Effect of paced breathing on ventilatory and cardiovascular variability parameters during short-term investigations of autonomic function

G. D. Pinna, R. Maestri, M. T. La Rovere, E. Gobbi, F. Fanfulla

Abstract

Paced breathing (PB) around 0.25 Hz has been advocated as a means to avoid confounding and to standardize measurements in short-term investigations of autonomic cardiovascular regulation. Controversy remains, however, as to whether it causes any alteration in autonomic control. We addressed this issue in 40 supine, middle-aged, healthy volunteers by assessing the changes induced by PB (0.25 Hz for 8 min) on 1) ventilatory parameters, 2) the indexes of autonomic control of cardiovascular function, and 3) the spectral indexes of cardiovascular variability. Subjects were grouped into group 1 (n = 31), if spontaneous breathing was regular and within the high-frequency (HF) band (0.15–0.45 Hz), or group 2 (n = 9), if it was irregular or slow (<0.15 Hz). In both groups, PB was accompanied by an increase in minute ventilation (both groups, P < 0.01), whereas tidal volume increased only in group 1 (P = 0.0003). End-tidal CO2 decreased by [median (lower quartile, upper quartile)] −0.2 (−0.5, −0.1)% (group 1, P < 0.0001) and −0.6 (−0.8, −0.5)% (group 2, P = 0.008). Mean R-R interval and systolic and diastolic pressure remained remarkably stable (all P ≥ 0.13, both groups). No significant changes were observed in spectral indexes of R-R and pressure variability (all P ≥ 0.12, measured only in group 1 to avoid confounding), except in the HF power of pressure signals, which significantly increased (all P < 0.05) in association with increased tidal volume. In conclusion, PB at 0.25 Hz causes a slight hyperventilation and does not affect traditional indexes of autonomic control or, in subjects with spontaneous breathing in the HF band, most relevant spectral indexes of cardiovascular variability. These findings support the notion that PB does not alter cardiovascular autonomic regulation compared with spontaneous breathing.

  • heart rate variability
  • controlled breathing
  • baroreflex sensitivity
  • spectral analysis

a large number of studies (9, 19, 35) have been published in the last two decades using the analysis of spontaneous oscillations of pulse interval and blood pressure as an indirect, quantitative, and noninvasive means to assess the autonomic control of cardiovascular function. Both physiopathological investigations and clinical applications have been carried out, and a rich variety of algorithms for analysis have been proposed. Among them, those based on spectral analysis of individual signals and of their mutual interrelationships have gained the highest interest, mainly because they would enable an interpretative link between different oscillatory components and specific physiological mechanisms (18, 23, 37). To control for confounding factors that might affect measured parameters, cardiovascular variability signals have been mostly collected during short-term recordings performed in dedicated laboratories, where environmental and physical conditions have been kept constant and emotional and sensorial stimuli have been avoided. Less attention, however, has been paid to respiration, despite much evidence that breathing characteristics (frequency, amplitude, and regularity) markedly affect beat-to-beat cardiovascular variability (1, 6, 22, 31, 32).

A major assumption underlying the interpretation of spectral analysis of cardiovascular variability signals is that oscillatory components lying within the so-called high-frequency (HF) band (0.15–0.45 Hz) are of respiratory origin, whereas those within the low-frequency (LF) band (0.04–0.15 Hz) are of pure cardiovascular origin (18). This assumption, in turn, implies that the frequency content of lung-volume oscillations is well within the HF band, a condition often violated in resting humans. In many subjects, for instance, breathing frequency slows down to the LF band, and entrainment of the cardiovascular rhythm around 0.1 Hz often occurs (36). In other subjects, a markedly irregular breathing pattern causes respiratory frequency components to spread over the LF band, increasing spuriously the frequency content of cardiovascular oscillations (22). Moreover, in subjects with a regular breathing activity within the HF band, the respiratory frequency is characterized by a marked variability both between and within individuals (15). Because the transfer between respiratory oscillations and cardiovascular signals is a frequency-dependent phenomenon (6, 7, 14, 31), this variability would likely translate into variability of HF spectral parameters.

Therefore, voluntary control of respiration at frequencies ≥0.2 Hz has been advocated as a means to avoid confounding and to standardize measurements in short-term evaluations of autonomic cardiovascular rhythms (6, 7, 10, 31, 32, 36). This procedure would also improve the reproducibility of some spectral parameters (30). General consensus on the use of controlled breathing protocols, however, has not yet been reached, and standardization guidelines are still lacking. A major reason is that contradictory results have been published so far on the effect of controlled-breathing procedures on autonomic cardiovascular control (7, 23, 25, 26).

The aim of this study was to determine whether a simple fixed-frequency paced-breathing protocol, suitable to be used routinely in physiopathological and clinical investigations, alters cardiovascular autonomic regulation compared with spontaneous breathing. We assumed that three major mechanisms may potentially account for such alteration. First, paced breathing may change respiratory parameters, thus affecting cardiovascular regulation both thorough a variety of reflex inputs from peripheral sensory systems and through a direct effect of central respiratory activity on the efficacy of basic cardiovascular reflexes (8, 9, 34). Second, cortical inputs associated with the act of voluntarily controlling breathing frequency may directly affect cardiovascular regulation (12, 13, 21). Third, an intrinsic effect of changing breathing frequency on autonomic control cannot be excluded, as some investigators (5, 11) have shown that slow breathing may, in fact, induce autonomic changes. Therefore, to reach the goal of our study, we first assessed whether paced breathing alters major ventilatory parameters (tidal volume, minute ventilation, and arterial blood-gas tensions) and then examined its effect on the indexes of autonomic control of cardiovascular functions (heart rate and systolic and diastolic arterial pressure) and on the spectral indexes of cardiovascular variability, including spectral baroreflex sensitivity. Regarding the latter measurements, we had to take into account that, according to the premises of the study, these indexes were likely severely biased in those subjects having slow or irregular breathing during spontaneous respiration. Accordingly, only respiratory parameters and traditional indexes of autonomic control were measured in these subjects.

METHODS

Subjects.

We studied 41 healthy volunteers, 37 men and 4 women [median age (lower quartile, upper quartile), 54 (49, 62) yr; range, 40–69 yr]. Age and gender distributions were chosen so as to be close to the typical demographic characteristics of cardiac-disease patients (16). The study was approved by the Ethical Committee of the Salvatore Maugeri Foundation (Pavia, Italy), and all subjects gave their written informed consent before participation.

Experimental protocol.

Recordings were carried out in our laboratory for autonomic assessment between 3:00 PM and 6:00 PM with the subjects in the supine position. After instrumentation and calibration were completed, subjects carried out a session of familiarization with the paced-breathing protocol. They were instructed to follow recorded instructions to breath in and out at a frequency of 0.25 Hz, with the inspiratory duty cycle [inspiratory time (Ti)/total breath time (Ttot)] set at 0.4. After an initial trial, they were asked whether they felt comfortable with the paced-breathing frequency or would they rather prefer to slightly increase or decrease it. Accordingly, an adjustment was made within ±10% of the target value (0.25 Hz). After signal stabilization was completed, subjects breathed spontaneously for 8 min and then breathed at their paced-breathing frequency for another 8 min. During these two sessions, we recorded ECG, lung volume (Respitrace Plus, Non-Invasive Monitoring Systems), noninvasive arterial blood pressure (Finapres 2300, Ohmeda), arterial oxygen saturation (Biox 3740, Ohmeda), and expired CO2 concentration (Oscaroxy, Datex-Ohmeda). The calibration of lung-volume measurements was carried out by taking a simultaneous 30-s recording of the respiratory flow using a Fleisch pneumotachograph (47304A, Hewlett-Packard). The flow was digitally integrated and regressed on the lung volume signal to obtain the calibration factor. To avoid artifacts in the arterial pressure signal, the Finapres self-adjustment was disconnected during the recordings and temporarily switched on between spontaneous and paced-breathing sessions to allow for recalibration.

Measurements.

The beginning and the end of the inspiratory and expiratory phase of each breath cycle were automatically identified in the lung-volume signal, and the corresponding Ti, Ttot, breathing frequency (1/Ttot), and tidal volume were computed. The time series of breath-by-breath minute ventilation was then generated by dividing tidal volume values by corresponding Ttot values. End-tidal CO2 (ETCO2) was measured in each expiratory phase as the mean value of the alveolar plateau of the CO2 curve in the region preceding the inspiratory drop.

Beat-by-beat R-R interval (resolution, 1 ms) and systolic and diastolic arterial pressure values were obtained from the ECG and arterial pressure signals with the use of custom-made software (17). The lung volume, R-R interval, and systolic and diastolic pressure time series were resampled at 2 Hz. Each recording was visually inspected, and a 5-min segment free from artifacts, large transients, or marked changes in the fluctuation pattern of the signals was interactively selected. This selection was aimed at satisfying the stationarity requisite for the proper application of the spectral-analysis techniques (27).

To avoid the effect of possible outliers, due to occasional sighs, for instance, representative figures of all respiratory parameters during spontaneous and paced breathing were computed as the median value of the corresponding time series. Following a common convention, representative figures of R-R interval and systolic and diastolic pressure were computed as the mean value of the corresponding time series.

After detrending was completed via least squares, second-order polynomial fitting, univariate spectral analysis was performed on lung volume, R-R interval, and systolic and diastolic pressure signals using the autoregressive method (Burg algorithm) with spectral decomposition (Johnsen and Andersen algorithm) and was verified by the classical nonparametric weighted-covariance approach (Parzen window with a bandwidth of 0.015 Hz) (27). Autoregressive model order was set at 26 but was interactively increased whenever main oscillatory components observed in the Blackman-Tukey spectrum were not resolved by the autoregressive one or when negative components appeared in the spectral decomposition table (27). To cope with the possible presence of more than one spectral component within each band, we computed the characteristic frequency and power of the LF and HF bands as, respectively, the barycentric frequency and sum of powers of the spectral components identified in each of them. The barycentric frequency is the weighted mean of the central frequencies of spectral components in the band with the use of the power of each component as weight. In these computations, we did not consider spectral components as having <10% of the overall power in the band, because they likely represented pure-noise contributions. LF and HF powers of the R-R interval were also computed in normalized units (37).

From the lung-volume spectrum, we computed the spectral leakage of respiratory components into the LF band as: Math This parameter gives the proportion of respiratory variability that is accounted for by oscillatory components in the LF band. Subjects with a spectral leakage of ≤15% during spontaneous breathing were allocated to group 1 (regular breathing in the HF band), whereas those with spectral leakage of >15% were allocated to group 2 (slow or irregular breathing).

We performed bivariate spectral analysis between systolic pressure and R-R interval time series (weighted covariance method with 0.03- Hz bandwidth Parzen window) and computed spectral baroreflex sensitivity by averaging the transfer function modulus within the LF band (28). Bivariate spectral analysis was also carried out to examine the coherence between lung volume and R-R interval.

Statistical analysis.

Because of a marked skewness in the distribution of some variables (mostly spectral parameters), descriptive statistics are given as median (lower quartile, upper quartile), unless otherwise specified. Normality was assessed by the Shapiro-Wilk test. Paired comparisons between spontaneous and paced breathing were performed by the Wilcoxon signed-rank test. Comparisons between groups were carried out by the Mann-Whitney U-test. The strength of linear association between pairs of variables was assessed by the Pearson coefficient of correlation. Because this measurement is greatly affected by the presence of outliers and by skewed distributions, the former were removed before computation and Spearman's rank correlation was used when skewed variables were involved in the relationship. An observed value was considered an outlier when it exceeded or was less than the upper or lower quartile ±1.5 times the interquartile range (38). A P < 0.05 was considered statistically significant, and all tests were two-sided.

RESULTS

In nine male subjects, the spectral leakage of respiratory components into the LF band was >15% (range, 42–92%). In three of them, this was due to a slow phasic respiratory oscillation generating a narrowband spectrum centered at 0.08, 0.12, and 0.13 Hz, respectively. This oscillation fully entrained the LF component of cardiovascular variability signals. In another five subjects, spectral leakage was caused by an irregular breathing pattern characterized by random marked changes in frequency and amplitude of the respiratory oscillation, generating a broadband spectrum with major components in the LF band and also, in three subjects, very LF (VLF, 0.01–0.04 Hz) band. Another subject showed recurring cycles of hyperventilation at LF followed by hypoventilation at HF, again causing spectral leakage in the LF and VLF bands. Representative examples of slow and irregular breathing, together with the changes induced by paced breathing, are given in Figs. 1, 2, and 3. In one female subject, the paced-breathing protocol caused anxiety and discomfort, and the recording was stopped. Accordingly, we studied 31 subjects in group 1 and 9 subjects in group 2. Age was similar in the two groups: 53 (46, 62) yr in group 1 versus 55 (55, 65) yr in group 2 (P = 0.14). An example of recorded signals in a subject with regular breathing in the HF band is given in Fig. 4.

Fig. 1.

Representative example of slow breathing at ∼0.13 Hz. During spontaneous breathing (left), lung volume (Lung Vol) oscillation fully entrained R-R interval, shown clearly by peak in R-R interval spectrum and coherence function at same frequency of peak of lung volume spectrum. During paced breathing (right), peak of lung volume spectrum shifted to 0.23 Hz, causing dramatic decrease in low-frequency (LF, 0.04–0.15 Hz) power of R-R interval and identical shift in peak of coherence. Similar results were obtained for arterial pressure variability signals. LF power of R-R interval, systolic pressure, and baroreflex sensitivity were, respectively, 145 ms2, 7.1 mmHg2, and 3.8 ms/mmHg during spontaneous breathing and 23 ms2, 1.9 mmHg2, and 1.8 ms/mmHg during paced breathing.

Fig. 2.

Representative example of irregular breathing. During spontaneous breathing (left), most of the power of lung volume signal was located within LF band. This caused power of R-R interval signal in same band to be inflated by respiratory components, as indicated by high coherence between lung volume and R-R interval. During paced breathing (right), power of lung volume signal was concentrated ∼0.23 Hz, with no spectral components within LF band. Consequently, LF power of R-R interval dramatically decreased, as well as coherence between lung volume and R-R interval in same band. Similar results were obtained for arterial pressure variability signals. LF power of R-R interval, systolic pressure, and baroreflex sensitivity were, respectively, 1,233 ms2, 20.0 mmHg2, and 5.2 ms/mmHg during spontaneous breathing and 114 ms2, 8.7 mmHg2, and 2.8 ms/mmHg during paced breathing.

Fig. 3.

Representative example of irregular breathing with alternating phases of hyperventilation and hypoventilation. During spontaneous breathing (left), spectral component of lung volume signal at 0.13 Hz caused coherent oscillations of R-R interval at same frequency. During paced breathing (right), power of lung-volume signal was concentrated ∼0.24 Hz, LF power of R-R interval at 0.13 Hz dramatically decreased, and coherence between lung volume and R-R interval became significant only around paced breathing frequency. Similar results were obtained for arterial pressure variability signals. LF power of R-R interval, systolic pressure, and baroreflex sensitivity were, respectively, 1,206 ms2, 7.2 mmHg2, and 9.1 ms/mmHg during spontaneous breathing and 197 ms2, 5.0 mmHg2, and 5.7 ms/mmHg during paced breathing.

Fig. 4.

Representative example of spontaneous and paced breathing in one subject with regular spontaneous breathing in high-frequency band (HF, 0.15–0.45 Hz). During spontaneous breathing (left), whole power of lung volume signal was located within HF band, and there was virtually no coherence between lung volume and R-R interval in LF band. LF power of R-R interval, systolic pressure, and baroreflex sensitivity were, respectively, 125 ms2, 5.5 mmHg2, and 3.9 ms/mmHg during spontaneous breathing and 157 ms2, 7.4 mmHg2, and 3.3 ms/mmHg during paced breathing (right).

Subjects with regular breathing in the HF band: group 1.

In group 1 subjects, breathing frequency during spontaneous breathing was normally distributed (P = 0.54), with a median value of 0.25 (0.21, 0.29) Hz and a range between 0.18 Hz and 0.36 Hz. The spectral leakage of respiratory components into the LF band was 3.5 (2.0, 6.0)%. As shown in Fig. 5, paced breathing forced the breathing frequency of all subjects to converge to ∼0.24 Hz, with a corresponding spectral leakage of 2.3 (1.9, 3.7)%. The Ti-to-Ttot ratio during paced breathing was very close to spontaneous breathing [0.45 (0.44, 0.48) vs. 0.44 (0.43, 0.46); P = 0.03] but was slightly greater than the target ratio (0.4; P < 0.0001).

Fig. 5.

Case profile plot of breathing frequency during spontaneous and paced breathing. Target frequency during paced breathing was 0.25 Hz, but subjects were allowed to adapt it within ±10% to improve their comfort during experiment. Each symbol represents a different subject from group 1; n = 31 subjects.

Relevant descriptive statistics of respiratory parameters are reported in Table 1. Tidal volume and minute ventilation were significantly greater during paced breathing than during spontaneous breathing. The scatterplot in Fig. 6 shows that in almost all subjects who had to reduce their spontaneous breathing rate, there was an increase in tidal volume (top left quadrant of the plot), whereas in those who had to increase the breathing rate, the change in tidal volume was rather scattered (bottom and top right quadrant of the plot). Overall, the correlation coefficient between the two changes was −0.56 (P = 0.0013). Changes in tidal volume and minute ventilation were highly correlated (r = 0.74; P < 0.0001).

Fig. 6.

Relationship between change in breathing frequency and corresponding change in tidal volume changing from spontaneous to paced breathing. Top, left: outlier, characterized by marked increased in tidal volume during paced breathing, is easily discernable. Correlation coefficient was −0.56 (P = 0.0013).

View this table:
Table 1.

Descriptive statistics of respiratory parameters during spontaneous and paced breathing in subjects with regular breathing in the HF band

Paced breathing was accompanied by a slight, albeit a highly statistically significant, decrease in ETMath and an increase in O2 saturation. As shown in Fig. 7, changes in ETMath were inversely related to changes in minute ventilation (r = −0.78; P < 0.0001).

Fig. 7.

Relationship between change in minute ventilation and corresponding change in end-tidal CO2 (ETFormula) changing from spontaneous to paced breathing. Correlation coefficient was −0.78 (P < 0.0001).

Table 2 reports descriptive statistics for cardiovascular parameters. Mean R-R interval and systolic and diastolic pressure did not show any significant systematic change passing from spontaneous to paced breathing. Similarly, there was no significant change in the LF power of all cardiovascular variability signals and in the HF power of the R-R interval. Conversely, the HF power of both arterial pressure signals significantly increased during voluntary control of respiration. Changes in HF power were moderately correlated to changes in tidal volume (Spearman's correlation, Θ = 0.49, P = 0.005; and Θ = 0.34, P = 0.06, for systolic and diastolic pressure, respectively).

View this table:
Table 2.

Descriptive statistics of cardiovascular parameters during spontaneous and paced breathing in subjects with regular breathing in HF band

Characteristic frequencies of LF and HF bands of R-R interval and systolic and diastolic pressure remained remarkably stable with the passing from spontaneous to paced breathing. Baroreflex sensitivity remained substantially stable in the two experimental sessions. Median values of this parameter appeared to be lower than what one might expect when using other estimation techniques. As shown in a previous study (28), this is indeed a peculiarity of the computation method used.

Subjects with slow or irregular breathing: group 2.

In group 2 subjects, paced breathing caused spectral leakage to dramatically decrease from 55 (54, 78)% to 11 (5, 16)% (P = 0.004). Respiratory and standard cardiovascular parameters are reported together in Table 3 (we did not measure spectral indexes in these subjects). Breathing frequency increased to ∼0.22 Hz, with almost no change in tidal volume. Minute ventilation significantly increased, and this change was higher than that observed in subjects with a regular spontaneous breathing in the HF band (P = 0.03 in the between-group comparison). The decrease in ETMath and increase in O2 saturation after increased ventilation, however, were comparable to those observed in the other group (P = 0.06 and P = 0.58, respectively, in the between-group comparison). The change in ETMath was negatively correlated with the change in minute ventilation (r = −0.74; P = 0.04). Paced breathing changed neither mean R-R interval nor mean systolic and diastolic arterial pressure.

View this table:
Table 3.

Descriptive statistics of respiratory and cardiovascular parameters during spontaneous and paced breathing in subjects with slow or irregular breathing

DISCUSSION

Breathing and cardiovascular oscillations.

Several studies have documented that the magnitude of respiratory-related fluctuations of R-R interval (i.e., respiratory sinus arrhythmia) dramatically changes according to breathing rate, exhibiting a band-pass behavior with the highest response ∼0.1 Hz and a progressive decrease of the gain as the respiratory frequency increases above this value (6, 15, 33). Consequently, when respiration becomes slow or irregular with major frequency components near the center of the LF band, a marked increase of the LF power of the R-R interval occurs. This condition is usually accompanied by a simultaneous attenuation or an almost absence of HF power of the same signal. Both these phenomena will therefore act as powerful confounders in spectral measurements of heart rate variability. As shown in this study, slow or irregular breathing patterns are more likely to occur in healthy subjects than is usually recognized. Indeed, they were found in 9 (22%) of 40 subjects who performed the recording protocol, with a spectral leakage of respiratory components into the LF band ranging from 42% to 92%.

Besides breathing frequency, tidal volume also affects respiratory sinus arrhythmia by modifying the gain of the passband and the steepness of the roll-off of the transfer function between lung volume and R-R interval (6, 15). These changes, however, have consistently been shown to be modest (2, 6, 15). Brown et al. (6), for instance, reported an average 7% increase of the gain between 0.1 to 0.4 Hz for an increase in tidal volume from 1.0 to 1.5 liters (6).

Although less studied, previous investigations (7, 22, 33) have shown that the oscillations induced by respiration on systolic and diastolic pressure have a frequency-dependent behavior similar to that of R-R interval. Hence, confounding would also affect spectral measurements of arterial pressure variability whenever slow or irregular respiration occurs. Moreover, because the computation of spectral baroreflex sensitivity involves the interrelationship between R-R interval and systolic pressure oscillations in the LF band, a biased measurement of this quantity would also be expected.

Voluntary control of breathing.

Voluntary control of breathing has been proposed by several authors (7, 10, 14, 31, 36) as an effective means to avoid confounding in short-term evaluations of cardiovascular variability. Indeed, by breathing regularly at a frequency centered in the HF band, spectral leakage of respiratory components into the LF band would be avoided and a correct attribution of LF and HF spectral measurements to physiological mechanisms would be possible. In addition, this procedure would allow a standardization of measurements between and within individuals.

In this study, we have proposed a simple breathing protocol based on the control of breathing frequency alone. Besides the fact that at ∼0.25 Hz the effect of tidal volume on respiratory-related R-R interval and systolic pressure oscillations is negligible (6, 7, 15), controlling only breathing frequency is much easier and less stressful than controlling tidal volume at the same time.

As shown in the three representative examples of Figs. 1, 2, and 3, paced breathing was very effective in regularizing breathing activity, which virtually eliminated spectral leakage into the LF band and dramatically changed spectral autonomic parameters. Although these examples provide further support to the need of controlling respiration in cardiovascular variability analysis, controlled-breathing protocols are still far from being widely used and are not mentioned in current guidelines for heart rate-variability measurements (37). Previous conflicting results on the effect of controlled breathing protocols on autonomic cardiovascular regulation have likely contributed to such a situation. Indeed, it has been suggested that voluntary control of breathing 1) enhances the vagal modulation of heart rate (18), 2) leads to a reduced parasympathetic influence in cardiovascular regulation (25), and 3) does not alter vagal modulation (26) or vagal tone (7, 26). These studies, however, used different experimental protocols and laboratory procedures and computed relevant parameters with the use of different algorithms. Moreover, in most studies, the spectral leakage of frequency components of respiration into the LF band was not assessed; therefore, a confounding in the measurement of variability indexes might have occurred. Furthermore, in some investigations, the controlled-breathing protocol was rather complex and cumbersome, which might have caused excessive mental effort and arousal (25).

Effect of paced breathing on respiratory parameters in subjects breathing in HF band.

In subjects with a regular spontaneous breathing in the HF band, spectral leakage of respiration into the LF band was, by definition, very low and became even lower during paced breathing. Interestingly, the inspiratory duty cycle during paced breathing was closer to spontaneous breathing than to the target value of 0.4. This finding has two possible alternative explanations: during paced breathing, either the subjects automatically tended to reproduce the spontaneous Ti-to-Ttot ratio independently of the external auditory input or they simply reacted late to the auditory cue for expiration, thus increasing the inspiratory portion of the breath cycle.

In most subjects, we found that paced breathing induced an increase in tidal volume and minute ventilation with a median percentage change of +18% and +14%, respectively. As shown in Fig. 6, in about half of the subjects, this increase could at least in part be explained by the well-known effect of automatic adjustment of breath amplitude brought about by a reduced breathing frequency (7, 22). It is clear, however, that in some subjects the change in tidal volume was disproportionally high compared with the simultaneous decrease in breathing frequency. Moreover, in most subjects who increased their breathing frequency, tidal volume still increased. Therefore, we are led to conclude that the act of voluntary controlling breathing frequency at around 0.25 Hz is accompanied in most subjects by an increased respiratory drive, which represents the effect of cortical inputs on the spontaneous respiratory pattern generator. A similar trend in the results was found by Bernardi and coworkers (3, 4) in two previous studies.

As a consequence of increased ventilation, ETMath slightly decreased and O2 saturation slightly increased during paced breathing, both with a high statistical significance. As shown in Fig. 7, reductions in ETMath were highly correlated with the increase in minute ventilation. Cooke at al. (7) and Badra et al. (1), using a similar protocol, found an average decrease of ETMath of about −0.45% (P < 0.05) and −0.7% (P < 0.05), respectively. Bernardi et al. (3) found that paced breathing increased O2 saturation on average from 95.4% to 96.1% (P < 0.05). Hence, a slight reduction of ETMath and a slight elevation of O2 saturation represent a consistent effect of paced breathing in supine normal subjects, simply reflecting a condition of mild hyperventilation.

Effect of paced breathing on respiratory parameters in subjects with slow or irregular breathing.

In subjects with slow or irregular spontaneous breathing, paced breathing caused a more pronounced increase in minute ventilation (median percentage change, 42%) but no change in tidal volume. Although the increase in ventilation simply reflected the increase in breathing frequency, the substantial stability of tidal volume was likely due to the effect of two opposite mechanisms. On the one hand, the act of voluntarily controlling breathing brought about an increase in central respiratory drive (as previously observed in group 1 subjects) with a corresponding tendency toward increasing tidal volume. On the other hand, the increase of breathing frequency was accompanied by a physiological tendency toward reducing tidal volume. As a consequence of augmented ventilation, ETMath decreased and O2 saturation increased with changes comparable to those observed in group 1. Hence, even in subjects with slow or irregular spontaneous breathing, paced breathing elicits a mild hyperventilation, which is mainly due to the increase in breathing frequency.

Effect of paced breathing on cardiovascular parameters in subjects breathing in HF band.

In group 1 subjects, the mean R-R interval remained remarkably stable when changing from spontaneous to paced breathing (Table 2), a result in agreement with several previous investigations (1, 4, 6, 7, 14, 32) that used the same breathing frequency and similar or different protocols. A similar remarkable stability was observed in systolic and diastolic pressure, again in agreement with previously published studies (1, 4, 32). These findings suggest that paced breathing at 0.25 Hz neither changes mean vagal and sympathetic nerve traffic to the heart nor increases tonic sympathetic activity toward peripheral resistance vessels.

Paced breathing did not cause any significant change both in the characteristic frequency and in the absolute and normalized power of the LF band of all cardiovascular variability signals. Therefore, it seems likely that, provided the frequency content of respiration is well within the HF band, paced breathing at ∼0.25 Hz does not alter efferent vagal and sympathetic modulations in the frequency range from 0.04 Hz to 0.15 Hz. Similar results were obtained in previous investigations (4, 32) that used the same paced-breathing frequency.

The change induced by paced breathing in the absolute and normalized HF power of R-R interval was very low and largely not significant, suggesting that vagal modulation was not altered by the breathing protocol. The same result was found by Sanderson et al. (32), whereas Bernardi et al. (4) and Pagani et al. (23) found a significant increase in the normalized HF power. We argue that these discrepancies might be due to 1) differences in the protocol and laboratory procedures (Pagani and coworkers, for instance, used a 0.33-Hz-paced frequency, and 7 of 16 subjects breathed through a mouthpiece connected to a spirometer), 2) lack of assessment of spectral leakage during spontaneous breathing [Pagani et al. (23) reported a “marked change in respiratory waveform” during paced breathing], and 3) differences in the experimental context, which might have caused excessive increase in tidal volume in some subjects [Bernardi et al. (4) reported that ventilation increased by 132 ± 36%].

At variance with previous results, the HF power of systolic and diastolic pressure significantly increased during paced breathing. This result most likely reflects the increase in tidal volume brought about by paced breathing, because at frequencies ∼0.25 Hz, the influence of respiration on arterial pressure is mainly mechanically mediated by variations in intrathoracic pressure (33). Sanderson and coworkers (32) found that during paced breathing, the HF power of systolic pressure almost doubled compared with spontaneous breathing, but statistical significance was not reached (32).

Finally, paced breathing did not affect spectral baroreflex sensitivity, in agreement with a previous investigation from Frederiks and coworkers (10), suggesting that sympathovagal balance was not altered by the procedure.

Effect of paced breathing on cardiovascular parameters in subjects with slow or irregular breathing.

In group 2 subjects, paced breathing did not induce any significant change both in mean R-R interval and systolic and diastolic blood pressure, despite the average increase in minute ventilation of 42%. Because in humans the hypocapnia elicited by voluntary hyperventilation causes tachycardia and a fall in total peripheral resistance (20), we are led to exclude that the slight reduction in ETMath observed in these subjects could have had any relevant cardiovascular effect. Similarly, we are led to infer that, within the range tested, the change in breathing frequency (median, 0.06 Hz; range, 0.0–0.14 Hz) did not cause any specific effect on vagal and sympathetic tones.

In conclusion, in this study we examined the respiratory and cardiovascular effects of paced breathing at ∼0.25 Hz both in subjects with a regular phasic breathing in the HF band (median breathing frequency, 0.25 Hz; range, 0.18–0.36 Hz) and in a group of subjects with slow or irregular breathing (median breathing frequency, 0.16 Hz; range, 0.08–0.24 Hz). We found that in both groups this breathing protocol caused a mild hyperventilation. It seems unlikely that these changes could have affected autonomic regulation. Indeed, an examination of standard indexes of autonomic control consistently indicated that paced breathing did not alter tonic vagal and sympathetic outflow. Moreover, the results from subjects breathing spontaneously in the HF band, in whom we could compute spectral indexes of cardiovascular variability, indicated that paced breathing did not cause any significant change in vagal and sympathetic modulations or in baroreflex sensitivity. In these subjects we observed only a significant increase in the HF power of systolic and diastolic arterial pressure signals, most likely brought about by increased tidal volume. According to current knowledge, however, this spectral parameter does not carry any relevant physiological value. Taken together, all these findings support the notion that paced breathing at 0.25 Hz does not alter cardiovascular autonomic regulation compared with spontaneous breathing. This conclusion is valid within the breathing frequencies considered in our study and cannot be extended to lower or higher breathing rates.

Because paced breathing at 0.25 Hz is a simple experimental procedure that avoids confounding and allows a better standardization in the measurement of spectral indexes of cardiovascular variability, it should be considered as a means to improve the informative content and the physiological and clinical value of these parameters. Recent findings from laboratory investigations in healthy humans (10) and clinical trials in pathological subjects (16) provide indirect support of this view. Because of the tendency of subjects to increase tidal volume while performing the paced breathing protocol, we would stress the need of a suitable familiarization session with paced breathing before signal recording with an online check of respiratory parameters.

Footnotes

  • The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

REFERENCES

View Abstract