Main

In the adult, there is evidence of considerable variation in components of the respiratory cycle with time (1–3). Cardiac-related changes in airflow have also been described (4–7). Despite this, there have been some studies showing a similarity between successive breaths in each individual subject but with differing patterns between subjects (8, 9).

In studies of lung mechanics in the newborn infant, it is often assumed that there are similarities in airflow between successive breaths and that it is only necessary to measure parameters in a small number of breaths to characterize these for an infant (10). However, a number of studies in newborn infants have shown considerable variability in breathing patterns in successive breaths, particularly during expiration (11–13).

It was, therefore, decided to study respiratory airflow in term newborn infants during quiet sleep in the first week after delivery, to examine the nature and degree of variability found in the patterns of airflow, and, in view of the findings in adults, to ascertain whether cardiac contraction contributes toward this variability in the newborn.

Dahlstrom et al. (5) showed that in adults during breath holding with an open glottis, there was an outward movement of gas with each cardiac diastole and an inward movement with each systole; they ascribed this phenomenon to cyclic changes in intrathoracic gas volume. It was, therefore, decided to study in addition a smaller number of infants during periodic breathing to examine the nature of airflow during the apneic intervals.

If cardiac contraction was found to influence airflow in the newborn infant, this could have implications for the measurement of lung mechanics in these babies.

METHODS

Subjects.

Twenty term infants of gestational age 37–41 wk were studied in the Neonatal Research Unit of the London Hospital during the first week after delivery. These infants were selected from a larger number of infants solely on the basis of whether they showed regular breathing for a continuous period of 240 s during quiet sleep while breathing room air, with no movement artifacts or sighs. All data satisfying these criteria were included in the analysis, no breaths being excluded. Episodes of quiet sleep were determined using the criteria of Prechtl and Beintema (14). Details of the infants are shown in Table 1. One infant (087) was born by cesarean section, all the others having vaginal deliveries. All the infants had Apgar Scores of 9 or 10 at 5 and 10 min after delivery. They had all been examined by a specialist neonatologist to confirm that they were appropriately developed for gestational age.

Table 1 Details of 20 babies and results * Postnatal age. † Time to peak tidal expiratory flow/TE. ‡ Integrated spectral power over range of frequencies for resp. (square-root scaling). § Integrated spectral power over range of frequencies for heart rate (square-root scaling). ¶ Phase and coherence between flow and ECG cosine wave (see text). ** Mean ± SD.
Full size table

No episodes of periodic breathing were present in the 20 infants being studied, and it was decided to use the data from five additional term infants. Two of these had periods of regular breathing during quiet sleep while breathing room air followed by episodes of spontaneous periodic breathing. The remaining three babies were from a study in which periodic breathing had been induced by hypoxia (15, 16). The Hospital Research Ethics Committee approved this study in which term infants were exposed to 15% O2 in the trunk plethysmograph for a period not exceeding 5 min or less if the Po2 measured by either of two transcutaneous Po2 electrodes fell below 40 mm Hg.

Periodic breathing was defined as bursts of breathing occurring at regular intervals separated by periods of apnea lasting at least 4 s, persisting for at least 3 min. Details of these five additional infants are shown in Table 2. One infant (201) was born by cesarean section, the other four having vaginal deliveries. All had Apgar Scores of 9 or 10 at 5 and 10 min after delivery and were appropriately developed for gestational age.

Table 2 Details of five babies with periodic breathing; RMS values of volume and flow during regular breathing and during periodic apneas * Postnatal age (h). † Root-mean-square of VT (cm3) during regular breathing and periodic apneas. ‡ Root-mean-square of airflow (cm3/s) during regular breathing and periodic apneas. § Mean ± SEM.
Full size table

Informed consent was obtained from all 25 mothers, and approximately half of them chose to be present during the recordings.

Recordings.

The ECG was recorded using a model 8811A Hewlett Packard preamplifier. Esophageal pressure was routinely measured using an air-filled balloon located at approximately the junction of the lower third and upper two-thirds of the esophagus (17) by using a Hewlett Packard differential pressure transducer type 267B, but these data were not used in the present analysis. Tidal breathing was recorded using the trunk plethysmograph (17–19) with a Hewlett Packard pressure transducer type 270. Tests for leakage in the system were performed at the beginning and at intervals during the recording, and the plethysmograph was calibrated for each infant at the end of the recording session. The characteristics and use of this particular plethysmograph have been fully described by Lewis (17). In summary, it had an empty volume of 27.5 L and was constructed of an aluminum-magnesium alloy to allow rapid equilibration with room temperature. The response time to the bursting of a balloon was approximately 10 ms to achieve a full-scale deflection. The small volume, especially with a baby and its bedding inside, and the high conductivity to heat of the plethysmograph ensured that adiabatic effects were minimized.

Analysis.

The data were recorded on a frequency-modulated tape recorder (TEAC). Data from some of these infants had been used in previously reported studies on heart rate variability and the ventilatory response to hypoxia (15, 20, 21). For the present study, all the data were redigitized using the Cambridge Electronic Design system (Cambridge, UK); this incorporates a 16-bit analog-to-digital conversion system with a 1-MHz internal frequency source. The ECG was digitized at 1000 Hz and tidal breathing at 100 Hz. No filtering was performed during the recording and digitization stages.

A computer program was written to calculate the intervals between the onset of successive QRS complexes of the ECG with a resolution of 1 ms by using the algorithm of Okada (22); these R-R intervals were used to calculate instantaneous heart rate sampled at 100 Hz (20). The tidal volume (VT) data were analyzed by a specially written computer program that used a low-pass Dolf-Chebyshev filter (23) set to pass all frequencies lower than 3.0 Hz without any attenuation of the signal to remove high-frequency noise. It was, thus, possible to determine with accuracy the onsets of inspiration and expiration for each successive breath. Airflow, sampled at 100 Hz, was calculated using a computer program written to differentiate the VT data without any shift in time.

For spectral analysis, any linear trend was first removed from the data for VT and airflow. When spectral analysis is carried out on a finite set of data, there is danger of both ends of the data producing spurious contributions to the spectrum (spectral leakage); this is avoided by “windowing” the data so that the beginning and end of the data fall smoothly to zero. This was carried out by applying a cosine taper to the first and last 5% of the data. Auto- and cross-spectral analyses were then performed (24, 25). The frequency resolution of the spectrum was 0.1 Hz.

The area under the curve in a power spectrum is equal to the variance (SD2) of the data with the ordinate showing the contribution of each frequency band to the total variance. The contribution of a range of frequencies to the total variance (spectral density over a specific frequency range) was calculated by summing the values included in this frequency range before taking square roots to convert them to their ordinary units. For cross-spectral analysis between flow and the ECG, the data were converted to zero mean and unit SD, therefore being dimensionless.

To avoid plotting the spectra in units for variance (cm6 for VT and cm6/s2 for flow), it was decided to plot them with the ordinate on a square-root scale, i.e. using the units cm3 for VT and cm3/s for flow, thus illustrating the relative magnitude of the spectral peaks in conventional units. The SD of each set of data has been shown on each graph of a spectrum.

Spectral analysis of the actual ECG recording showed a main peak at the heart rate (Fig. 3) but with up to 15 harmonics of gradually diminishing amplitude at higher multiple frequencies as the Fourier analysis attempted to reproduce the original waveform. For cross-spectral analysis, this difficulty was overcome by creating a new time series for each infant with a cosine wave being interpolated between the positions of the onset of each R wave; each peak in this time series, thus, coincided with each R wave of the ECG. The spectrum of this derived “ECG cosine wave” had a single peak at the same frequency as the fundamental peak of the spectrum of the original ECG but without the multiple harmonics of the ECG spectrum shown in Figure 3.

Figure 3
figure 3

Spectral power for first 60 s of data for ECG, VT, and airflow for baby 067 with square-root scaling on the ordinate. Note:x axis for ECG is from 0–30 Hz; for the remaining spectra, x axis is from 0–5 Hz.

Full size image

Statistical methods.

The Stata statistical package (26) was used for summary statistics, ANOVA, and multiple regression. The null hypothesis was rejected if p < 0.05.

RESULTS

Variables of the respiratory cycle.

The inspiratory VT, durations of inspiration (TI) and expiration (TE), and duration of each breath (TTOT) were calculated for each breath of each of the 20 infants over the whole 240 s of data, all breaths being included in the analysis. The overall mean values and coefficients of variation (CV) of these variables were as follows: VT 14.1 cm3 (CV 8–26%), TI 0.73 s (CV 8–23%), TE 0.98 s (CV 7–25%), and TTOT 1.71 s (CV 5–18%).

Figure 1 shows the inspiratory and expiratory airflow in the first 20 breaths of the first infant studied. Considerable variations in the shapes of both inspiratory and expiratory flow can be seen. Similar variations in airflow between successive breaths were found in all infants.

Figure 1
figure 1

Flow data for the first 20 breaths in baby 065, inspiration upward.

Full size image

The time to peak tidal expiratory flow, expressed as a proportion of TE (TPTEF/TE), was calculated for each breath in each infant. As would be expected from Figure 1, there was a wide scatter in the results. The mean values and variability of the ratio TPTEF/TE for each infant for the whole 240 s of data are shown in Table 1. The mean for the 20 infants was 0.42 [95% confidence interval (CI) 0.36–0.49]. The mean within-subject CV was 31% (95% CI 25–37%). Table 1 also shows the mean respiratory and heart rates for each infant calculated over the whole 240 s.

Figure 1 suggested the presence of small perturbations in airflow distorting the smooth pattern of airflow into and out of the lungs. It was, therefore, decided to perform spectral analysis of the data to determine the frequencies of these variations in airflow.

Spectral analysis.

Figure 2 shows the first 12 s of data for the ECG, VT, and airflow in the second infant studied. Although the shapes of the changes in VT are relatively smooth, small higher-frequency components may be seen distorting the shape of inspiratory and expiratory airflow.

Figure 2
figure 2

First 12 s of data for baby 067. From above downward:1) ECG;2) VT, inspiration upward;3) airflow, inspiration upward. Note: the VT show changes about the mean, without a true zero; the horizontal line for airflow is a true zero.

Full size image

Figure 3 shows the spectrum of each of these variables for the first 60 s of data in the same infant as in Figure 2 with the ordinates plotted on a square-root scale. The spectrum of the ECG has been plotted up to 30 Hz to illustrate the multiple harmonics found in this type of data, whereas the remaining spectra have been plotted from 0.0 to 5.0 Hz. The mean respiratory rate for this infant during the first 60 s was 33.1 min−1, i.e. 0.55 Hz, the mean heart rate being 107.5 beats per minute, i.e. 1.8 Hz. It will be seen in Figure 3 that there is a large peak at this respiratory rate (0.55 Hz) in each of the spectra for VT and flow, with a smaller peak at the first harmonic of this respiratory rate (1.1 Hz) and small peaks at higher frequencies including one at 1.8 Hz, the mean heart rate, in both spectra. To characterize these peaks in the spectra, two histograms were calculated for each 60 s of data:a) a histogram of the reciprocals of TTOT (i.e. the variations in respiratory rate) and b) a histogram of the reciprocals of the R-R intervals of the ECG, i.e. variations in heart rate in this section of data. Figure 4 shows the results of superimposing these two histograms on the spectra for airflow in two different infants, one with a relatively high mean heart rate (above) and the other with a lower mean heart rate (below). It will be seen that the main peak in each spectrum coincides with the histogram of respiratory rate and that the histogram of heart rate variations coincides with one of the subsidiary peaks at a higher frequency. Similar results were found in all the infants.

Figure 4
figure 4

Spectral power for airflow for first 60 s of data for baby 076 (above) and baby 128 (below) with square-root scaling. The left histogram in each case is that of the reciprocals of TTOT for this section of data, and the right histogram is that of the reciprocals of the R-R intervals of the ECG for this section of data.

Full size image

To quantify these results, integrated spectral power for airflow was calculated over the range of frequencies found for a) breathing and b) heart rate; these were calculated for each successive 60 s of data, thus providing four estimates for each infant. The power was then transformed by taking square roots for the results to be in the original units, cm3/s. ANOVA showed that there were no significant differences in respiratory rate, heart rate, power over the range of frequencies for breathing, and power over the range of heart rate among these four successive intervals of 60 s in each infant. The mean results of the whole 240-s period in each infant were, therefore, calculated and are shown in Table 1.

Table 1 also shows that the ratio of the mean power at the heart rate to the mean power at the respiratory rate for all the infants was 2.76/16.15, i.e. 0.171. This implies that the variations in flow due to the heartbeat were about one-sixth of the tidal airflow. Multiple regression of dependent variables (power at the heart rate, power at the respiratory rate, and their ratio) against the variables body weight, postnatal age, mean respiratory rate, mean heart rate, and VT was performed. This showed that power over the range of respiratory rates was directly related to mean respiratory rate (p < 0.01) and VT (p < 0.01), but none of these variables affected the power found at the heart rate or the ratio of power at the heart rate to that at the respiratory rate.

Time relationship between ECG and changes in airflow.

The temporal relationship between the ECG and the changes in airflow at the heart rate was then examined. Cross-spectral analysis was performed between airflow and the ECG cosine wave (see “Methods”) to determine the frequencies shared in common between airflow and the heart rate. Figure 5 shows the results in the same infant shown in the lower part of Figure 4. The shape of the spectrum for airflow remains the same; the spectrum for the ECG cosine wave shows a main peak at 1.7 Hz, the same frequency as the center of the histogram for R-R intervals shown in the lower part of Figure 4. The cross-spectrum shows a main peak corresponding to the heart rate at 1.7 Hz. At this frequency, the phase was 0.93 of a cycle. This value for phase implies that the peak in airflow at the cardiac rate in this infant occurred at an average of either 0.93 of a cycle at this frequency before the R-wave or at 0.07 (1.00–0.93) of a cycle after the R-wave. The coherence at this frequency was 0.54; as coherence is analogous to the correlation coefficient (r) squared, this corresponds to a value of 0.73 for the correlation between flow and heart rate at this frequency. The small peak at 0.6 Hz in the cross-spectrum reflects the respiratory sinus arrhythmia in this infant.

Figure 5
figure 5

Data for airflow and ECG cosine wave (above), their spectra (lower left and middle), and their cross-spectrum (lower right) for first 60 s of data for baby 128 with square-root scaling (normalized, dimensionless units).

Full size image

The results for all the infants are shown in Table 1. It will be seen that there was considerable variation in mean phase between different infants. Multiple regression was then performed with phase as the dependent variable and with weight, postnatal age, mean heart and respiratory rates, TI, TE, TTOT, and maximal inspiratory and maximal expiratory flow rates as the independent variables. The results of this analysis showed that four of these independent variables had a significant effect on phase. The magnitude of phase was positively related to TE (p < 0.01) and inversely related to TTOT (p = 0.01), maximal inspiratory flow rate (p < 0.05), and maximal expiratory flow rate (p < 0.05). All the other independent variables had no significant effect upon phase. A similar multiple regression for coherence showed no significant results for all the independent variables.

Periodic breathing.

To ascertain whether there were indeed changes in respiratory airflow in the absence of breathing movements, it was decided to analyze flow during the apneic intervals found during periodic breathing in the group of five infants detailed in Table 2. Figure 6 shows 12 s of data during an episode of periodic breathing in one of these additional infants with spontaneous periodic breathing. Small perturbations in airflow are noticeable during the apneic interval. Figure 7 shows the data for airflow over the central 7 s of this apneic interval in Figure 6 together with the power spectrum and histogram of reciprocals of the R-R intervals. Despite the instability of the power spectrum over such a small amount of data, this Figure nevertheless shows changes in airflow occurring at approximately the same frequency as the heart rate.

Figure 6
figure 6

Twelve seconds of data for baby 201 during spontaneous periodic breathing to show apneic interval between two bursts. VT has no true zero; airflow has a true zero.

Full size image
Figure 7
figure 7

Flow data for baby 201 over the central 7-s period of apnea shown in Figure 6 (above); spectral power with square-root scaling of the ordinate (below). The histogram is that of the reciprocals of the R-R intervals of the ECG during this 7-s period.

Full size image

Because of this difficulty of performing spectral analysis on short sections of data, the mean values of the root-mean-square (RMS) of the oscillations in tidal volume and airflow during each apneic interval lasting 4 s or more were calculated for each of these five infants. The corresponding RMS values of VT and airflow during successive 10-s intervals of regular breathing during quiet sleep in the same infant were also calculated; the mean results are shown in Table 2. There did not appear to be any difference in the findings in the two air-breathing infants compared with the three breathing the hypoxic gas mixture. The mean RMS values of the pseudosinusoidal oscillations in VT during apnea were approximately one-eleventh of those occurring during regular breathing (0.34/4.44), whereas mean airflow during the apneic intervals was between one-fifth and one-sixth of the amplitude of airflow oscillations during regular breathing (3.15/17.48).

DISCUSSION

The findings in this study indicating an effect of cardiac contraction on airflow in the newborn infant raise a number of questions regarding the measurement and interpretation of changes in airflow in the newborn.

Variability of TPTEF/TE.

There is an increasing interest in the ratio of the TPTEF to TE in studies of lung mechanics in newborn infants, and a number of studies have used the mean of approximately 10 breaths in its calculation, e.g. Stick et al. (10). Using inductance plethysmography, they found a mean within-subject variation of TPTEF/TE of 13% (range 6–21%), far lower than in the present study. A number of studies have demonstrated the presence of expiratory braking and retarded airflow in newborn infants in a significant number of breaths (11, 13), but this problem is not discussed in many recent papers. Stocks et al. (12), in a large study, have measured TPTEF/TE under a wide variety of conditions. They found that by 10 breaths, most infants were within only 25% of their true mean. In those infants studied during the first 2 wk after delivery, they found a mean value of 0.49 ± 0.11 SD, with a wider within-subject variability than in 6-wk postnatal age infants. They concluded that this wider within-subject variability shortly after delivery may limit the usefulness of this variable. The value for TPTEF/TE of 0.42 (95% CI 0.36–0.49) found in the 20 infants in the present study does not differ significantly from their estimates.

These effects of cardiac contraction on airflow can theoretically be removed from flow data, but most filtering methods now in use also reduce the amplitude of the respiratory signal.

Studies in adults relevant to the findings.

Luisada (4) in 1942 documented many reports, some dating from the last century, on changes in airflow and pressure induced by the heartbeat. Using a low-pressure transducer and also a Fleisch pneumotachograph, he described in detail the features of the “pneumocardiogram.” These findings of cardiac-related changes in airflow in the adult have been confirmed in a number of studies (5–7).

Itoh et al. (27), using the impedance method of measuring breathing in adults, described three peaks in the spectrum of the signal: one at the respiratory rate (fR), one at its first harmonic (2fR), and an independent peak at the fundamental frequency for the pulse or heart rate (fP). They estimated this latter volume as up to 100 mL for an adult (this would be about one-sixth of a VT of 600 mL). This is of the same order of magnitude found in the present study.

Studies in the newborn infant.

I have been unable to find any studies of changes in airflow in relation to the cardiac cycle in the newborn infant. There may be a number of reasons for this:a) the presence of these perturbations in airflow may be considered artifacts or noise in the data;b) the frequency response of the equipment used to measure breathing or airflow: for example, Lewis (17) demonstrated that the use of a pneumotachograph attached to a face mask resulted in the signal being delayed compared with that from the trunk plethysmograph, with serious consequences for the measurement of dynamic compliance;c) failure in many reported studies to detail the specifications of commercially obtained software packages, especially in regard to any filtering used;d) the method of measuring breathing: the use of a face mask or nasal prongs and a pneumotachograph can alter the pattern of breathing either by stimulation of the trigeminal region, by increasing dead space, or by increasing the work of breathing (28–30). Thoracic impedance pneumography ignores the contribution of diaphragmatic contraction to breathing in the newborn. The inductance pneumograph (Respitrace) is difficult to calibrate accurately in the newborn infant; the whole body plethysmograph, usually constructed of perspex and having a large volume, is complex thermodynamically, and the nonlinear effects may be within the range of the breathing frequencies. The trunk plethysmograph used in the present study, by contrast, enables the infant to breath room air without any interference with the external airways or any change in dead space. Adiabatic effects are minimized by the large surface/volume ratio and the high conductivity to heat of the metal construction.

Spectral analysis.

This measures the average frequency components found in data over a period of time and is not ideal for identification of the precise temporal relationships between airflow and instantaneous changes in the heart rate. Pichot et al. (31) have shown the superiority of the discrete wavelet transform over the Fourier transform in quantifying heart rate variability, but as yet no method has been suggested for examination of instantaneous relationships between two time series. Examination of the data in the present study showed that the R-wave of the ECG occurred at apparently random times during the inspiratory and expiratory parts of the respiratory cycle, and it cannot be assumed that the cardiac effect on airflow is independent of the phase of the respiratory cycle. For example, right and left ventricular ejection times during spontaneous breathing in the newborn have been shown to depend on the phase of the respiratory cycle (32).

There is also the problem of the harmonics of the fundamental respiratory frequency that appear in the spectrum. This can be a problem when the heart rate is an exact multiple of the respiratory rate; some of the infants in this study showed such multiples over short intervals of time but the majority did not, and, in all cases, the frequency due to the heartbeat, identified by the reciprocals of the R-R intervals of the ECG, coincided with an independent peak in the power spectrum.

Possible explanations for the present findings.

These include the following:a) Direct pressure of chambers of the heart on adjacent lung tissue, but, in their study, West and Hugh-Jones (7) also found the effect at sites distant from the heart and even at the mouth. b) The respiratory pump: with the decrease in intrathoracic pressure during inspiration, there is an increase in flow from the superior and inferior vena cavae into the right atrium and, hence, the thoracic cavity. c) During right ventricular ejection, the pulmonary vasculature becomes engorged with blood. d) After left ventricular contraction, a large proportion of the cardiac output will leave the thoracic cavity. All of these factors may affect lung stiffness or pressure differences within and outside the airways and, hence, compliance at different phases of the respiratory cycle. e) The skull of the newborn infant being very compliant, it is possible that with each cardiac contraction an increased amount of arterial blood leaves the plethysmograph for the head, in excess of venous return from the head. This would also apply to impedance and inductive plethysmography. However, these cardiac-related changes in airflow are also found in adults whose skulls are far less compliant and using different methods of measuring airflow.

Some of these questions could be answered in the newborn infant by noninvasive methods of measuring changes in flow of blood through the great vessels entering and leaving the thoracic cavity in relation to the phase of the respiratory cycle. An example of such an approach is the study by Toska and Eriksen (33) in adults in which they have shown that stroke volume is inversely related to heart rate and that both are affected by the phase of the respiratory cycle (the respiratory changes in stroke volume persist even after the abolition of heart rate variation by the administration of atropine).

Possible significance of the present findings.

It is important that the present findings be confirmed by other workers using other methods of measuring airflow. If confirmed, these findings have implications for studies of airflow and lung mechanics in the term newborn infant: breaths should not be excluded because they are not typical or show artifacts, and many successive breaths need to be analyzed to obtain reliable estimates. Stocks et al. (12) concluded that shortly after delivery, this may not be possible.

In addition, cardiac function may have to be taken into account when assessing respiratory function in the newborn infant. It would also be of importance to examine changes in airflow in preterm and term infants with patent ductus arteriosus and various types of congenital heart disease; this cardiac effect on airflow may have a different magnitude and significance in such infants than in the healthy term newborns described in this study.

An interesting finding is that the time delay between the changes in flow and depolarization of the cardiac ventricular muscle (QRS complex) varied considerably between the different infants and that some of this variation could be accounted for by differences in TE and TTOT and the peak rates of inspiratory and expiratory flow in the different infants. This suggests that the interactions between respiratory airflow and cardiac contraction are likely to be highly complicated and nonlinear. It would be useful to perform a fuller analysis of this type of data to characterize the mathematical dynamics of airflow; this could have important implications for the control of breathing and cardiac output in the newborn infant, both during regular and periodic breathing.