequency,
0.020.05 Hz), LF (low frequency, 0.050.15 Hz), HF (high frequency, 0.150.5 Hz). Y axis . . . spectral density, amplitude (PSD);
[ms2 HzS1]. Z axis . . . time (s), Supine1 (not evaluated), Standing (evaluated phase), Supine2 (evaluated phase). The spectral
performance of each component (power; [ms2]) is represented by an area delineated by amplitude (Y axis) and frequency
range of the component (X axis).

Please cite this article in press as: R. Metelka et al., Heart rate variability evaluation in the assessment of cardiac autonomic neuropathy in
patients with type 2 diabetes, Cor et Vasa (2017), http://dx.doi.org/10.1016/j.crvasa.2017.05.001
CRVASA-499; No. of Pages 10

e4 cor et vasa xxx (2017) e1e10

Table 1 Basic clinical and laboratory data concerning the series of patients with type 2 diabetes.
Mean SD Median Max Min
Age (years) 51.1 10.7 51 75 25
BMI [kg/m2] 32.2 4.3 32.5 40.4 24.8
Waist circumference [cm] 110.1 10.2 107 130 90
Diabetes duration (years) 5.2 6.7 2 32 0
Glycated hemoglobin [mmol/mol] 58.6 23.0 49 118 29
Total cholesterol [mmol/l] 4.57 1.37 4.56 1.07 8.54
TG [mmol/l] 2.47 2.52 1.62 0.60 13.54
HDLc [mmol/l] 1.11 0.31 1.10 0.54 1.88
LDLc [mmol/l] 2.49 0.87 2.45 0.78 5.85
apoB [g/l] 0.99 0.28 0.97 0.51 1.81
Glomerular ltration [ml/s/1.73 cm2] 1.4 0.2 2 2 1
ACR [mg/mmol/l] 0.9 1.1 0 4 0
PWV [cm/s] 9.4 2.3 8.7 16.1 6.0
Legend: BMI (body mass index), serum levels: glycated hemoglobin, total cholesterol; TG, triglycerides; HDLc, HDL cholesterol; LDLc, LDL
cholesterol; APoB, apolipoprotein B; ACR, urine albumin/creatinine quotient; PWV, pulse wave velocity.

used the method described by Salinger et al. [11] to obtain basal
HRV data in all patients. This method enables to perform both
the basic Ewing battery of reex tests and SAHRV consisting of
the workup of 5 consecutive minute tachograms during
orthoclinostatic SSS test (300 normal R-R intervals) with
classical Fourier analysis. The acquired data were compared
with published clearly dened normatives of SAHRV indices
of each author.
All patients were examined following a light breakfast with
no intake of caffeine or theins, at least 12 h of abstinence from
smoking, and 24 h with no medication affecting the cardio-
vascular system. The evaluation took place between 8 and
11 a.m. at an ambient room temperature (2224 8C), in a calm
room, after instructions about the principle of testing were
given. Resting ECG was obtained, autonomic functions tested,
and echocardiography performed. After ANS examination
taking 4045 min, patients lled a questionnaire concerning
their autonomic functions [12]. The prole of the studied
patient series was completed with laboratory and clinical data
from their last outpatient check-ups: glycated hemoglobin
(GH), glomerular ltration rate (GFR), urine albumin/creatinine
index (ACR) plus lipidogram. Of anthropometric parameters,
BMI, waist circumference, and triglyceride waist (TG waist)
were taken into account. Some of the studied patients
underwent the measurement of pulse wave velocity (PWV)
using the Sfygmocor f.AtCor Medical device on another
occasion.
Autonomic functions were tested using the diagnostic
system DiANS PF8/PF7 produced by Dimea Group, Ltd. [13] with
telemetric transfer of a single-lead ECG signal (thoracic belt)
and respiratory rate to the computer. Using the methodology
of Salinger et al. [11], our patients underwent the following
parts of a testing battery dened by Ewing: deep breathing
while seated (6 cycles/min), orthostatic test (SupineStanding)
after a 5-min pause. Then we continued with tachogram
recording while supine and breathing at a spontaneous rate
after another 5-min pause; this tachogram was obtained for
the purposes of spectral analysis during an orthoclinostatic
test (Supine1StandingSupine2, SSS test). While obtaining the
tachogram for spectral analysis, ECG, respiratory rate and
spectral analysis were monitored on-line. The entire recording
was archived, including ECG, after the elimination of artifacts.
In this test, spectral analysis of tachogram was performed on a
segment following steady-state achievement (1 min) after the
change of body position. The recording segment used for
spectral analysis involved a 256 R-R window with 300 normal
R-R intervals, i.e. some 5 min.
After the SSS test, another tachogram involving 300 normal
R-R intervals was obtained while supine and performing non-
deep, rate-controlled breathing (paced breathing PB,
12 cycles/min) for spectral analysis. The testing was complet-
ed with recording of a tachogram during Valsalva maneuver
while seated. None of the patients suffered from severe
hypertension or advanced diabetic retinopathy, i.e. conditions
that would contraindicate the use of the above maneuver. At
the very end of evaluation, some tests can be repeated
following a short pause.
During the testing, blood pressure was measured to exclude
hypotension, measurements taking place every time before
and after each test plus each 5 min while standing during the
orthoclinostatic test.
The following indices of cardiovascular autonomic function
tests according to Ewing were assessed: I E during deep
breathing test (the difference between mean heart rate values,
HR between inspiration and expiration), I/E (mean inspiratory
and expiratory HR ratio), R-Rmax/R-Rmin in the orthoclinostatic
test (the longest R-R divided by the shortest R-R while
standing), BI (brake index %, BI = (R-Rmax R-Rmin)  100/R-
Rresting), VR index during Valsalva maneuver (Valsalva ratio,
the highest HR value while expiration against resistance
divided by the lowest HR value after termination of the
expiration against resistance) [14]. The acquired values were
compared with available normatives for given age. We chose
the so called Ewing total score as the basis of assessment (ETS,
i.e. number of tests with abnormal ndings) [12].
Spectral analysis of the R-R tachogram (SAHRV) [11] in each
period of the recording in a steady-state (300 normal R-R
intervals) was performed using Fourier transformation analy-
sis in a 256 R-R window in a frequency range 0.020.5 Hz. The
spectral area was divided to the following frequency (spectral)
components: VLF (very low frequency, 0.020.05 Hz), LF (low
frequency, 0.050.15 Hz), and HF (high frequency, 0.150.5 Hz).
The obtained recordings were evaluated both qualitatively
(visually), assessing the dynamics of spectral power in

Please cite this article in press as: R. Metelka et al., Heart rate variability evaluation in the assessment of cardiac autonomic neuropathy in
patients with type 2 diabetes, Cor et Vasa (2017), http://dx.doi.org/10.1016/j.crvasa.2017.05.001
CRVASA-499; No. of Pages 10
cor et vasa xxx (2017) e1e10
individual spectral components within a 3D graph, and
quantitatively, using the spectral power of individual basic
spectral components during each 5-min phase of the
orthoclinostatic test (Supine1StandingSupine2, SSS) and
during rate-controlled, non-deepened breathing (PB, 12 breath
cycles per minute). The following spectral indices were
evaluated during the spectral analysis of the SSS test: total
spectral power in the range 0.020.5 Hz (TP, total power),
spectral powers of the low-frequency (LF) and high-frequency
(HF) components, LF/HF power ratio, and so called normalized
components LFnu and HFnu (relative representation of LF and
HF components within the spectral power after subtraction of
the omitted VLF component (0.020.05 Hz) from the TP. LFnu
was calculated as (PwrLF/Total Pwr PwrVLF)  100 and HFnu
as (PwrHF/Total Pwr PwrVLF)  100. The measured indices of
SAHRV are presented in linear units of the spectral power [ms2]
or after their natural logarithmic transformation (ln ms2).
The other indices assessed were as follows: length of the
R-R interval (R-R [ms]) and mean breathing frequency
(VF, [c/min, Hz 1 ]).
Dynamic changes of the individual basic indices of SAHRV
(TP, LF, HF, LF/HF, R-R, VF) were assessed in the standing
position during SSS test, in the supine position during the last
phase of orthoclinostatic test (Supine2), and during 5 min of
rate-controlled breathing (PB). While standing, the vagal
activity, i.e. the vagal modulation of HR is maximally
suppressed; sympathetic inuence on HR becomes more
prominent together with baroreectoric pressure stimulation.
TP decreases, mostly so in HF and less in LF. After lying down
again (Supine2) and during PB, the vagal modulation of heart
rate becomes the most pronounced this corresponds to
absolute increase or just redistribution of the total spectral
power in HF (HF and HFnu) plus decrease in LF power, LFnu and
LF/HF index [14,15]. The initial supine position (Supine1) is
considered adaptive and without adequate standardization;
this position was thus not evaluated [16]. The VLF index was
not assessed [5].
In post processing analysis, Stejskal et al. [8] start from the
assessment of age-dependent (range 1270 years) SAHRV
indices during the SSS test, forming several complex param-
eters characterizing sympathovagal interaction. These are as
follows index of vagal activity (VA), decreasing with age, and
index of sympathovagal balance (SVB), increasing with age.
The composite of both indices forms a so called total score (TS),
which can be related to the calendar age and enables the
calculation of a representative parameter described as
functional age of autonomic nervous system (ANS). The
assessment of current autonomic cardiovascular functioning
consists of the evaluation of normality of the acquired
complex indices (TS, VA, SVB) with respect to values
corresponding to the calendar age of the individual on a point
scale (from +5 to 5): normal value borderline value
abnormal value of the given index. The nal outcome we used
in our presented work was the difference between the so called
functional age of ANS and calendar age [8]. The normative
data, calculation of the functional age of ANS according
Stejskal et al. [8], are the included in software of diagnostic
system DiANS PF8/PF7.
When assessing the normality of obtained data according
to Vlkov et al. [9], measured values of spectral indices (TP, LF,
Please cite this article in press as: R. Metelka et al., Heart rate variability evaluation in the assessment of cardiac autonomic neuropathy in
patients with type 2 diabetes, Cor et Vasa (2017), http://dx.doi.org/10.1016/j.crvasa.2017.05.001
e5
HF [ms2], LF/HF ratio) for each patient with no statistic
workup were compared with published normatives (TP, LF, HF,
LF/HF) for corresponding age decades. Since the distribution of
data was irregular and great intraindividual variability was
expected, norms represented by median and percentiles were
used. Values between median and 15th percentile were
considered normal for given age. Decrease of the value below
the 15th percentile can be described as possible (early, mild)
CAN, values below the 5th percentile correspond to denite
(severe) CAN.
The third type of analysis followed the methodology of
Howorka et al. [10]. Parameter cumLFHF, dened as the sum
of spectral power of LF and HF components in all 3 positions
during the orthoclinostatic test (LF Supine1 + HF Supine1 + LF
Standing + HF Standing + LF Supine2 + HF Supine2) in linear
values [ms2] was calculated; natural logarithmic transforma-
tion of this sum was then obtained [ln ms2]. To compare the
results of each patient with an age-matched norm, we used the
nomogram created by the authors of the above methodology in
both linear and logarithmic values. Following options are
available: no CAN, early CAN, and severe CAN.
Statistical analysis
Data were analyzed using statistical software IBM SPSS
Statistics, version 22. Mutual dependence of quantitative
parameters was evaluated with Spearman's correlation
analysis. Paired data were compared with Wilcoxon's paired
test. Mutual dependence of quantitative (sex, presence of
peripheral neuropathy, ophthalmologic ndings, renal
neuropathy, treatment with BB, treatment with ACEi,
smoking) and quantitative parameters was assessed and
comparison of selected patient groups in quantitative
parameters was performed using MannWhitney U test.
Fisher's exact test served to compare groups of patients in
qualitative parameters. Normality of data was tested with
ShapiroWilk test. The level of signicance was 0.05 in all
tests.
While evaluating the results within our series of patients,
relatively homogenous with respect to age (age 51.1  10.7
years), no signicant dependence of any evaluated parameter
(indices within Ewing battery, all spectral analysis indices,
results of clinical evaluations e.g. echocardiography) on the
calendar age was detected, making it possible to perform
statistical tests across the entire series (n = 43). Nevertheless,
some parameters in which signicant differences were
detected were also evaluated within decades of age later to
check the general results (4150 years: n = 15, 5160 years:
n = 15).
Results
Cardiovascular reex tests
Based on the analysis of the entire series according to Ewing
total score (ETS) [12], absence of CAN (ETS = 0) was pronounced
in 11.6%, possible CAN (ETS = 1) in 32.6%, and manifest CAN
(ETS = 23) in 55.8% patients. No hypotensive reaction or
manifest orthostatism were seen in any patient.
CRVASA-499; No. of Pages 10

e6 cor et vasa xxx (2017) e1e10

Spectral analysis of HRV-application of different

methodologies

As for SAHRV evaluation according to Stejskal et al. [8], most

probands achieved results signicantly different from their
age norm (Graph 1). The difference between so called
functional and calendar age in the entire series was highly
signicant (mean 21.8  12.9 years, median 23.5 years,
p < 0.0001). The author of this work considered the above
results fully in line with his subjective qualitative visual
evaluation of dynamics concerning individual parameters in
the SSS test in a 3D graph. Graph comparing the functional and
calendar age clearly shows a signicant representation of
more severe CAN. Descriptive characteristics of individual
spectral indices (TP, LF, HF, LFnu, HFnu, LF/HF, VF, R-R)
concerning the most numerous group of patients aged 4150
and 5160 years are summarized in Table 2.
Rate-controlled, non-deepened breathing (PB, f = 12 cycles/
Graph 1 Comparison of functional age of ANS and calendar
min) led, as expected, to even greater change on respiration-
age according to Stejskal et al. [8].
dependent, vagally modulated HF component than that
observed under orthoclinostatic load in Supine2 during SSS
test. Within the entire group, signicant difference in spectral
indices LF ( p < 0.01), HF ( p < 0.0005), LF/HF ( p < 0.0004), LFnu

Table 2 Spectral indices of SAHRV concerning patients aged 4160 years during Supine1StandingSupine2 test and
during rate-controlled ventilation.
Standing Age Supine2 Age Ventilation Age

4150 y 5160 y 4150 y 5160 y 4150 y 5160 y

2
TP linear [ms ] Mean 256.2 258.8 TP Mean 483.2 297.7 TP Mean 394.9 358.2
SD 177.3 333.7 SD 479.3 183.7 SD 318.2 243.4
Median 197.0 116.9 Median 318.5 293.7 Median 278.3 302.5

LF [ms2] Mean 105.5 148.2 LF Mean 158.0 82.8 LF Mean 94.0 71.6
SD 82.8 302.4 SD 124.9 41.4 SD 71.0 47.7
Median 93.0 52.9 Median 98.0 78.7 Median 88.3 59.5
Minimum 8.9 8.4 Minimum 19.6 8.4 Minimum 23.5 8.6
Maximum 304.3 1263.8 Maximum 384.2 156.7 Maximum 262.9 156.3

HF [ms2] Mean 54.9 24.9 HF Mean 197.6 112.3 HF Mean 264.4 167.1
SD 54.4 25.9 SD 371.4 121.4 SD 252.7 195.9
Median 36.1 14.5 Median 85.0 62.8 Median 185.9 97.3

LF/HF Mean 2.74 5.62 LF/HF Mean 1.50 1.72 LF/HF Mean 0.76 1.10
SD 2.18 4.40 SD 1.27 1.89 SD 0.83 1.53
Median 2.20 4.11 Median 1.18 1.06 Median 0.37 0.72

LFnu Mean 65.2 81.1 LFnu Mean 51.5 51.9 LFnu Mean 34.0 40.5
SD 16.4 25.1 SD 20.2 20.3 SD 22.0 20.7
Median 68.5 81.2 Median 53.9 51.3 Median 26.6 40.4

HFnu Mean 34.7 23.0 HFnu Mean 48.5 48.1 HFnu Mean 66.0 58.8
SD 16.4 14.8 SD 20.2 20.3 SD 21.9 20.6
Median 31.5 18.8 Median 46.1 48.7 Median 73.4 55.0

Respiratory Mean 0.24 0.28 Respiratory Mean 0.25 0.24 Respiratory Mean 0.20 0.20
rate Hz 1 SD 0.08 0.05 rate Hz 1 SD 0.05 0.05 rate Hz 1 SD 0.00 0.00
Median 0.26 0.27 Median 0.24 0.24 Median 0.20 0.20
Legend: Analyzed in age groups 4150 years, n = 15, and 5160 years, n = 15. Test: SupineStandingSupine2: descriptive characteristics during
Standing and Supine2 are given. Ventilation test: during 5 min of rate-controlled, non-deepened ventilation 12 cycles/min in the supine
position (0.2 Hz 1). Indices concerning spectral analysis of heart rate variability (SAHRV) during spontaneous ventilation: [ms2], TP (total power,
total spectral performance within the range 0.020.5 Hz), low-frequency component (LF, 0.050.15 Hz), high-frequency component (HF, 0.15
0.5 Hz), LF/HF ratio and so called normalized components LFnu and HFnu describing the relative representation of LF and HF components in the
total performance after subtraction of the VLF component (0.020.05 Hz), which was not analyzed. Respiratory rate in the given position [Hz 1].

Please cite this article in press as: R. Metelka et al., Heart rate variability evaluation in the assessment of cardiac autonomic neuropathy in
patients with type 2 diabetes, Cor et Vasa (2017), http://dx.doi.org/10.1016/j.crvasa.2017.05.001
CRVASA-499; No. of Pages 10

cor et vasa xxx (2017) e1e10 e7

( p < 0.0004) and HFnu ( p < 0.001) was found, showing a more
pronounced absolute and also relative increase in HF and
lowering of LF during breathing (PB) than following orthocli-
nostasis, Supine2. During breathing (PB) HF component
dominated in the increase of total spectral power (TP),
although TP increase was not signicant. If the HF median
in Supine2 was at least twice that seen while standing, PB
added another, approximately twofold signicant increase in
the HF spectral power comparing Supine2 (see above,
p < 0.0005)! This reserve of vagal HR modulation was found
in 45% patients (n = 16), who underwent the SSS test and then
the PB test (n = 37). Descriptive characteristics of this feature
are summarized in Table 2.
The evaluation of SAHRV according to Vlkov et al. [9],
using non-statistic comparison between values of spectral
analysis indices measured in individual patients and pub-
lished medians and percentiles for appropriate age norms, has
failed when performed simultaneously for all spectral param-
eters (TP, LF, HF, LF/HF). Many patients could not have been
ranked because their values ranged across multiple diagnostic
strata according to percentiles. Evaluation was thus performed
for 2 key parameters, i.e. for TP (total spectral power) and HF
(high-frequency component), just with respect to Supine2
phase. Based on the TP-Supine2 and HF-Supine2 values, as
many as 51% of patients met the parameters corresponding to
population norm, i.e. TP and HF values between the median
and 15th percentile; 28% of patients suffered from mild CAN
(between 15th and 5th percentile), and 21% from severe CAN
(less than 5th percentile). The median, 15th and 5th percentile
values published by Vlkov et al. [9] are shown in Table 3.
Using spectral analysis with CAN severity assessment
based on the cumulative SAHRV index cumLFHF according
to Howorka et al. [10], only 9.3% patients were CAN-free, 72%
suffered from its mild (early) form and 18.6% from its severe
form.
Table 3 Normative data after orthoclinostasis, SAHRV
according to Vlkov et al. [9].
Supine2 position Normative data SAHRV, according
to Vlkov et al. [9], Supine1
StandingSupine2 test
Age 4150 y 5160 y
2
TP [ms ] Median 574
15th percentile 147
5th percentile 95
LF [ms2] Median 315
15th percentile 59
5th percentile 45
Correlation analysis Ewing tests and SAHRV
While calculating the mutual correlations, moderate correla-
tion was found between Ewing total score (ETS) and individual
SAHRV indices expressed in both linear [ms2] and logarithmic
[ln ms2] units during orthoclinostasis (Supine2, [ms2]: TP r =
0.4, p < 0.006, LF: r = 0.31, p < 0.04, HF: r = 0.45, p < 0.003)
and during controlled breathing (TP, [ms2]: r = 0.56, p < 0.0001,
LF: r = 0.38, p < 0.018, HF: r = 0.52, p < 0.001).
Moderate correlation was found between ETS and HRV
assessment using the complex parameter functional age of
ANS (r = 0.37, p < 0.015); the same was true for ETS and HRV
assessment using the cumulative power cumLFHF expressed
in linear [ms2] or logarithmic [ln ms2] units (r = 0.46,
p < 0.002).
Clinical indices
No signicant correlation was found between ETS or SAHRV
indices during orthoclinostatic or rate controlled breathing
test on one hand and the following clinical or laboratory
parameters on the other: known type 2 diabetes mellitus
duration, BMI, waist circumference, triglyceride waist,
presence of retinopathy, peripheral neuropathy or nefropathy
(according to GFR, ACR) or QTc length on ECG.
No signicant correlations were found between SAHRV
parameters on one hand and smoking, treatment with beta
blockers or treatment with ACE inhibitors on the other.
Among the echocardiographic parameters, signicantly
higher prevalence of diastolic dysfunction (with an absolute
preponderance of mild impairment) was detected in patients
with lowered HRV according to the basic indices TP, LF
( p < 0.01 to p < 0.009) during Supine2, according to the complex
index functional ANS age ( p < 0.03) and even according to the
cumulative parameter cumLFHF concerning the SSS test
( p < 0.04). No signicant association was found between SAHRV
parameters and other echocardiographic parameters (global
longitudinal and circular strain, epicardial fat amount).
The only stabile and highly signicant association was
represented by mild to moderate correlation (r = 0.440.53,
signicance 0.0010.02) between the glycated hemoglobin (GH)
value and basic spectral SAHRV indices during orthoclinos-
tasis for Supine2 (TP: r = 0.47, p < 0.001; LF: r = 0.38, p < 0.01;
HF: r = 0.43, p < 0.004), during rate-controlled breathing (PB;
TP: r = 0.46, p < 0.004; LF: r = 0.36, p < 0.02; HF: r = 0.53,
289
p < 0.001). GH value correlates with parameter functional age
102
of ANS (r = 0.51, p < 0.0004) or cumulative power cumLFHF
60
(r = 0.44, p < 0.003) as well.
126
38
Rate-controlled breathing
24

HF [ms2] Median 209 131 Concerning the rate-controlled breathing (PB) test (n = 37), we
15th percentile 71 35
took the group of patients who underwent both the orthocli-
5th percentile 34 19
nostatic and PB tests and divided it into two according to
LF/HF Median 1.57 0.78 whether they showed a marked increase in the HF component
15th percentile 0.63 0.30
(for which we chose a so called working title vagal reserve)
5th percentile 0.15 0.16
during the PB test, the cut-off criterion being HF-PB increase by
Legend: SAHRV, spectral analysis of heart rate variability; TP, total more than 100% compared to HF-Supine2. Patients with a
spectral performance; LF, low-frequency component; HF, high-
demonstrable PB vagal reserve had compensated their
frequency component. Units [ms2].
diabetes better with respect to glycated hemoglobin (GH).

Please cite this article in press as: R. Metelka et al., Heart rate variability evaluation in the assessment of cardiac autonomic neuropathy in
patients with type 2 diabetes, Cor et Vasa (2017), http://dx.doi.org/10.1016/j.crvasa.2017.05.001
CRVASA-499; No. of Pages 10
e8 cor et vasa xxx (2017) e1e10
Median of GH with present vagal reserve (n = 16) during PB was
46 mmol/mol; without this reserve (n = 21) it was 56 mmol/
mol. In spite of small number of tested patients, there was no
signicant difference in GH values. As for other clinical
characteristics, no other difference was found using PB in
the context of SAHRV.
Discussion
Methods
The problem related to absence of age-stratied normative
data has been present since the earliest history of SAHRV
measurement [7]. It is clear that despite progress in methodo-
logical approaches and possibilities of computer-assisted data
analysis enabling easier completion of tests and faster interpre-
tation of acquired data, it is necessary to stick meticulously to
standardized test conditions and be aware of the fact that
interpretation of individual HRV indices requires both methodo-
logical and clinical experience. Complete testing, i.e. the
completion of the entire Ewing test battery and spectral analysis
at the same time takes some 40 min. Ewing tests required good
patient's cooperation. Nevertheless, from the study results, it
was evident that both approaches to ANS examination (Ewing's
battery of tests, SAHRV) are not competitive in the end, each
providing valuable and complementary information about
cardiovascular autonomic reactivity.
The results of the presented work, showing signicant
dynamic changes of the spectral power, conrmed the
unequivocal usefulness of SAHRV testing during autonomic
load (Supine1StandingSupine2, SSS test). This methodology
also allows better standardization of testing than sole
relaxation before a single recording in the supine position.
Standing up within the SSS test is equivalent to resetting of
most contaminating inuences and, moreover, the SSS test
makes it possible to perform spectral analysis following a clear
vagal stimulation, i.e. clinostasis (repeated supine, Supine2).
In line with observations of other authors [17,18], the
presented work pointed out a rarely exploited possibility to
potentiate autonomic load by pronounced vagal stimulation
using rate-controlled, non-deepened breathing (paced breath-
ing PB, 12 breathing cycles per minute). The advantage and
pre-requisite of standard performance of PB makes possible to
acquire tachogram for spectral analysis containing the
necessary 300 R-R intervals without causing any hemodynam-
ic or respiratory alteration of the patient's state.
Demonstration of signicantly higher values of HF-PB than
HF-Supine2 illustrates the real possibility to detect a vagal
reserve even in situations where classical evaluation of SAHRV
after repeated Supine2 during the orthoclinostatic SSS test shows
a minimal vagal modulation of HR shown in the HF area.
The effect of PB or changes in tidal volume on SAHRV or the
use PB to distinguish merged HF from LF component in
probands with low respiratory rate (below 10 c/min) was
already mentioned [6,7,17,18]. However, the inclusion of
standardized PB (5 min, 12 c/min) as a part of the diagnosis
and quantication of CAN grade using SAHRV (including
detection of so called vagal reserve) has not been published
in the literature so far.
Please cite this article in press as: R. Metelka et al., Heart rate variability evaluation in the assessment of cardiac autonomic neuropathy in
patients with type 2 diabetes, Cor et Vasa (2017), http://dx.doi.org/10.1016/j.crvasa.2017.05.001
Normative data
While testing HRV using Ewing reex tests and SAHRV, we
conrmed an excellent usability of the methodology dened
by Salinger et al. [11].
The use of index functional age of ANS or of the
difference between this index and calendar age of a patient
tested according to Stejskal et al. [8] has been successful in the
view of the authors. It offers a quality instrument enabling to
detect even small changes in the SAHRV parameters,
applicable especially during longitudinal follow-up of patients
and after therapeutic interventions. Other outcome parame-
ters dened in this methodology (TS, VA, SVB) were not used in
our work because of the small series of patients studied. Their
analysis is planned in future, after the study is extended to
involve prospective follow-up and therapeutic interventions.
We were not successful in applying available normative
data published by Vlkov et al. [9] (median and percentile
rating) for the same processing SAHRV method by Salinger
et al. [11], especially because of too a wide range of the norm
and too clouded border between possible and denite CAN
(Table 3).
Assessment of HRV using the calculation of cumulative
power of LF + HF components in all SSS test phases (cumLFHF)
according to the stratication normogram provided by
Howorka et al. [10] seems to be a robust and quick screening
test for CAN. The validity of this method was conrmed in the
presented work by a moderate correlation between cumLFHF
and the results of Ewing tests (ETS, r = 0.46, p < 0.002). The
drawback of cumulative index cumLFHF use can be described
as its lower potential with respect to longitudinal follow-up,
allowing for quantitative evaluation total HRV changes only,
not for quantication of changes concerning each ANS branch
(sympathetic or vagal).
Patient selection
It is known that CAN accompanies a wide range of diseases
[19]. The purpose of the study was primarily methodological,
i.e. to increase the uniformity of CAN evaluation in routine
practice, with the possibility of quantication of changes in
CAN over time, regardless of its cause. Therefore, the selection
of patients was concentrated on the group with the highest
probability of the subclinical form of CAN, best represented by
diabetes mellitus type 2 as integral part of the metabolic
syndrome. The inclusion of co-morbidities, also affecting the
VSF, was therefore not inconsistent with the study targets.
Only patients with severe structural heart disease were excluded.
Methodologically and ethically, examination conditions (e.g.
drug withdrawal) were modied so that the diagnostic procedure
was standardized while not substantially disrupting the treat-
ment and health status of the examined patients.
Clinical aspects of the study
Under the given conditions, no signicant correlation between
SAVSF parameters on the one hand and co-morbidities,
smoking, beta-blocker therapy, or ACE inhibitors on the other
hand has been demonstrated to allow the study to meet the
secondary study targets. The presented cross-sectional work
CRVASA-499; No. of Pages 10
cor et vasa xxx (2017) e1e10
on a small sample of type 2 diabetics showed signicant
autonomic dysfunction in most of them, without dependence
on duration of diabetes. Clearly, it supports the recommenda-
tion that assessment of ANS activity in diabetes type 2 should
be made already at the diagnosis of disease within the initial
staging of the disease [12]. The observed positive correlation
between SAHRV evaluated using SSS test and PB (or CAN
grade) on one hand and glycated hemoglobin value on the
other should be highlighted here. This means that a
signicantly better HRV (partial CAN regression) is consistent
with a better compensation of diabetes according to GH. This
fact points to the concurrence of metabolic (diabetes) and
functional (ANS) disease history with the need to monitor both
in parallel.
Conclusions
1. Authors conrm an excellent usability of the HRV assess-
ment by SAHRV in short ECG recordings by the methodology
of Salinger et al. [11] under conditions of autonomic load
using the orthoclinostatic Supine1StandingSupine2 test
(SSS test), as well as the usability of this methodology for the
performance of cardiovascular reex tests (Ewing's battery).
They clearly conrm the positive effect and necessity of
autonomic load during short-term SAHRV assessment in
order to gain better information about autonomic cardio-
vascular reactivity of ANS and to have a standardizing
feature. The basic type of load for SAHRV is represented by
orthoclinostatic test consisting of Supine1StandingSu-
pine2 phase. Further vagal provocation by rate-controlled,
non-deepened breathing (12 cycles/min) can be seen as
facultative load option increasing the usability of short-
term SAHRV.
2. Analyzing the potential of methods available in the Czech
Republic for the evaluation of cardiovascular autonomic
reactivity (ANS), an optimal methodological variant en-
abling quantication of pathology/normality of HRV was
found in using short-term SAHRV during orthoclinostatic
SSS test with calculation a complex indices, such as so-
called functional age of ANS according to Stejskal et al. [8].
3. Evaluation of the presence and severity of CAN with a
cumulative index describing total HRV according to
Howorka [10] showed good discrimination ability within
the initial screening for CAN, albeit without the possibility
to distinguish between the impairment of the sympathetic
and vagal branch of ANS. The usability of this method in
prospective follow-up will have to be conrmed during a
longer prospective study.
4. The presented cross-sectional work on a group of type 2
diabetics demonstrated signicant autonomic dysfunction
in most of them, without dependence on duration of
diabetes. Clearly, it supports the recommendation that
assessment of ANS reactivity in type 2 diabetes should be
made already at the diagnosis of disease within the initial
staging. The CAN grade correlates well with metabolic
compensation of diabetes.
Since the main goals of presented study were achieved, it is
conceivable to switch from cross-sectional to a prospective
Please cite this article in press as: R. Metelka et al., Heart rate variability evaluation in the assessment of cardiac autonomic neuropathy in
patients with type 2 diabetes, Cor et Vasa (2017), http://dx.doi.org/10.1016/j.crvasa.2017.05.001
e9
study and to test the hypothesis that CAN is partially
reversible by therapeutic interventions.
Limitation of study
Although the main objective of the work (selection of the
appropriate methodology) has been met, the basic patient
sample has to be extended several times in order to verify the
suitability of used methodology for longitudinal follow up and
assessing therapeutic interventions.
It can be assumed that some of the predicted dependencies
between the HRV and the investigated clinical parameters (e.g.
echocardiographic, pulse wave velocity) are not conclusive due
to small dataset size; they will be studied further.
Conict of interest
Author state that they have no competing interests.
Ethical statement
The study was conducted in line with ethical standards
dened in the Helsinki declaration.
Informed consent
All patients signed an informed consent prior to being
included in the study.
Funding body
Supported by Ministry of Health, Czech Republic - conceptual
development of research organization (Faculty Hospital
Olomouc, FNOL-00098892).
references
[1] V. Spallone, D. Ziegler, R. Freeman, et al., Toronto
Consensus Panel on Diabetic Neuropathy, Cardiovascular
autonomic neuropathy in diabetes: clinical impact,
assessment, diagnosis, and management, Diabetes/
Metabolism Research and Reviews 27 (7) (2011) 639653.
[2] A.I. Vinik, T. Erbas, C.M. Casellini, Diabetic cardiac
autonomic neuropathy, inammation and cardiovascular
disease, Journal of Diabetes Investigation 4 (1) (2013) 418.
[3] G. Dimitropoulos, A.A. Tahrani, M.J. Stevens, Cardiac
autonomic neuropathy in patients with diabetes mellitus,
World Journal of Diabetes 5 (1) (2014) 1739.
[4] D.J. Ewing, C.N. Martyn, R.J. Young, et al., The value of
cardiovascular autonomic function tests: 10 years
experience in diabetes, Diabetes Care 8 (5) (1985) 491498.
[5] Task Force of the European Society of Cardiology and the
North American Society of Pacing and Electrophysiology,
Heart rate variability. Standards of measurement,
physiological interpretation and clinical use. Special report,
Circulation 93 (1996) 10431065.
CRVASA-499; No. of Pages 10
e10 cor et vasa xxx (2017) e1e10
[6] M.W. Agelink, R. Malessa, B. Baumann, et al., Standardized
tests of heart rate variability: normal ranges obtained from
309 healthy humans, and effects of age, gender, and heart
rate, Clinical Autonomic Research 11 (2) (2001) 99108.
[7] D. Nunan, G.R. Sandercock, D.A. Brodie, A quantitative
systematic review of normal values for short-term heart
rate variability in healthy adults, Pacing and Clinical
Electrophysiology 33 (2010) 14071417.
[8] P. Stejskal, R. lachta, M. Elfmark, et al., Spectral analysis of
heart rate variability: new evaluation method, Acta
Universitatis Palackianae Olomucencis Gymnica 32 (2)
(2002) 1318.
[9] E. Vlkov, J. Bednak, . Burov, et al., Spektrln analza [15]
variability srden frekvence normativn data, esk a
Slovensk Neurologie a Neurochirurgie 73/106 (6) (2010)
663672.
[10] K. Howorka, J. Pumprla, A. Schabmann, Optimal parameters
of short-term heart rate spectrogram for routine evaluation
of diabetic cardiovascular autonomic neuropathy, Journal
of the Autonomic Nervous System 69 (23) (1998) 164172.
[11] J. Salinger, J. Opavsk, P. Stejskal, et al., The evaluation of
heart rate variability in physical exercise by using the
telemetric VariaPulse TF 3 system, Acta Universitatis
Palackianae Olomucensis. Gymnica 28 (1998) 1323.
[12] S. Lacigov, Z. Ruav, A. Jirkovsk, et al., Doporuen
postup diagnostiky a lby diabetick neuropatie (2016),
(Recommendation for the diagnosis and treatment of
diabetic neuropathy. Standards of Czech diabetes
association (2016)), DMEV-Diabetes, Metabolismus,
Please cite this article in press as: R. Metelka et al., Heart rate variability evaluation in the assessment of cardiac autonomic neuropathy in
patients with type 2 diabetes, Cor et Vasa (2017), http://dx.doi.org/10.1016/j.crvasa.2017.05.001
Endokrinologie, Viva-asopis pro postgraduln
vzdlvn 19 (2) (2016) 5763.
[13] J. Salinger, M. Gwozdziewicz, Systmy pouvan pro
vyeten krtkodob variability srden frekvence (Systems
for assessment of short-term heart rate variability in: K.
Javorka (Ed.), Variabilita frekvencie srdca: Mechanismy,
hodnotenie, klinick vyuitie (Heart Rate Variability:
Mechanism, Assessment, Clinical Application), OSVETA,
Martin, 2008, 5760, ISBN: 978-80-8063-269-4.
[14] J. Opavsk, Autonomn nervov systm a diabetick
autonomn neuropatie. Klinick aspekty a diagnostika,
Galn, Praha, 2002, , ISBN: 80-7262-194-7.
P. Stejskal, J. Salinger, Spektrln analza srden frekvence,
Medicina Sportiva Bohemica & Slovaca 2 (1996) 3342.
[16] P. Stejskal, R. lachta, M. Elfmark, et al., The effect of age on
short-term heart rate variability, Acta Universitatis
Palackianae Olomucensis. Gymnica 29 (1999) 717.
[17] E. ujov, P. Stejskal, A. Jakubec, et al., Respiration
frequency and spectral analysis of heart rate variability,
Acta Universitatis Palackianae Olomucensis. Gymnica 34
(2004) 4347.
[18] T.E. Brown, L.A. Beightol, J. Koh, et al., Important inuence
of respiration on human R-R interval power spectra in
largely ignored, Journal of Applied Physiology 75 (1993)
23102317.
[19] R. Freeman, Cardiac autonomic syndromes, in: M. Malik
(Ed.), Clinical Guide to Cardiac Autonomic Tests, Kluwer
Academic publishers, The Netherlands, 1998, 357391,
ISBN: 0-7923-5178-9.