Assessment of Cardio-Respiratory Relationship during and after Exercise in Healthy Recreative Male Subjects: A Pilot Study

Featured Application

This is an original article in which both cardiac and respiratory signals were measured simultaneously during exercise under submaximal heart rate along with the determination of predictors in recovery. We strongly believe that these results will have important implications in the area of exercise physiology and sports medicine.

Abstract

Background: Understanding the responses of the cardio and respiratory systems during exercise, as well as their coupling in post-exercise recovery, is important for the prescription of exercise programs in physically recreative subjects. Aim: In this work, we aimed to set up an adjusted experiment to evaluate the relations and changes in parameters obtained from an analysis of cardiac and respiratory signals under three physiological conditions: relaxation, exercise, and post-exercise recovery. Material and Methods: Simultaneously recorded ECG (RR intervals) and respiratory signal during relaxation, bicycle ergometry exercise until submaximal heart rate (HR), and recovery in 10 healthy men were analyzed. The exercise included consecutive phases of 3 min in duration with a constant workload. Parasympathetic cardiac control (RMSSD), heart rate (HR), breathing frequency (BF), and respiratory cycle amplitude (RCA) were calculated. Anthropometric data were also collected. Results: Based on time series analysis, our results show that: (1) during exercise, an increase in HR was related to a reduction in HR variability and RMSSD, while an increase in BF was related to an increase in RCA, and (2) during recovery, HR and RCA significantly decreased, while RMSSD had a biphasic response. The results of multiple linear regressions showed that the averaged HR, RMSSD, and BF during 3 min segments of recovery were determined by several calculated and collected parameters. Conclusions: The parameters from the analysis of respiratory signals and RR interval time series under conditions of relaxation and exercise, along with anthropometric data, contributed to the complexity of the post-exercise recovery of cardiopulmonary parameters after submaximal HR exercise in healthy recreative males.

1. Introduction

The assessment of cardio–respiratory coupling (CRC) in various physiological and pathological conditions has been the focus of many studies over the years. The cardiac and respiratory systems interact in many different forms, one of them being well known as respiratory sinus arrhythmia, in which heart rate changes are related to respiration cycle–RR intervals are shortened during inspiration and lengthened during expiration. Moreover, the complexity of CRC interactions is reflected in the non-linear dynamics between the systems, and these various forms of coupling can coexist at different time scales [1,2]. Nowadays, it is clear that the regulation of cardio–respiratory coupling has consequences, not only on the level of blood gases and maintaining homeostasis, but also on other central functions such as arousal [3].

One of the areas of intensive research in quantifying CRC is the area of sports medicine. Exercise, which, according to Caspersen et al. [4], represents “planned, structured, repetitive, and purposive physical activity”, is an integral component of both sports and recreational activities. It has been shown that physical exercise contributes to increasing the ability of oxygen consumption and to a more economical cardiac response during exercise at submaximal heart rate intensity [5,6,7]. These aforementioned factors can consequently impact improvements in overall human work capacity, particularly through the enhancement of two vital systems whose interaction is not sufficiently researched, especially in exercise and recovery. In general, it is well established that physical activity affects cardio–respiratory coupling. It is shown that highly intensive training leads to an increase in CRC strength. However, it has still not been clarified whether this increase in CRC strength in physical activity has its origin in enhanced mechanics or enhanced neural circuits between the cardiac and respiratory centers in the medulla in trained people [1,8]. Furthermore, inconsistent results are obtained, because various studies employ different protocols and modalities of physical activity, as well as its duration and intensity. Also, body posture in experimental protocols should not be neglected [9].

Little is known about the correlations between cardiac and respiratory parameters in different conditions: at rest, in exercise, and during post-exercise recovery. It is questionable as to what extent the resting heart rate and parasympathetic activity affect changes in both the cardiac and respiratory systems during exercise, i.e., to what extent they do not depend only on ongoing physical activity. The interplay between the heart rate, breathing frequency, and respiratory cycle amplitude is of particular importance. Recent studies have implied the differential control of breathing frequency and tidal volume [10,11]. The former changes in relation to a muscle afferent input, central command, and the level of physical exertion, while the latter is more susceptible to metabolic inputs such as changes in PaCO₂ and metabolic acidosis. Also, there is a difficulty in reproducing “real” exercise conditions in studies, as this is the only way to evaluate the integrative roles of the various compensatory mechanisms which simultaneously influence cardiac and respiratory parameters in physical activity [12].

Furthermore, the post-recovery period is of great importance, as it shows the returning dynamics of parameters to pre-exercise levels. Immediate changes following physical activity are of special interest as the reactivation of parasympathetic activity occurs and sympathetic activity slowly withdraws. Although the recovery period is interpreted as a reverse process in relation to exercise, the particular timeframe and the rate by which all parameters return to baseline values are of importance, as this might be indicative of low autonomous “reserve” in a person or dysautonomia [9,13]. In a healthy population, the timeframe for returning the values of these parameters to pre-exercise levels, as well as the parasympathetic response after exercise, depend on numerous factors, both in terms of chronic exposure to harmful external environmental conditions and experimentally induced acute internal non-pathological states, e.g., in hypoxia through high-altitude simulation [14,15,16]. In addition, it is of particular importance to examine the differences in responses to exercise in terms of the behavior of autonomic heart control centers, as these parameters precisely indicate the interplay of the ANS branches.

In this study, we aim to evaluate the behavior of cardio–respiratory relationship under parasympathetic withdrawal during exercise under submaximal heart rate conditions and following its reactivation in healthy young men. To our knowledge, this is the first study in which the heart and respiratory signal parameters are assessed simultaneously. Finally, we want to determine whether HR, RMSSD, and BF in recovery are dependent on other parameters examined at rest and during exercise.

2. Materials and Methods

2.1. Participants

Ten recreationally active men (mean age = 30.6 ± 2.7 years) participated in our study. We enrolled 12 men, but 2 of them were excluded from the study due to arrythmias. The anthropometric data of the participants are shown in

Table 1. Anthropometric values and body composition measures.

Index	Mean Value ± SD
Height (cm)	184 ± 6
Total body water (L)	55 ± 6
Protein mass (kg)	15 ± 2
Minerals mass (kg)	5.1 ± 0.5
Body fat mass (kg)	20 ± 10
Soft lean mass (kg)	71 ± 8
Fat-free mass (kg)	75 ± 8
Body mass (kg)	100 ± 20
Skeletal muscle mass (kg)	43 ± 5
Body mass index (kg/m2)	28 ± 5
Percent body fat (%)	20 ± 9
Skeletal muscle index (kg/m2)	9.5 ± 0.8

. All subjects filled out a physical activity readiness questionnaire for everyone [17]. The participants were actively engaged in cyclic aerobic physical exercise (brisk walking, running, and cycling) for 75–150 min per week, depending on the intensity (moderate to high intensity: 65–85% of HR_max). At a younger age, they were active athletes in various sports disciplines, but in the last 5 years, none of the participants had been involved in sports activities. The inclusion criteria for the study involvement were to be engaged in recreational physical activity three times per week and the absence of acute or chronic illnesses. At the time of performing the experiment, the participants did not have any cardio–respiratory pathological conditions and they were familiarized with the use of bicycle ergometers.

2.2. Experimental Protocol

The protocol of this study was approved by the Ethics Committee of the Faculty of Medicine, Belgrade University (Ref. Numb. 29/XII-18). The participants were instructed to restrict their intake of food and drinks that included caffeine and alcohol and to avoid physical exercise 24 h before the experiment. The experiment was carried out on three consecutive days between 8 a.m. and 3 p.m., and the measurements were performed in environmentally controlled conditions: (21–23) °C [18,19,20].

The experimental protocol started with the measurement of anthropometric values and body composition in the Laboratory for Sports Medicine and Exercise Therapy at the Institute of Medical Physiology, Faculty of Medicine, University of Belgrade. The participants’ mass was measured with a Seca 214 Portable Stadiometer (Cardinal Health, Columbus, OH, USA). Body composition was analyzed with the InBody 370S (Body Composition Analyzer, Seoul, Republic of Korea).

Further procedures were realized in the Laboratory of Biosignals at the Institute of Biophysics in Medicine, Faculty of Medicine, University of Belgrade. In the first experimental phase, the participants were in a supine relaxed position for 15 min. At the end of the relaxation phase, the participants were sitting for 3 min, due to blood pressure adaptation. After finishing the first phase, the participants were included in the physical exercise phase with submaximal effort. For this phase, a Keiser-M3 (Keiser™, Chicago, IL, USA) spinning bike ergometer was used, which represents an alternative to standard cycle ergometers [21]. In this study, physical exercise is defined according to the modified Young Men’s Christian Association (YMCA) test protocol for submaximal effort, which was modified by adding two additional phases of 3 min in duration with an equivalent increase in resistance (Figure 1,

Table 2. Levels of individually dosed workload in watts (W).

Level	HR < 80 (Beats/min)	HR = 80–89 (Beats/min)	HR = 90–100 (Beats/min)	HR > 100 (Beats/min)
I	25	25	25	25
II	125	100	75	50
III	150	125	100	75
IV	175	150	125	100
V	200	175	150	125
VI	225	200	175	150

) [22].

Our pilot study showed that modified physical exercise was better for provoking submaximal effort in the participants. Physiological effort was assessed by measuring the heart rate (HR). For this experiment, the level of workload was the main external parameter. Physical exercise was realized through six resistance levels (Figure 1 and Table 2). The first level was the warm-up phase (3 min with 25 W intensity). According to the HR in this level, the resistance for the next levels was defined. The cadence was 55–65 revolutions per minute (RPM) in all exercise phases. In level two, the workload was 125 W, and in each subsequent level, the participants saw an increase of 25 W. Reaching the submaximal HR (Equation (2)) determined the end of the physical exercise phase. The submaximal HR values were calculated with a percentage calculation of age-predicted maximal heart rate [23].

HR_max = 220 − age

(1)

HR_submax = (220 − age) × 0.85

(2)

HR_max represents maximal HR and HR_submax represents submaximal HR according to APMHR. Upon finishing the exercise, the participants were positioned in the supine position. This represented the recovery phase of the experiment, in which we continued with signal monitoring for 15 min.

2.3. Data Acquisition

The measured signals of the ECG and respiratory rhythm are shown in Figure 2. The ECG and respiratory data were obtained with a sampling frequency of 1000 Hz and resolution of 16 bits using a Biopac MP100 system and AcqKnowledge 3.9.1 software (BIOPAC system, Inc, Santa Barbara, CA, USA). RR intervals (time intervals between the successive heart beats) and BB intervals (time intervals between the successive inhalations) were determined.

The ECG and respiratory signals were analyzed using an OriginPro 8.6 (OriginLab Corporation, Northampton, MA, USA) software. The time series of the RR intervals and respiratory signals were divided into consecutive segments of 180 s. In each RR interval segment, the average values of RR and parasympathetic activity (RMSSD) were calculated. Similarly, the average values of BB and respiratory cycle amplitude (RCA) were calculated from each 180 s segment of the respiratory signal. The HR and BF were calculated from the average RR and BB values according to the equations:

$$\mathrm{H}\mathrm{R}=\frac{1}{\overline{\mathrm{R}\mathrm{R}}}\times 60(\mathrm{b}\mathrm{e}\mathrm{a}\mathrm{t}\mathrm{s}/\mathrm{m}\mathrm{i}\mathrm{n})$$

(3)

HR—heart rate and $\overline{\mathrm{R}\mathrm{R}}$—the average RR interval value in analyzed segment.

$$\mathrm{B}\mathrm{F}=\frac{1}{\overline{\mathrm{B}\mathrm{B}}}(\mathrm{H}\mathrm{z})$$

(4)

BF—breathing frequency and $\overline{\mathrm{B}\mathrm{B}}$—average BB interval value in analyzed segment.

$$\mathrm{R}\mathrm{M}\mathrm{S}\mathrm{S}\mathrm{D}=\sqrt{\frac{\sum _{i=1}^{N-1}{\left({\mathrm{R}\mathrm{R}}_{i+1}-{\mathrm{R}\mathrm{R}}_i\right)}^2}{N-1}}(\mathrm{m}\mathrm{s})$$

(5)

RMSSD—root mean square of successive RR interval differences, which is used as a noninvasive measure of parasympathetic cardiac control [24].

We had five 180 s segments in the relaxation phase, six in the physical exercise phase, and five segments in the post-exercise recovery phase. To obtain two series with the same number of points, we performed equally equidistant resampling of the RR series using the mean RR value for each subject. In addition, we resampled the respiratory signal according to the mean RR values using linear interpolation between the two adjacent samples. The resampling procedure was conducted as in the work of Kapidžić et al. [25].

2.4. Sample Entropy and Cross-Sample Entropy

Sample entropy (SampEn) is a measure of time series irregularity (unpredictability). It was developed by Richman and Moorman [26] as a refinement of the approximate entropy introduced by Pincus [27]. For some time series, SampEn is defined as the negative natural logarithm of the conditional probability that two sequences, similar for m points, remain similar within tolerance r at the next point (i.e., for m + 1 points), where self-matches are not included:

$$SampEn\left(N,m,r\right)=-\mathrm{l}\mathrm{n}\frac{A^m\left(r\right)}{B^m\left(r\right)}$$

(6)

where B^m is the probability that two sequences will match for m points, while A^m is the probability that two sequences will match for m + 1 points and N is the number of equidistant data points. A time series with a small number of similarities is characterized by large values of sample entropy, which indicates its higher unpredictability (irregularity). Cross-approximate entropy as a measure of asynchrony between two time series was also introduced by Pincus [27]. We used cross-sample entropy (Cross-SampEn) as a measure of the asynchrony between the RR interval and respiratory signal time series [27,28]. This measure is a non-linear measure of time series irregularity derived from the probability of finding a similarity between two signals.

Each equally resampled RR and Resp series was normalized by a standardized procedure; after subtracting the mean value, the demeaned series was divided by the standard deviation of the time series. In this paper, SampEn of RR intervals is denoted by SampEnRR, SampEn of respiratory signal is denoted by SampEnResp, and Cross-SampEnRRResp is the cross-sample entropy between the two signals.

2.5. Statistics

The data were analyzed using the SPSS software package (SPSS version 17, Chicago, IL, USA). The Wilcoxon comparison test for related samples was used to compare the data obtained in each experimental phase and between the experimental phases. The results are presented as mean values +/− standard deviations. The correlation between the variables was quantified by the Pearson correlation coefficients. A stepwise multiple linear regression analysis was applied to find which variables and parameters predicted HR, RMSSD, and BF in the recovery phase of the experiment. Statistical significance was set at p ≤ 0.05.

3. Results

In Figure 3, we present the averaged values of HR, RMSSD, BF, and RCA calculated from the 3 min segments collected during all three phases of the experiment: relaxation, bicycle ergometry exercise, and post-exercise recovery. In the following four tables (

Table 3. Statistical significance of differences between the heart rate (HR) values calculated from 3 min segments in different conditions: Relaxation (Rel), Exercise (Exe), and Recovery (Rec). Statistically significant values are in bold.

HR
	Exe1	Exe2	Exe3	Exe4	Exe5	Exe6	Rec1	Rec2	Rec3	Rec4	Rec5
Rel1	0.005	0.005	0.005	0.005	0.005	0.008	0.005	0.005	0.005	0.005	0.005
Rel2	0.005	0.005	0.005	0.005	0.005	0.008	0.005	0.005	0.005	0.005	0.005
Rel3	0.005	0.005	0.005	0.005	0.005	0.008	0.005	0.005	0.005	0.005	0.005
Rel4	0.005	0.005	0.005	0.005	0.005	0.008	0.005	0.005	0.005	0.005	0.005
Rel5	0.008	0.008	0.008	0.008	0.012	0.008	0.008	0.008	0.008	0.008	0.008
Exe1		0.005	0.005	0.005	0.005	0.008	0.285	0.241	0.074	0.009	0.005
Exe2			0.005	0.005	0.005	0.008	0.005	0.005	0.005	0.005	0.005
Exe3				0.005	0.005	0.008	0.005	0.005	0.005	0.005	0.005
Exe4					0.007	0.008	0.005	0.005	0.005	0.005	0.005
Exe5						0.008	0.005	0.005	0.005	0.005	0.005
Exe6							0.008	0.008	0.008	0.008	0.008
Rec1								0.007	0.013	0.013	0.009
Rec2									0.114	0.047	0.028
Rec3										0.013	0.007
Rec4											0.022

,

Table 4. Statistical significance of differences between the parasympathetic activity (RMSSD) values calculated from 3 min segments in different conditions: Relaxation (Rel), Exercise (Exe), and Recovery (Rec). Statistically significant values are in bold.

RMSSD
	Exe1	Exe2	Exe3	Exe4	Exe5	Exe6	Rec1	Rec2	Rec3	Rec4	Rec5
Rel1	0.009	0.008	0.008	0.008	0.005	0.008	0.878	0.005	0.005	0.005	0.114
Rel2	0.022	0.008	0.008	0.008	0.005	0.008	0.799	0.005	0.005	0.005	0.037
Rel3	0.037	0.008	0.008	0.008	0.005	0.008	0.878	0.005	0.005	0.005	0.009
Rel4	0.013	0.008	0.008	0.008	0.005	0.008	0.386	0.005	0.005	0.005	0.017
Rel5	0.038	0.012	0.012	0.012	0.008	0.012	0.260	0.008	0.008	0.008	0.008
Exe1		0.011	0.008	0.008	0.005	0.008	0.047	0.028	0.047	0.074	0.878
Exe2			0.011	0.008	0.008	0.017	0.011	0.374	0.314	0.110	0.051
Exe3				0.515	0.086	0.161	0.008	0.038	0.011	0.011	0.008
Exe4					0.110	0.674	0.008	0.038	0.011	0.011	0.008
Exe5						0.859	0.005	0.028	0.009	0.005	0.005
Exe6							0.008	0.038	0.015	0.008	0.008
Rec1								0.017	0.028	0.037	0.241
Rec2									0.386	0.037	0.007
Rec3										0.007	0.005
Rec4											0.059

,

Table 5. Statistical significance of differences between the breathing frequency (BF) values calculated from 3 min segments in different conditions: Relaxation (Rel), Exercise (Exe), and Recovery (Rec). Statistically significant values are in bold.

BF
	Exe1	Exe2	Exe3	Exe4	Exe5	Exe6	Rec1	Rec2	Rec3	Rec4	Rec5
Rel1	0.036	0.015	0.005	0.005	0.005	0.008	0.508	0.878	0.126	0.005	0.721
Rel2	0.047	0.005	0.005	0.005	0.005	0.008	0.359	0.221	0.017	0.005	0.241
Rel3	0.022	0.012	0.007	0.005	0.005	0.008	0.386	0.114	0.017	0.005	0.285
Rel4	0.037	0.012	0.009	0.005	0.007	0.008	0.295	0.974	0.022	0.005	0.515
Rel5	0.050	0.013	0.011	0.011	0.008	0.012	0.110	0.051	0.015	0.008	0.314
Exe1		0.285	0.009	0.005	0.005	0.008	0.102	0.203	0.285	0.005	0.028
Exe2			0.007	0.005	0.005	0.008	0.028	0.037	0.093	0.005	0.005
Exe3				0.009	0.007	0.008	0.011	0.022	0.047	0.005	0.005
Exe4					0.059	0.012	0.007	0.013	0.028	0.005	0.005
Exe5						0.008	0.005	0.009	0.013	0.005	0.005
Exe6							0.008	0.008	0.011	0.008	0.008
Rec1								0.646	0.005	0.721	0.508
Rec2									0.139	0.475	0.445
Rec3										0.919	0.047
Rec4											0.139

and

Table 6. Statistical significance of differences between the respiratory cycle amplitude (RCA) values calculated from 3 min segments in different conditions: Relaxation (Rel), Exercise (Exe), and Recovery (Rec). Statistically significant values are in bold.

RCA
	Exe1	Exe2	Exe3	Exe4	Exe5	Exe6	Rec1	Rec2	Rec3	Rec4	Rec5
Rel1	0.508	0.169	0.017	0.013	0.013	0.017	0.114	0.445	0.508	0.285	0.285
Rel2	0.508	0.114	0.022	0.017	0.013	0.025	0.114	0.445	0.575	0.333	0.508
Rel3	0.445	0.139	0.022	0.017	0.013	0.025	0.114	0.575	0.445	0.386	0.445
Rel4	0.441	0.093	0.022	0.017	0.005	0.012	0.059	0.445	0.721	0.386	0.086
Rel5	0.767	0.139	0.038	0.028	0.011	0.028	0.086	0.767	0.859	0.515	0.594
Exe1		0.005	0.005	0.005	0.007	0.012	0.285	0.646	0.799	0.333	0.333
Exe2			0.005	0.005	0.009	0.012	0.575	0.285	0.059	0.013	0.017
Exe3				0.017	0.025	0.012	0.285	0.028	0.009	0.007	0.007
Exe4					0.047	0.050	0.203	0.009	0.007	0.007	0.005
Exe5						0.050	0.009	0.007	0.005	0.005	0.005
Exe6							0.012	0.012	0.012	0.012	0.012
Rec1								0.007	0.005	0.005	0.005
Rec2									0.009	0.005	0.005
Rec3										0.022	0.022
Rec4											0.959

), statistical significance for the comparisons between all calculated values for each variable is given. In these tables, we omit comparisons between the values obtained during relaxation, because there were no statistically significant differences, except for HR between Rel1 and Rel2 (p = 0.047).

As expected, during submaximal HR exercise, HR increased and parasympathetic cardiac control was subsequently reduced. Besides increased BF, we also quantified a statistically significant increase in RCA (Figure 3, Table 5 and Table 6).

During recovery, HR and RCA uniformly significantly decreased, while changes in BF were almost not statistically significant. Parasympathetic activity during post-exercise recovery had a biphasic response. In the first 3 min segment, we obtained a high value of RMSSD which was not statistically different from the values obtained during relaxation (Table 4). In the second three-minute segment, we saw a drop in RMSSD, which was followed by raised values until the end of recovery.

In the first step of the examination of cardio–respiratory coupling, we were interested in the mutual behavior of HR and BF in different physiological conditions, and a statistically significant correlation was obtained only during exercise (Figure 4). Further, we analyzed the relationships between HR with RMSSD and BF with RCA under three conditions. We found a statistically significant correlation between RMSSD and HR only during relaxation. A low HR is related to high parasympathetic activity in untrained subjects, but this relationship ceased to exist at higher HR values and reduced RMSSD during recovery and exercise (Figure 5). Interestingly, a relationship between RMSSD and BF was found during exercise. A relationship between RCA and BF was not found in relaxation (p = 0.398). During exercise, there was weak positive correlation (r = 0.324, p = 0.013), and in recovery, a weak negative correlation (r = −0.34, p = 0.016).

Then, we applied the sample entropy method to calculate the regularity of the RR intervals and respiratory signal time series, as well as the cross-sample entropy, to calculate their synchrony. The SampEnRR values were the highest during relaxation and the lowest during exercise (Figure 6). There were no statistically significant differences between the SampEnResp values in the three different conditions. Finally, asynchrony between the rhythms was the highest in relaxation and the lowest during exercise.

The Pearson correlation coefficients between the subjects’ fat mass and calculated cardio–respiratory parameters are given in

Table 7. Pearson coefficient of correlation with subjects’ fat mass.

	HR	RMSSD	BF	RCA
Rel1	0.753	−0.602	0.364	−0.267
Rel2	0.845	−0.796	0.561	−0.311
Rel3	0.831	−0.723	0.445	−0.255
Rel4	0.752	−0.612	0.449	−0.225
Rel5	0.775	−0.567	0.577	−0.253
Exe1	0.879	−0.771	0.349	−0.420
Exe2	0.658	−0.669	0.506	−0.393
Exe3	0.393	−0.574	0.297	−0.443
Exe4	0.108	−0.519	−0.019	−0.372
Exe5	−0.095	−0.517	0.177	−0.309
Exe6	−0.065	−0.645	0.156	−0.162
Rec1	0.487	−0.799	0.472	−0.445
Rec2	0.453	−0.785	0.777	−0.409
Rec3	0.625	−0.738	0.681	−0.107
Rec4	0.735	−0.812	0.375	−0.117
Rec5	0.707	−0.712	0.646	−0.144

Statistically significant values are in bold.

. During the recovery period, we found strong negative correlations between the fat mass and parasympathetic activity, and a positive correlation of the fat mass with BF and HR at the end of the recovery.

The results of the multiple linear regression for the averaged values of HR, RMSSD, and BF obtained during recovery are given in

Table 8. Multiple linear regression analysis for the averaged HR over 3 min segments during post-exercise recovery in the supine position.

Dependent Variable	Model	Coefficients				Model Summary
		B	SE	Beta	Sign.	R2	Sign.
HRRec1	Const	−88.108	48.066		0.141	0.779	0.020
	HRExe3	1.496	0.393	0.883	0.020
HRRec2	Const	−39.098	36.869		0.349	0.740	0.028
	HRExe3	1.030	0.306	0.860	0.028
HRRec3	Const	39.498	0.169		0.003	1.000	0.001
	HRExe1	1.361	0.005	1.357	0.002
	BFExe5	−93.153	0.304	−0.488	0.002
	HRRel2	−0.585	0.003	−0.517	0.003
	BMI	−0.114	0.008	−0.049	0.043
HRRec4	Const	3.118	1.280		0.248	1.000	0.001
	HRExe1	1.039	0.013	1.132	0.008
	BFExe5	−67.553	0.741	−0.386	0.007
	Age	0.545	0.007	0.150	0.008
	Fat mass	−0.253	0.012	−0.316	0.030
HRRec5	Const	20.203	12.613		0.184	0.838	0.010
	HRRel4	0.884	0.194	0.915	0.010

,

Table 9. Multiple linear regression analysis for the averaged RMSSD over 3 min segments during post-exercise recovery in the supine position.

Dependent Variable	Model	Coefficients				Model Summary
		B	SE	Beta	Sign.	R2	Sign.
	Const	0.017	0.001		0.001
	RMSSDRel3	1.320	0.001	0.785	0.001
RMSSDRec1	BFExe2	−0.149	0.001	−0.369	0.001	1.000	0.001
	BFExe6	0.041	0.001	0.085	0.001
	BFRel4	−0.008	0.001	−0.029	0.005
	Const	0.054	0.001		0.003
	HRExe2	−0.001	0.001	−0.824	0.003
RMSSDRec2	BFExe4	0.186	0.002	1.230	0.008	1.000	0.002
	BFExe3	−0.099	0.002	−0.845	0.012
	RMSSDRel3	−0.104	0.005	−0.248	0.025
RMSSDRec3	Const	−0.026	0.002		0.008
	BFExe5	0.128	0.001	0.728	0.001	0.999	0.001
	Fat Mass	−0.001	0.001	−0.924	0.001
	Height	5.75 × 10−5	0.001	0.036	0.037
	Const	0.071	0.010		0.005
RMSSDRec4	HRExe1	−0.001	0.001	−0.903	0.002	0.978	0.001
	BFExe5	0.078	0.017	0.405	0.018
	Const	−0.044	0.008		0.012
RMSSDRec5	RMSSDRel5	0.589	0.054	0.905	0.002	0.979	0.003
	BFExe4	0.118	0.020	0.479	0.010

and

Table 10. Multiple linear regression analysis for the averaged BF over 3 min segments during post-exercise recovery in the supine position.

Dependent Variable	Model	Coefficients				Model Summary
		B	SE	Beta	Sign.	R2	Sign.
	Const	0.076	0.001		0.001
	BFRel5	0.573	0.001	0.746	0.001
BFRec1	RMSSDExe4	−5.041	0.001	−0.319	0.001	1.000	0.001
	HRExe4	0.001	0.001	0.072	0.001
	RCARel3	0.001	0.001	0.007	0.001
	Const	0.081	0.001		0.004
	BFRel4	0.991	0.001	0.917	0.001
BFRec2	Body Mass	0.001	0.001	0.189	0.001	1.000	0.001
	HRExe6	−0.001	0.001	−0.064	0.002
	RCAExe2	−0.001	0.001	−0.015	0.008
BFRec3	Const	−0.141	0.001		0.001
	BFRel4	1.100	0.001	0.931	0.001
	HRRel5	0.003	0.001	0.198	0.001	1.000	0.001
	RCAExe5	−0.002	0.001	−0.037	0.001
	HRExe4	6.06 × 10−5	0.001	0.005	0.002
	Const	−0.125	0.036		0.041
BFRec4	BFRel4	0.678	0.042	0.821	0.001	0.994	0.001
	HRExe1	0.003	0.001	0.312	0.008
	Const	0.116	0.002		0.011
	BFRel3	0.537	0.003	0.698	0.003
BFRec5	RMSSDExe6	−12.997	0.157	−0.814	0.008	1.000	0.001
	RMSSDExe4	6.432	0.137	0.450	0.014
	Body Mass	0.001	0.001	0.072	0.033

.

4. Discussion

The main goal of our study was to clarify the cardio–respiratory relationship and its compensatory mechanisms in three different conditions: relaxation, exercise with submaximal HR, and post-exercise recovery in an adjusted experimental protocol in young healthy recreative male subjects. To our knowledge, this is the first study which included simultaneous measurements of the heart and respiratory signals in the previously mentioned conditions.

In the relaxed state, we did not notice any statistical significance in the 3 min segments in RMSSD, BF, and RCA (Table 4, Table 5 and Table 6). HR was statistically significant between the first two 3 min segments, but with borderline significance. Potentially, this borderline significance could be attributed to the physiological response of the organism to the environment during the experiment. In support of this statement, the results of the study by Lequeux et al. indicate significantly higher HR values in subjects measured in clinical settings compared to HR values measured at home [29].

Also, there was no correlation between any of the examined parameters, except for HR and RMSSD, which showed a strong negative correlation (Figure 5). The latter result was expected, as in the relaxed state, the vagal tone of the heart is prevalent and directly influences HR levels [30,31,32]. According to the research by Gourine and Ackland (2019), high vagal activity in the relaxed state is significantly associated with a high exercise capacity [33]. Furthermore, SampEnRR and CrossSampEnRRResp showed higher values, which are indicative of the unpredictability and complexity of heart signals and the asynchrony between the heart and respiratory signals, respectively (Figure 6).

In exercise, as expected, the HR value increased, while the parasympathetic regulation of HR decreased (Table 3 and Table 4)—the relationship between these parameters ceased to exist at higher HR values. It is well known that a rapid increase in HR is governed by rapid parasympathetic withdrawal, as well as by an increased sympathetic tone during physical activity. Upon commencing exercise, conditions have to be set in relation to the intensity of exercise in order to maintain an adequate blood pressure and organ perfusion. It is thought that there are a few mechanisms which lead to these changes. In the first place, inputs from higher cortical centers activate the cardiovascular center in the brainstem, which resets the baroreceptor activity to a higher operating point [34,35]. Also, there is a feedback from muscle mechanoreceptors and metaboreceptors, as well as an increase in venous return to the right atrium due to peripheral muscle action. The result is decreased parasympathetic activity and increased sympathetic activity. The sympathetic tone is more pronounced as the intensity of physical activity becomes higher [35]. Furthermore, it has been shown that the ratio of parasympathetic to sympathetic activity operates at 4:1 during relaxation and shifts to 1:4 during high-intensity physical exercise [36]. In our study, the participants did not reach the level of maximum parasympathetic withdrawal, which may be significant in the interpretation of the results.

Further, we found significant increases in BF and RCA during exercise (Table 5 and Table 6). However, we did not notice any correlation between these two parameters. It is of note that there was a rapid increase in BF immediately after applying the workload intended for warming up, while RCA was significantly increased in comparison to the relaxation state after the second 3 min segment of exercise. There are numerous studies which have pointed out the differential regulation of BF and RCA during exercise. Namely, breathing frequency is primarily regulated by muscle afferents III and IV and “central command” [12]. The contribution of each is still debated, however, evidence suggests that the muscle afferent input is of greater importance during moderate exercise, while the “central command” has a larger contribution in high-intensity exercise. Oppositely, RCA is primarily regulated by metabolic inputs such as PaCO₂ and, to a lesser extent, the level of arterial H⁺ [10,11,12].

In our study, we found a weak positive correlation between BF and RCA in exercise, which can be explained by the different modes of their regulation and relative independence from each other in low and moderate physical activity [11,12]. Namely, studies have shown that the influence of metabolic inputs on RCA is not proportional. While RCA increases with an increase in PaCO₂ during physical activity, it is increased to a certain level, and any other increase in PaCO₂ is not followed by an additional increase in RCA. However, BF can be indirectly augmented in high-intensity exercise. It is assumed that non-pleasant sensations of severe hypercapnia induce higher-brain centers to induce BF. Nicolò et al. proposed the name “unbalanced interdependence”, where RCA is influenced more by BF than vice versa—the fine-tuning of RCA based on the levels of BF [10]. Even at rest, a voluntary decrease in BF leads to changes in RCA in order to match the level of alveolar ventilation with the volume of CO₂ expired [37,38]. While one can argue that RCA cannot be further increased due to an insufficient time to inflate the lungs in high-intensity exercise, it is difficult to distinguish the levels of influence of PaCO₂ and metabolic acidosis on RCA alone. We assume that the dependence of BF and RCA is more pronounced in high-intensity exercise than in submaximal HR settings, which might explain their weak positive correlation in our study.

With further analysis, we found a strong negative correlation between BF and RMSSD in exercise, as well as a strong positive correlation between BF and HR (Figure 5 and Figure 6). It is assumed that the sympathetic tone increases, at least in some part, because of the activation of muscle afferents III and IV, which are stimulated by various stimuli–mechanical distortions, increases in tissue temperature, and metabolic products of muscle contraction, etc. These afferents innervate lamina I in the spinal cord, as well as the intermediolateral column, which, in turn, innervates the brainstem regions which control breathing. This sequence of events leads to a rise in sympathetic activity and a decline in parasympathetic activity [39]. It is possible that the observed negative correlation between BF and RMSSD was due to the activation of β₂ adrenergic receptors in the airway smooth muscles, causing bronchodilatation with a subsequent reduction in airflow resistance [40]. This would make the breathing frequency easier to increase with the intensity of physical activity—the principle of minimal effort. Also, this correlation might not be confounded. Namely, it should be emphasized that there was no statistically significant difference in BF between the first and second 3 min segment. As it concerns RMSSD, the level of parasympathetic activity dropped until the minimum in the first two segments of exercise, and there was no statistically significant difference between the segments afterwards (Table 4 and Table 5). It can be concluded that a significant rise in BF occurs in the circumstances of minimal parasympathetic activity and a high sympathetic tone. A strong positive correlation between BF and HR makes the latter statement more plausible, maintaining the optimal ventilation–perfusion level. Moreover, this is in accordance with the metabolic regulation of RCA and its role in minimizing the changes in PaCO₂—there was no correlation between RMSSD and RCA, but a weak correlation between BF and RCA.

Further, the SampEnRR values were the highest during relaxation and the lowest during exercise (Figure 6). This is expected, as an increased sympathetic tone leads to shorter and regular RR intervals, leading to a decreased entropy value. However, we did not notice any statistically significant differences between the SampEnResp values between relaxation and exercise, which might put the applied method for determining the entropy of respiratory signals in question (Figure 6). Taking into consideration that there was no correlation between BF and RCA in relaxation and a weak positive correlation between BF and RCA in exercise, it should be expected that the entropy of respiratory signals drops. Finally, asynchrony between the rhythms was the highest in relaxation and the lowest during exercise, with a significant difference (Figure 6). This can be explained by the influence of the dominant sympathetic tone on both signals in exercise and the lack of its influence in the relaxation state. However, the muscle contraction mode should be considered, as Weippert et al. showed that, in low-intensity physical activity, dynamic and isometric contractions can have different influences on autonomic control on cardiac and respiratory parameters [41].

In post-exercise recovery, we observed a biphasic response of parasympathetic activity. There was an instant rise in parasympathetic activity reflected in a high value of RMSSD in the first 3 min segment. This was not statistically different from the values obtained during relaxation (Figure 3, Table 4). The rise in RMSSD was followed with a decline in activity in the second 3 min segment, followed again by raised values until the end of recovery. HR and RCA uniformly and significantly decreased in recovery, while BF significantly decreased in relation to exercise, but without significant changes in its values in all 3 min segments of recovery.

Immediately upon the cessation of exercise, the “central command” and peripheral input from the activated muscles were abolished, resetting the baroreceptor reflex again to a lower operating point [35,42]. This could explain the sudden rise in parasympathetic activity and sudden drop in both HF and BF (Figure 4). Also, the effect of changing posture should not be neglected. After the exercise, the participants were left to recover in a lying supine position. This change in position could explain such a rise in RMSSD in the first 3 min segment—blood pressure is always higher in the standing position, and it would not require such a level of parasympathetic activation if participants were left in the standing position to recover [9]. Furthermore, the results of the study by Barak et al. support our findings by demonstrating that the recovery rate after submaximal HR exercise is higher in the supine position compared to the sitting position for participants [43].

Afterwards, a more gradual time-dependent recovery of parameters was observed, probably due to the compensation of metabolic changes after the physical activity. This is in accordance with the observed weak negative correlation between BF and RCA.

In relation to the relaxed state, HR was significantly higher in recovery in all five 3 min segments (Table 3). BF did not differ from the values in the relaxed state, except in the third and fourth 3 min segments, where we observed a significant increase in BF in relation to the relaxed state (Table 5). This happened in parallel with the lowest level of parasympathetic activity in recovery, although we failed to find any correlation between BF and RMSSD in the recovery period (Table 4). RCA did not differ significantly in the recovery period in comparison to the relaxed state in any time segment (Table 6). In conclusion, the timeframe for recovery in our study was probably not long enough in order to obtain all of the examined parameters at the pre-exercise state levels. In general, the duration of recovery is primarily dependent on the intensity of the physical activity, as well as on its modality and duration [35]. In addition, the recovery duration may also be influenced by the age of the participants and their exercise capacity level. It has been found that athletes require less time to recover compared to non-athletes [43]. Moreover, it has been proven that prepubertal children, as well as trained adults, recover faster and have a quicker parasympathetic reactivation after a maximal HR running test compared to untrained adults [44]. Furthermore, different environmental conditions can impact the speed of recovery after exercise. The results of Haddad et al. [15] showed that hypoxia significantly delays parasympathetic reactivation after submaximal HR intensity exercise. This makes a comparison of the results of studies rather difficult, because each includes a different protocol, duration, and intensity of physical activity.

By analyzing the entropy values, we observed a significantly higher SampEnRR value in the recovery period in comparison to exercise, but still significantly lower than in the relaxed state (Figure 6). This was expected, as in the recovery period, the parasympathetic response is activated and the sympathetic response withdraws slowly, resulting in a more complex cardiac signal. However, we did not obtain any statistical difference in the SampEnResp values between any state, even though there were marked changes in BF and RCA (Figure 6). Asynchrony between the signals in recovery was not significantly different in comparison to exercise, nor in comparison to the relaxed state (Figure 6). This could be the result of a specific pattern of changes in the autonomic nervous system in the recovery period, in which signals slowly regain the characteristics of those in the relaxed state in terms of asynchrony and complexity.

Lastly, we ran a multiple regression analysis model to determine the relationship between the average 3 min HR, RMSSD, and BF segments in recovery and the other examined parameters (Table 8, Table 9 and Table 10). We found that the average HR rate in the first and second 3 min HR segments in recovery was strongly predicted by the average HR value achieved in the third 3 min segment of exercise. It is possible that the behavior of parasympathetic activity influenced the observed results. Namely, RMSSD was already minimal in the third 3 min segment in exercise and remained minimal until the end of exercise. Oppositely, in the first 3 min segment of recovery, the RMSSD value was immediately increased with the opposite behavior of the sympathetic activity—a gradual increase during exercise and gradual withdrawal during recovery. Further, this is more plausible with the observed BF in the fifth 3 min segment in exercise as a negative HR predictor in the third and fourth 3 min segments in recovery, as the sympathetic tone was the highest during the last segment in exercise and the effects of its withdrawal were emphasized in the last segments of recovery. The new balance between the two autonomic branches might have resulted in the HR in the fourth 3 min segment in relaxation being a positive predictor of the HR in the last segment of recovery. Our results from the multiple regression analysis are correlated with the observation of autonomic behavior in post-exercise recovery in the study of Garcia et al. [9].

When it comes to RMSSD, the dominant overall predictor of parasympathetic activity in all five 3 min segments was BF, along with RMSSD in the relaxed state. The sudden rise in RMSSD in the first 3 min segment and somewhat settled RMSSD in the last 3 min segment of recovery were strongly predicted by RMSSD in third and fifth 3 min segments in the relaxed state, respectively (Table 9). In the early time segments of recovery, BF in early exercise was a positive predictor of RMSSD. The same pattern was observed for later segments—BF in the last time segments of exercise was a positive predictor of RMSSD in the late recovery. Once again, this suggests the fine tuning of the two autonomic branches according to the level of their activities in exercise and recovery.

BF itself was strongly negatively predicted in the first and last 3 min segment of recovery by the level of RMSSD in the late time segments of exercise, while BF in the intermediate time segments was more predicted by the level of BF in the relaxed state (Table 10). These results are correlated with the observed upregulation of BF by the muscle afferents and “central command” in exercise and vice versa after the cessation of physical activity [12].

Finally, it is of note to mention that we found a strong negative correlation between the fat mass and parasympathetic activity during the recovery period and a positive correlation of the fat mass with BF and HR at the end of the recovery (Table 7). Our results are in line with the study by Carvalho et al. [45], where a negative correlation was also observed between the fat mass and parasympathetic activity. Moreover, they observed a lower heart rate variability in obese people [45]. Also, the study by Cornell et al. demonstrated that the percentage of body fat significantly negatively correlated with heart rate recovery after exercise testing at the submaximal HR level [46]. This suggests that it takes longer for obese people to recover in comparison to the healthy population. However, the exact mechanisms of how obesity impairs cardiac autonomic modulation remain to be elucidated in further studies.

In the end, we shall acknowledge a few limitations of our study. The sample size of ten participants is small, but we are convinced that the basic alterations in HR, BF, RMSSD, and ROC during the experimental protocol with a larger sample size study will have the same directions of change and preserved interrelationships. Despite being a pilot study, a larger sample size would enhance the robustness of our findings. In addition, including female participants could unveil intriguing gender-specific differences in the examined parameters. In an earlier study, we found sex differences between the nonlinear properties of cardiac signals and between the nonlinear properties of respiratory signals obtained in the supine position [25]. Hence, it could be expected that these are persistent in the form of cardio–respiratory coupling during exercise and recovery in healthy female participants. Other limitations pertain to the absence of concomitant blood pressure signal assessments, as well as the omission of determining the respiratory exchange ratio and assessing neuroendocrine influences.

5. Conclusions

In this pilot study, we proposed a new adjusted experimental protocol for the assessment of cardio–respiratory coupling. Moreover, we analyzed parameters derived from simultaneously recorded cardiac (RR interval) and respiratory signal time series, both before, during, and after exercise, in healthy men. Our investigation unveiled the behavior and interrelationships among the examined parameters within these conditions. Finally, we observed that post-exercise recovery parameters were influenced by the cardiopulmonary indices obtained before and during exercise, as well as by anthropometric data.