CARDIOVASCULAR JOURNAL OF AFRICA • Vol 23, No 2, March 2012
AFRICA
69
late percentage body fat using the Jackson and Pollock equa-
tions.
23,24
Body mass index (BMI) [mass (kg)/height (m
2
)] and
waist-to-hip ratio (waist:hip circumferences) were calculated and
the HRV measurement protocol was then followed.
Heart rate variability measurement
The participants were fitted with the Suunto t6 heart rate monitor
(HRM) (Suunto; Vantaa, Finland). The electrodes on the trans-
mitter were wet with water and were placed on the chest against
bare skin to ensure good skin contact. Participants were tested
while lying supine with the total testing time lasting 20 minutes.
This time was divided into 15 minutes of rest followed by a five-
minute measurement of IBIs. The IBIs were then transferred to a
laptop (HP ProBook) computer where the data were stored in the
Suunto team manager software program (Firstbeat Technologies,
Ltd; Jyvaskyla , Finland).
The data were then exported as a text file to the HRV analy-
sis software (Kubios heart rate variability software version 2.0;
Biosignal Analysis and Medical Imaging Group, Department of
Physics, University of Kuopio, Kuopio, Finland) for analysis of
the following HRV parameters (1) time domain: resting heart
rate (RHR), R–R intervals (IBI), standard deviation of normal-
to-normal intervals (SDNN), mean square root differences
of the standard deviation (RMSSD), percentage of beats that
changed more than 50 ms from the previous beat (pNN50); and
(2) frequency domain: low-frequency normalised units (LFnu),
high-frequency normalised units (HFnu), low-frequency to high-
frequency ratio in normalised units (LF/HFnu).
The Suunto t6 HRM and Kubios program
20
comply with
guidelines recommended by the Taskforce of the European
Society of Cardiology and the North American Society of
Pacing and Electrophysiology standards for measurement of
HRV.
4
Before processing, the IBIs were manually corrected for
ectopic/missed beats. There is currently no universal method for
identifying and editing ectopic beats. The amount and type of
editing of IBI data has different effects on various HRV indices.
25
In the present study, manual editing or interpolation
25
of the IBI
intervals was performed using the following guidelines: if a
significantly higher IBI (representing an ectopic beat) was noted,
then that reading was deleted and the average of the two adjacent
IBIs replaced the deleted one. If a significantly lower value
(representing a missed beat) was noted, that IBI was deleted and
replaced with the previous IBI. If the ectopic or missed beats
exceeded 20% of a participant’s overall five-minute recording,
the participant was not included in the analysis.
2
This occurred
in six participants (four males and two females), with the result
that only 44 participants were included in the final analysis (
n
=
21 males and
n
=
23 females).
Once the IBIs were imported into the Kubios program, the
software automatically analysed the HRV in both time and
frequency domains. Power spectral analysis was performed using
the autoregressive (AR) algorithm in accordance with the recom-
mendations.
4
The AR algorithm was used as it yields improved
resolution, especially for short-term HRV measurements.
20,21
This algorithm creates a power spectral analysis with distinct
frequency bands, namely high frequency (HF), low frequency
(LF) and very low frequency (VLF).
The LF component has been proposed as reflecting both
sympathetic and parasympathetic effects on the heart and occurs
in a band between 0.04 and 0.15 Hz. However, researchers noted
that the low-frequency band is influenced by baroreceptor-
mediated regulation of blood pressure and reflects predominantly
sympathetic activity.
26
The HF (0.15–0.4 Hz) band corresponds
with respiratory sinus arrhythmia (RSA) and is said to reflect
parasympathetic activity.
26
Chemoreceptor processes, thermoreg-
ulation, and the renin–angiotensin system have been linked with
the VLF band. We did not use data collected from the VLF range
for this study.
Statistical analysis
The data were summarised using routine descriptive statistics
(mean
±
SD). Reliability measures were determined for day
2 versus day 3 and day 3 versus day 4. Absolute and relative
reliability of several HRV parameters was calculated using the
procedures described by Hopkins.
27
Hopkins has argued that
the statistical analysis used in reliability studies should include
observed values and confidence limits of the typical error. These
measures are sufficient to characterise the reliability of a meas-
ure and they substantially enhance comparison of the reliability
of tests, assays or equipment.
27
Absolute reliability is the degree to which repeated measure-
ments vary for individuals. This type of reliability is expressed
either in the actual units of measurement or as a proportion of
the measured values (dimensionless ratio).
28
Absolute reliability
was calculated using the typical error of measurement (TEM)
and TEM as a percentage (TEM%), expressed as percentage of
the mean score.
27
Sport and exercise science reliability studies have rarely
reported the separate analysis of homoscedastic and hetero-
scedastic data.
28
These parameters show how the measurement
error relates to the magnitude of the measured variable. When
the amount of random error increases as the measured values
increase, the data are said to be heteroscedastic. When there is
no relation between the error and the size of the measured value,
the data are described as homoscedastic.
27
Homoscedastic errors
are expressed in the actual units of measurement (TEM) but
heteroscedastic data are measured on a ratio scale. With homo-
scedastic errors, the raw data are analysed with conventional
parametric analyses, but heteroscedastic data are transformed
logarithmically before analysis or investigated with an analysis
based on ranks.
28
In the present study, the HRV data for each parameter were
examined using the technique described by Hopkins
27
and were
found to be homoscedastic. TEM and TEM% were calculated
with 90% CI for day 2 versus day 3 and for day 3 versus day 4,
using a spreadsheet downloaded from
.
The TEM is also known in statistical terms as the coefficient of
variation (CV).
Furthermore, for reliability studies, it has been suggested
that relative reliability is presented together with absolute reli-
ability.
28
Relative reliability is the degree to which individuals
maintain their position in a sample with repeated measurements
and is represented in the form of correlation coefficients. In this
study we used interclass correlations (ICC) with 95% CI.
27
The
main advantage of ICC over Pearson’s correlation is that it can
be used when more than one retest is being compared with a
test.
28
Various categories of reliability are based on the ICC. An
ICC above 0.8 is usually regarded as good to excellent reliability,
whereas an ICC between 0.6 and 0.8 may be taken to represent
substantial reliability.
3
