Effect of Pursed-Lip Breathing in Patients With COPD: Linear and Nonlinear Analysis of Cardiac Autonomic Modulation

Abstract

Objective: The aim of the present study was to evaluate the effect of pursed-lip breathing (PLB) on cardiac autonomic modulation in individuals with chronic obstructive pulmonary disease (COPD) while at rest. Methods: Thirty-two individuals were allocated to one of two groups: COPD (n = 17; 67.29 ± 6.87 years of age) and control (n = 15; 63.2 ± 7.96 years of age). The groups were submitted to a two-stage experimental protocol. The first stage consisted of the characterization of the sample and spirometry. The second stage comprised the analysis of cardiac autonomic modulation through the recording of R-R intervals. This analysis was performed using both nonlinear and linear heart rate variability (HRV). In the statistical analysis, the level of significance was set to 5% (p ≤ 0.05). Results: PLB promoted significant increases in the SD₁, SD₂, RMSSD and LF (ms²) indices as well as an increase in α₁ and a reduction in α₂ in the COPD group. A greater dispersion of points on the Poincaré plots was also observed. The magnitude of the changes produced by PLB differed between groups. Conclusion: PLB led to a loss of fractal correlation properties of heart rate in the direction of linearity in patients with COPD as well as an increase in vagal activity and impact on the spectral analysis. The difference in the magnitude of the changes produced by PLB between groups may be related to the presence of the disease and alterations in the respiration rate.

Keywords :

• respiration
• heart rate
• autonomic nervous system

Introduction

Pursed-lip breathing (PLB) is widely performed in lung rehabilitation programs as well as during activities of daily living among patients with chronic obstructive pulmonary disease (COPD) due to the fact that this technique provides a number of benefits to different systems, with the relief of symptoms and an increase in quality of life (1–3).

Among the observed changes with the maneuver used stands out its influence on the cardiac autonomic modulation, this emphasizes the importance of this technique owing to autonomic nervous system dysfunction present in this type of patients (4–6). PLB also exerts an influence on cardiac autonomic modulation. In a study carried out by Ramos et al. (3) using heart rate variability (HRV) as the method of analysis, the authors report that PLB reduces the sympathovagal balance through an increase in vagal performance stemming from the reduction in respiratory rate (f). The conclusions of the study were based on differences found in the root mean square of successive ­differences (RMSSD), but this result was not reproduced by the linear analysis performed in the frequency domain.

Traditional HRV analyses based on statistical procedures are limited with regard to the description of the complexity of biological systems, since cardiovascular regulation mechanisms interact in a nonlinear fashion. Therefore, nonlinear analysis methods founded on chaos theory are more sensitive for the detection of this complexity. Detrended fluctuation analysis (DFA) and the Poincaré plot are among the different indices used for this purpose (7,8).

DFA allows quantifying fractal correlation properties in non-stationary time series (9). A Poincaré plot is a graphic representation of a time series on the cartesian plane and is widely used for the quantitative and qualitative analysis of HRV (10). According to the literature, the indices obtained with these methods allow the analysis of physiological processes, such as the effect of breathing on the autonomic nervous system (ANS) (11,12) and diseases, such as COPD (13).

Based on the above-mentioned aspects, the authors of the present study hypothesize that PLB can alter the behavior of the nonlinear dynamics of HR. A search of the pertinent literature revealed no previous studies addressing the effect of PLB on cardiac autonomic modulation using a nonlinear analysis of HRV.

Thus, the aim of the present study was to analyze the effect of PLB on cardiac autonomic modulation in patients with COPD using DFA and Poincaré plots together with a linear analysis of HRV in the time and frequency domains.

Methods

Population

Thirty-two male individuals were divided into two groups [COPD (n = 17 –GOLD II –5 patients, GOLD III –9 patients, GOLD IV –3 patients) and control (n = 15)] based on the criteria established by the Global Initiative for Obstructive Lung Disease (GOLD) (14). The constitution of the control group and the anthropometric features were performed paired with the characteristics of COPD in order to avoid these variables influences on the analysis of cardiac autonomic modulation.

The following were not included in the study: smoking, alcoholism, exacerbation of COPD in the previous 2 months, medication that might affect cardiac autonomic modulation, metabolic and/or cardiac disease, restrictive pattern or non-reproducible curves in the spirometry and volunteers did not perform adequately PLB.

All volunteers received explanations regarding the objectives and procedures of the study and signed a statement of informed consent agreeing to participate. All procedures received approval from the Human Research Ethics Committee of the Faculdade de Ciências e Tecnologia –FCT/UNESP, Presidente Prudente Campus (Brazil) under process n° 24/2010.

Procedures

The procedures were carried out individually between 8 am and 12 pm in a room with temperature between 21 and 23°C and relative humidity between 40 and 60%. To reduce the anxiety of the volunteers, no people other than the subject and evaluator were permitted in the room during the data collection. All participants were submitted to a two-stage protocol, with each stage held on different days to avoid the interference of one stage on the other.

The first stage consisted of the identification of the individuals and the measurement of the anthropometric data for the determination of the body mass index (BMI). Weight was determined on a digital electronic scale (Welmy –W110 H –200 Kg –Brazil). Height was determined using a stadiometer with a precision of 0.1 cm and maximum capacity of 2 m (Sanny –Brazil). Spirometry was then performed for the classification of the volunteers using a Spirobank spirometer (MIR, Italy), following the criteria of the Spirometry Directive (15).

The second stage comprised the procedures necessary for the determination of cardiac autonomic modulation before, during and after PLB. The volunteers remained at rest seated in a chair for one hour and were asked to remain awake and quiet. This period was divided into three 20-minute phases –the first and third with spontaneous breathing and the second with PLB. Changes between phases were determined by the evaluator by means of a verbal command to the volunteer.

For this stage, the volunteers were instructed to refrain from alcohol and stimulating beverages, such as coffee and tea, the day prior to the data collection. Moreover, individuals who took maintenance medication, such as bronchodilators, anti-inflammatory mucolytic agents, etc., were asked to disrupt use 12 hours prior to the data collection. In addition, all volunteers received instructions in relation to completion of the RFL and, if necessary, the evaluator performed the maneuver with the patient to train him prior to the data collections.

Throughout the second stage, beat-to-beat HR was recorded using a Polar S810i heart meter (Polar Electro, Kempele, Finland) and f was determined by a count of the number of breathing cycles completed in one minute, determined in the 3rd and 17th minute of each 20-minute phase based on the observation of the thoracic dynamics of the volunteer.

Analysis of heart rate variability

For the analysis of HRV, beat-to-beat HR was recorded throughout the second stage with a sampling rate of 1000 Hz using a HR Polar S810i receiver (Polar Electro, Finland). This equipment has previously been validated for the capture of beat-to-beat HR and its use in HRV analysis (16–18). The series of intervals were submitted to a digital filtering process complemented by manual filtering for the elimination of premature, ectopic beats and artifacts. Only series with more than 95% sinus beats were included in the study (19). HRV analysis was ­performed using the most stable five minutes of the tracing in each 20-minute phase.

Detrended fluctuation analysis (DFA)

DFA was applied to the R-R interval time series for the analysis of the fractal properties of HR. For such, the R-R series was integrated using the following equation:

in which Y(k) is the kth term of the integrated series (k = 1, 2,  .  .  .  , N), R-R(i) is the ith value of the R-R intervals and R-Rave is the average of the R-R intervals of the original series of length N:

Next, the integrated time series is divided into intervals of length n (n = 1, 2,  .  .  .  , N). The local tendency of the series is calculated in each of these intervals by a straight line of least squares adjusted to the data. The Y coordinate of this line is denoted by Yn(k). Next, the integrated series Y(k) is detrended by subtracting the local tendency Yn(k) in each interval. For a given interval of size n, the characteristic size of the fluctuation for the integrated and detrended series is calculated as follows:

This procedure is repeated for all intervals of size n, obtaining a relationship between the mean number of fluctuations F(n) and the size of the n intervals. A linear relationship on a log-log plot indicates a scale exponential law, based on the following:

in which α is the scale exponent, which is calculated by a linear regression on a log-log plot. The short-term fractal component (alpha 1), corresponding to 4 to 11 beats, the long-term fractal component (alpha 2), ­corresponding to periods greater than 11 beats, and the ratio between these components (alpha 1/alpha 2) were calculated (13,20).

A software program was used for the DFA. This program is available at PhysioNet (http://www.physionet.org/), which is an online forum that stores biomedical signals and software programs for analyzing these ­signals (21).

Poincaré plot

A Poincaré plot is a representation of a time series on a cartesian plane. In the present study, each point of the plots was determined by an interval of the R-R series ­correlated to the subsequent interval (10). The ­quantitative analysis of the plots was expressed by the standard deviation of the instantaneous beat-to-beat variability (SD1), the long-term standard deviation of the continuous R-R intervals (SD2) and the SD1/SD2 ratio. The qualitative analysis of the plots was performed using the figures formed by its attractor, as follows (22):

1. Figure in which an increase in the dispersion of the R-R intervals occurs with the increase in the intervals –characteristic of a normal plot;

2. Figure with small overall beat-to-beat dispersion and no long-term increase in the dispersion of the R-R intervals.

Analysis of heart rate variability: Linear methods

The root mean square of successive differences (RMSSD) between normal adjacent R-R intervals and the standard deviation of all R-R intervals (SDNN) were used for the analysis of HRV in the time domain (21). The low frequency (LF: 0.04 to 0.15 Hz) and high frequency (HF: 0.15 to 0.40 Hz) spectral components in squared milliseconds and the LF/HF ratio were used for the analysis of HRV in the frequency domain. Spectral analysis was performed using the Fourier transform algorithm (22,23).

The HRV analysis software (Kubios, Biosignal Analysis and Medical Image Group, Department of Physics, University of Kuopio, Finland) was used for the analysis of these indices (24).

Data analysis

Descriptive statistics were used for the analysis of the data on the profile of the population, with the results expressed as mean, standard deviation and absolute numbers. The Shapiro-Wilk test was used to determine the distribution of the HRV data (normal or non-normal).

For the intragroup analysis, repeated-measure analysis of variance followed by Tukey's test were used for parametric data and Friedman's test followed by Dunn's test were used for nonparametric data. For the intergroup analysis, the unpaired Student's t-test was used for data with normal distribution and the Mann-­Whitney test was used for data with non-normal distribution. The level of significance was set to 5% (p < 0.05).

The calculation of the study power (GraphPad ­StatMate version 2.00 for Windows, GraphPad Software, San Diego, California, USA) based on the size of the sample analyzed and a 5% level of significance (two-tailed test) ensured a test power greater than 80% in the determination of differences among the variables

Results

Characterization of sample

Table 1 displays the anthropometric and spirometric data of the two groups. The COPD group had significantly lower weight, FEV₁ and FEV₁/FVC in comparison to the control group.

Table 1.  Anthropometric and spirometric variables of COPD and control groups

Variables	COPD	Control
Age (years)	67.29 ± 6.87 (68)[63.76 –70.82]	63.2 ± 7.96 (61)[57.92 –66.74]
Weight (Kg)	70.29 ± 11.05^a[64.61 –75.97]	81.85 ± 16.77[72.56 –91.14]
Height (m)	1.66 ± 0.05[1.63 –1.68]	1.69 ± 0.06[1.65 –1.73]
BMI (Kg/m^2)	25.54 ± 4.44[23.25 –27.82]	28.41 ± 4.9[25.7 –31.13]
FEV1/FVC	45.3 ± 9.38 (44.4)^a[40.48 –50.12]	77.25 ± 9.73 (79.6)[71.86 –82.65]
FEV1 (% predicted)	42.19 ± 14.2^b[34.89 –49.5]	102.89 ± 21.88[90.76 –115.01]

Mean ± standard deviation (median) [confidence interval]; ^a Statistically significant difference in relation to control group (Mann-Whitney test; p < 0.05); ^b Statistically significant difference in relation to control group (unpaired t-test; p < 0.05); Abbreviations: BMI (body mass index), FEV₁ (forced expiratory volume in first second); FEV₁/FCV (forced expiratory volume in first second/forced vital capacity ratio).

Table 2 displays the HRV data obtained from the nonlinear analysis. In the COPD group, PLB led to a significant increase in SD₁ and SD₂ in comparison to phase I and a reduction in the α2 index in comparison to phases I and III. In the control group, PLB led to a significant increase in SD₁ in comparison to phase I and in both SD₂ and α1 in comparison to phases I and III. A reduction in the SD₁/SD₂ ratio in phase II in comparison to phase I and a reduction in the α2 index in phase II in relation to phases I and III were also found in the control group. The intergroup analysis revealed significantly lower SD₁ and SD₂ in the COPD group in all phases of the protocol and higher α2 index values in the COPD group with the execution of PLB.

Table 2.  Analysis of nonlinear HRV indices in each phase of protocol

Variable		Phase I	Phase II	Phase III
SD1 (ms)	COPD	9.45 ± 5.5 (7.6)^ae[6.62 –12.28]	12.32 ± 6.8 (12.2)^f[8.82 –15.82]	10.1 ± 5.24 (8.5)^e[7.4 –12.79]
	Control	14.04 ± 5.14 (13.7)^a[8.7 –27.6]	17.72 ± 5.95 (15.8)[14.42 –21.01]	15.38 ± 5.92 (13.0)[12.1 –18.66]
SD2 (ms)	COPD	31.86 ± 12.83^be[25.26 –38.46]	46.14 ± 27.07^f[32.21 –60.06]	35.61 ± 15.02^e[27.89 –43.34]
	Control	46.66 ± 17.52 (40.8)[36.95 –56.36]	76.49 ± 29.51 (65.5)^c[60.14 –92.84]	50.73 ± 19.96 (44)[39.67 –61.79]
SD1/ SD2	COPD	0.29 ± 0.1 (0.26)[0.24 –0.34]	0.28 ± 0.11 (0.23)[0.22 –0.34]	0.28 ± 0.07 (0.27)[0.24 –0.31]
	Control	0.30 ± 0.08 (0.29)^a[0.26 –0.35]	0.24 ± 0.06 (0.23)[0.21 –0.27]	0.31 ± 0.06 (0.29)[0.27 –0.34]
α1	COPD	1.12 ± 0.18[1.03 –1.22]	1.23 ± 0.29[1.08 –1.38]	1.17 ± 0.20[1.07 –1.28]
	Control	1.05 ± 0.16[0.96 –1.14]	1.38 ± 0.12^d[1.31 –1.4]	1.08 ± 0.14[1.00 –1.16]
α2	COPD	0.97 ± 0.23 (1.01)[0.85 –1.08]	0.62 ± 0.22 (0.64)^cf[0.51 –0.74]	0.85 ± 0.23 (0.89)[0.73 –0.97]
	Control	0.94 ± 0.09[0.89 –0.99]	0.48 ± 0.17^d[0.38 –0.57]	0.88 ± 0.16[0.79 –0.97]

Mean ± standard deviation (median) [confidence interval]; ^aStatistically significant difference in relation to phase II (Friedman's test followed by Dunn's test; p < 0.05); ^bStatistically significant difference in relation to phase II (repeated-measure analysis of variance followed by Tukey's test; p < 0.05); ^cStatistically significant difference in relation to phases I and III of protocol (Friedman's test followed by Dunn's test; p < 0.05); ^dStatistically significant difference in relation to phase I and III of protocol (repeated-measure analysis of variance followed by Tukey's test; p < 0.05); ^eStatistically significant intergroup difference in each phase of protocol (Mann-Whitney test; p < 0.05); ^fStatistically significant intergroup difference in each phase of protocol (unpaired t-test; p < 0.05).

Figure 1 displays the Poincaré plots illustrating the changes caused by PLB in comparison to the spontaneous breathing performed at the beginning and end of the protocol. PLB led to an increase in the dispersion of points in both groups.

Figure 1.  Poincaré plots of phases of protocol in COPD and control groups.

Table 3 displays the HRV indices analyzed in the time and frequency domains. In the COPD group, PLB led to a significant increase in SDNN, RMSSD and LF (ms²) in comparison to phase I. In the control group, PLB led to an increase in SDNN, LF (ms²) and the LF/HF ratio in comparison to phases I and III as well as an increase in RMSSD in comparison to phase I. The intergroup analysis demonstrates that the COPD group had significantly lower SDNN, RMSSD and LF (ms²) in all phases of the protocol and significantly lower HF (ms²) in phases I and III in comparison to the control group.

Table 3.  Analysis of linear HRV variables in each phase of protocol

Variables		Phase I	Phase II	Phase III
SDNN (ms)	COPD	23.61 ± 9.65^ad[18.65 –28.58]	33.87 ± 19.52^e[23.83 –43.91]	26.23 ± 11.09^d[20.53 –31.94]
	Control	34.55 ± 12.71 (29.7)[27.51 –41.59]	55.58 ± 21.08 (49.1)^c[43.91 –67.26]	37.56 ± 14.57 (32.4)[29.49 –45.64]
RMSSD (ms)	COPD	13.34 ± 7.77 (10.8)^bd[9.34 –17.33]	17.39 ± 9.61 (17.2)^e[12.45 –22.34]	14.27 ± 7.39 (12.1)^d[10.46 –18.07]
	Control	19.81 ± 7.24 (19.3)^b[15.8 –23.82]	25 ± 8.41 (22.3)[20.34 –29.65]	21.72 ± 8.37 (18.3)[17.08 –26.36]
LF (ms^2)	COPD	211.76 ± 297.5 (119)^bd[58.79 –364.73]	1182.35 ± 1846.8 (526)^d[232.77 –2131.9]	243.52 ± 259.45 (112)^d[110.13 –376.93]
	Control	365.8 ± 358.06 (195)[167.49 –564.11]	2149.13 ± 1640.5 (1732)^c[1240.5 –3057.7]	518.2 ± 594.67 (299)[188.85 –847.55]
HF (ms^2)	COPD	62.94 ± 79.82 (37)^d[21.89 –103.99]	120.94 ± 181.21 (58)[27.76 –214.12]	76.17 ± 100.56 (33)^d[24.47 –127.88]
	Control	144.86 ± 127.67 (104)[74.15 –215.58]	133.46 ± 99.51 (102)[78.35 –188.58]	182 ± 235.66 (99)[51.48 –312.52]
LF/HF	COPD	5.58 ± 4.84 (3.19)[3.09 –8.07]	16.21 ± 17.25 (12.03)[7.34 –25.08]	4.91 ± 4.47 (4.01)[2.62 –7.21]
	Control	3.07 ± 3.07 (2.17)[1.37–4.78]	26.67 ± 21.08 (15.0)^c[13 –36.35]	3.58 ± 2.84 (2.29)[2.01 –5.15]

Mean ± standard deviation (median) [confidence interval]; ^aStatistically significant difference in relation to phase II (repeated-measure analysis of variance followed by Tukey's test; p < 0.05); ^bStatistically significant difference in relation to phase II (Friedman's test followed by Dunn's test; p < 0.05); ^cStatistically significant difference in relation to phases I and III of protocol (Friedman's test followed by Dunn's test; p < 0.05); ^dStatistically significant intergroup difference in each phase of protocol (Mann-Whitney test; p < 0.05); ^eStatistically significant intergroup difference in each phase of protocol (unpaired t-test; p < 0.05).

Figure 2 shows the f behavior throughout the study protocol. The control group had lower f in comparison to the COPD group. Moreover, significant reductions were found in both groups during PLB in comparison to the spontaneous breathing performed before and after PLB.

Figure 2.  Behavior of respiratory rate during protocol; mean ± standard deviation; ^#significant difference in phase 2 in relation to phase 1 and phase 3 in both groups (repeated-measure analysis of variance followed by Tukey's test); *significant intergroup difference in each phase of protocol (unpaired t-test; p ≤ 0.05).

Discussion

The findings of the present study suggest that PLB performed at rest led to the loss of the fractal correlation properties of HR in both groups. An increase in vagal activity and impact on the spectral analysis were also observed with PLB. The magnitude of the changes differed between groups, which may be related to the presence of COPD and alterations in respiratory rate.

The nonlinear analysis of heart rate have been increasing interest, since there is evidence that the mechanisms involved in cardiovascular regulation likely to interact with each other in a nonlinear way (25). These methods describes the fluctuation complex of the heartbeat and can separate non-linear behavior structures in the time series of heart beats more appropriately than linear methods (26). With this allows better discrimination between a person with normal or changed physiology (19) and a more sensitive autonomic nervous system analysis.

The nonlinear findings were based on DFA and the qualitative analysis of Poincaré plots. In the analysis of the exponents obtained from DFA, values close to 1 characterize fractal behavior, whereas values close to 0.5 represent unpredictability, in which no correlation between values is found, and values close to 1.5 represent linearity, demonstrating strongly correlated behavior (27,28).

In the COPD group, PLB led to a tendency toward an increase in alpha 1 values in relation to baseline, which ceased when the breathing technique was interrupted. This suggests that PLB leads to the loss or disarrangement of short-term fractal correlation properties of the HR dynamics toward more linear dynamics. While performing the pursed lips breathing the movements of inspiration and expiration influences the behavior of heart rate, inducing increased of heart rate during inspiration and a reduction in expiration (3). This behavior of the heart rate induced by the PLB has periodic profile and constant, and is interpreted by analysis of Detrended Fluctuation Analysis (DFA) as a linear pattern.

In a study addressing the effects of different breathing patterns on the nonlinear dynamics of HR, Pentila et al. (29) report similar alpha 1 behavior for a respiratory rate close to 6 irpm.

In both groups, PLB led to a statistically significant reduction in the long-term fractal component (alpha 2), which was more accentuated in the control group. This reduction in alpha 2 may be explained by the increase in variability in the power or the LF spectrum, which diminishes the time of the alpha 2 analysis window (29).

A qualitative analysis of the Poincaré plots demonstrates that PLB led to a greater dispersion of points, indicating an increase in HRV. However, this increase was less accentuated in the COPD group.

Regarding the linear indices, the COPD group demonstrated diminished HRV, characterized by a reduction in the activity of both autonomic branches. Pantoni et al. (6) and Carvalho et al. (13) report similar findings. This reduction may have influenced the magnitude of the autonomic responses to PLB, which was lesser in the COPD with regard to the majority of indices analyzed, possibly related to the presence of the disease.

PLB promoted an increase in RMSSD and SD₁, which the literature reports are indicators of parasympathetic activity (12). These findings are in agreement with those reported by Ramos et al. (3), who also found an increase in RMSSD using PLB, which was explained by the reduction in f stemming from this breathing technique.

PLB also produced a significant increase in SDNN and SD₂, indicating that the technique led to greater variation in the R-R intervals and, consequently, better overall variability, which may be explained by the increase in vagal activity during the execution of the procedure. The lower values found in the case group in comparison to the control group may be related to autonomic alterations exhibited by individuals with COPD.

In the spectral analysis, the reduction in f stemming from PLB accentuated the frequency peak in the LF band, as demonstrated by the higher absolute values in comparison to spontaneous breathing. Respiratory rates close to 6 irpm (0.1 HZ) are reported to promote an evident peak in the predetermined section of the LF band (12,30). In the present study, the magnitude of this was greater in the control group, which may be explained by the more accentuated reduction in f, with values close to 5 irpm.

The increase in the LF band during PLB suggests an explanation for the behavior of the HF (ms²) in the present study. The non-significant increase in HF (ms²) in the COPD and even the reduction in the control group may be associated with the diminished f caused by the breathing technique, which was reflected in the displacement the most evident frequency peak in the LF band.

The present study has limitations that should be addressed. The COPD group, which was classified based on the GOLD criteria (14) had five patients classified as GOLD II (50% < FEV₁ < 80% of predicted), nine ­classified as GOLD III (30% < FEV₁ < 50% of predicted) and three classified as GOLD IV (FEV₁ < 30% of predicted). Despite being a heterogeneous group with regard to disease severity, Camillo et al. (5) state that this aspect does not significantly affect the analysis of HRV. Moreover, while the sample was matched for age and BMI to minimize the influence of these variables on the ANS, the groups had individuals with arterial hypertension, which the literature describes as a risk factor that affects the ANS (31). However, the number of individuals with hypertension was similar in both groups, thereby minimizing the influence of this factor.

Considering the widespread use of this breathing technique among pacients with COPD, it is important to know the physiological aspects involved in its implementation to allow security to professionals and patients who use the technique.

Conclusion

Pursed-lip breathing affected the behavior of the autonomic nervous system in both groups studied, characterized by a loss of fractal correlation properties of heart rate in the direction of linearity due to the periodicity provided by the maneuver as well as an increase in vagal activity and impact on the spectral analysis. The magnitude of the changes produced by pursed-lip breathing differed between groups, which may be related to the presence of the disease and alterations in the respiration rate.

Declaration of Interest Statement

The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.