INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

FREY AND COLLEAGUES (11) RECENTLY argued that tidal flow is an integrated output of the neural respiratory oscillator in the brain stem, which, in turn, reflects the processing of interacting chemo- and stretch-receptor feedback mechanisms in addition to lung and chest wall passive mechanics. This viewpoint is premised on prior findings that inspiratory and expiratory phases are interdependent and are determined by multiple factors, including respiratory mechanics (e.g., Refs. 2, 3, 6, 7, 9, 15, 20), postinspiratory respiratory drive (13, 25, 26), peripheral chemoreceptors (24), and sleeping patterns (23). Accordingly, they proposed that such complex neuromechanical respiratory control is better characterized by considering the entire periodic tidal flow waveform (TFW), as opposed to simple indexes derived from a limited number of points (2, 20), e.g., time to peak tidal expiratory flow-to-expiratory time ratio (TPTEF/TE).

To characterize flows, Frey et al. (11) quantified the harmonic content of the TFW power spectrum in terms of two related spectral shape indexes: 1) k_d, a harmonic distortion index that defines, irrespective of spectral shape, the extent to which a periodic signal deviates from a single sine wave; and 2) s, the exponent of the power law describing the flow spectral energy vs. frequency. The latter index was justified by their finding that the TFW harmonics of both healthy (longitudinally at 1, 6, and 12 mo) and diseased lungs do indeed follow a power law functional form. They also found that 1) s was significantly lower in chronic lung disease (CLD) infants compared with healthy subjects, 2) k_d was significantly increased with maturation in healthy infants between 1 and 6 mo, and 3) s and k_d exhibited less breath-to-breath variability compared with TPTEF/TE in all infant groups and may, consequently, provide more robust noninvasive means to assess lung pathophysiology and maturational changes.

Very low birth weight (VLBW) preterm babies are a challenging population in whom widely accepted noninvasive methods to assess respiratory mechanics and control remain lacking (1, 18). These infants often suffer from significant lung mechanical dysfunction due to their surfactant deficiency, substantial chest wall distortion and asynchrony during breathing, and incomplete airway and alveolar development (8, 17). Moreover, given their prematurity, it is likely that their receptor-feedback mechanisms are immature. Indeed, because of these VLBW infant characteristics, noninvasively deriving tidal flow s and k_d indexes may carry clinical diagnostic value pertaining to respiratory mechanics and control of breathing in this population.

The primary aims of this study were 1) to test whether, like older infants, tidal flow harmonics of VLBW infants also followed a power law functional form, and 2) to determine whether equivalent s and k_d data may be obtained from distal measurements of tidal flow at the chest wall [respiratory inductance plethysmography (RIP); s_RIP and k_d,RIP, respectively] compared with proximal measurements at the mouth [face mask pneumotachography (P-tach); s_P-tach and k_d,P-tach, respectively]. Tidal flows measured distally are desirable for avoiding the potential artificial and variable effects of face mask placement and manipulation on the infant's breathing pattern and, consequently, on TFW indexes, for avoiding the frequently occurring and difficult-to-control air leaks around the mask, and for facilitating less intrusive, longer term measurements.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Subjects

Measurements and Protocol

Airway opening flow (ao) was measured by using a neonatal screen P-tach (Hans Rudolph, Kansas City, MO) attached to a neonatal face mask. Leaks around the face mask were removed by repositioning the face mask until none were apparent. Monitoring of tidal ventilation was obtained by real-time integration of ao. In a few cases, petroleum jelly was used around the mask to ensure that no leaks were present.

Airway opening (Pao) and esophageal pressures (Pes) were measured by using pressure transducers (MP45, Validyne, Northridge, CA). To measure Pes, each infant received a neonatal esophageal balloon catheter (Ackrad Laboratories, Cranford, NJ) inserted so that the balloon was, in the esophagus, at the level of the lower one-third of the trachea. Proper positioning of the esophageal balloon was checked by continuous on-line monitoring of Pes and adjusted until a high correlation (r² > 0.90) between Pao and Pes tracings with the airway occluded was obtained (12, 18).

Linearity of the P-tach was verified, and the frequency response was flat upwards of 12 Hz. Flow, Pao, and Pes were zeroed and calibrated at the beginning of each experiment. RIP tidal ventilation data were estimated from both RC and Abd data, and an absolute calibration was obtained by direct matching to corresponding P-tach flow-volume data, as previously described (8, 17).

All signals were sampled during spontaneous breathing at 100 Hz, monitored on-line (MP 100, BioPac Systems, Santa Barbara, CA), and stored on a computer for later analysis. Data were collected in stretches of 30-60 s. To facilitate quiet sleep, infants were generally fed via a nasogastric tube after instrumentation was complete and before initiation of measurements. At completion of feeding, the nasogastric tube was withdrawn to avoid possible measurement artifacts. Measurements were done with NCPAP and oxygen support discontinued. Respiratory rate, heart rate, and oxygen saturation were continuously monitored at the bedside.

Data Analysis

Tidal flow indexes. The primary aim of this study was to compare three tidal flow indexes from airway opening (TPTEF/TE, k_d, and s) and RIP (k_d and s) tidal ventilation data. Stretches of 10-15 consecutive spontaneous P-tach and RIP TFW data were chosen from each infant for analysis. RIP flow was estimated from the derivative of RIP volume. These data were specifically chosen such that they were free of air leak around the mask (confirmed off line) and free of gross artifact for all measured signals (Pes, Pao, ao, RC, and Abd). We determined TPTEF/TE from the expiration phase of each of the P-tach-measured TFW (2, 20).

Harmonic distortion. The deviation from a pure f_br sinusoid can be quantified as k_d and is defined as the square root of the relative power in the tidal flow above the fundamental or f_br (22, 27). Accordingly, a sine wave would have k_d = 0%, whereas increasingly higher values of k_d would indicate the presence of stronger higher frequency content.

Power law analysis. If the spectral harmonics of tidal flows follow a power law functional form, then the frequency dependence of TFW harmonics will be characterized by a linear decrease as a function of frequency (f) when plotted on log-log scales. This is mathematically defined by the following

(1)

where Eq. 1 is the logarithm of the power law equation ao(f) = A * f^s with A = log(c). Note, a smaller (larger) value of s indicates greater (lower) content of harmonics at higher frequencies, resulting in less (more) single sinusoidal shape of the TFW.

Statistical methods. For each infant, breath-to-breath TFW from the entire flow signal were used to determine TPTEF/TE, f_br (in breaths/min), tidal volume (VT; in ml/kg), and the associated lung mechanical properties. Least squares (MATLAB, The MathWorks, Natick, MA) estimates of lung resistance (RL) and elastance (EL) were derived by multiple-regression analysis. Here, the time (t) domain transpulmonary pressure (Ptp = Pao  Pes) data over the entire breath were described in terms of ao(t), VT(t), and the elastic recoil pressure at end expiration (P₀) via a series resistance-elastance model of the lungs (8, 18); i.e., Ptp(t) = RL · ao(t) + EL · VT(t) + P₀.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

We studied 16 VLBW infants [8 boys and 8 girls, gestational age: 28 wk (median); birth weight: 1,114 g (median)] with varying severities of RDS (with or without apnea of prematurity) between postnatal days 1 and 14 (median = 4 days; Table 1). All babies were on NCPAP support (4-6 cmH₂O) with FI_O2 varying from room air (21%) up to 43%.

Figure 1 shows an example of TFW and corresponding VT data measured directly at the mouth by using a P-tach compared with the RIP-derived flow and volume data. Respiratory rate (median: 87 breaths/min), VT (median: 3.7 ml/kg), and TPTEF/TE (median: 0.61) varied substantially among the infants (Tables 1 and 2). Their estimated lung mechanical properties (RL: median 24.5 cmH₂O · l¹ · s and EL: median 564 cmH₂O/l) varied appreciably also (Table 1).

View larger version (23K): [in this window] [in a new window]   Fig. 1.   Example of tidal flow (top), volume (middle), and ribcage (RC) and abdominal (Abd) band respiratory inductance plethysmography (RIP) data (bottom) from a very low birth weight (VLBW) infant (28 wk, 1,169 g). Top and middle: RIP-derived flow and volume waveforms (dotted lines) are superimposed on corresponding pneumotachograph (P-tach) waveforms (solid lines) to illustrate their time domain similarity. au, Arbitrary units.

A total of 202 individual breaths was analyzed in all patients (10-15 breaths per subject). The overall mean ± 95% CI of TPTEF/TE was 0.62 ± 0.08. We found a consistent power law behavior of the breath-to-breath TFW harmonics in all infants. The averaged TFW spectra vs. frequency in individual subjects are shown in Fig. 2. Importantly, the power law functional form was similar when derived from direct flow measurements at the mouth (Fig. 2, top) or indirectly through VT excursions measured distally at the chest wall (Fig. 2, bottom).

View larger version (27K): [in this window] [in a new window]   Fig. 2.   Averaged P-tach (top) and RIP (bottom) derived flow power spectra shown for each infant (shaded lines) on a log-log plot. Note, for each infant, frequencies were normalized to the corresponding breath frequency (fbr) or fundamental such that fbr = 1. The mean ± SD flow power spectra for all infants are represented by the symbols and error bars.

The mean and 95% CI of TFW indexes derived from all infants and their observed RCV% are summarized in Table 2. The s and k_d, derived separately from measurements at the mouth [s_P-tach 3.73 ± 0.23% (CI) and k_d,P-tach = 38.8 ± 4.6%] and the chest wall (s_RIP = 3.51 ± 0.30% and k_d,RIP = 42.8 ± 4.8%) were of comparable value (Table 2; means of both indexes were similar by unpaired t-test) and were closely correlated (Fig. 3). Yet small but systematic between-method differences in the values of both indexes were statistically significant (paired t-test). Linear regressions illustrating the high correlations between the s (s_RIP = 1.17 × s_P-tach + 0.85; r² = 0.81) and k_d (k_d,RIP = 0.90 × k_d,P-tach + 8.0; r² = 0.76) indexes were estimated independently from RIP vs. P-tach (Fig. 3).

View larger version (20K): [in this window] [in a new window]   Fig. 3.   Breath-by-breath comparison of RIP vs. P-tach derived spectral indexes from all infants (open circles): power law exponent (s) derived from measurements at the chest wall (sRIP) vs. s derived from measurements at the mouth (sP-tach; top) and harmonic distortion index (kd) derived from measurements at the chest wall (kd,RIP) vs. kd derived from measurements at the mouth (kd,P-tach; bottom). Comparison of infant-averaged RIP vs. P-tach derived spectral indexes (shaded triangles) is shown with corresponding linear regression fit (solid line).

Whereas s and k_d did not exhibit any dependence on either infant weight or age, both showed considerable within- and between-patient variations (Fig. 3). The s and k_d variability was similar, irrespective of flow measurement method (RIP vs. P-tach) when quantified in terms of the breath-to-breath or RCV% (Table 2). These results showed relatively lower s RCV% (s_P-tach: 16.2 ± 8.1%; s_RIP: 13.1 ± 2.9%) compared with both k_d (k_d,P-tach: 23.6 ± 13.0%; k_d,RIP: 18.8 ± 5.3%) and TPTEF/TE (P-tach: 22.2 ± 9.5%), whose RCV% were statistically similar (Table 3). The fact that s had the lowest variability was consistent with findings of Frey et al. (11), but, unlike their data in larger infants, TPTEF/TE in VLBW babies did not exhibit significantly greater breath-to-breath variations compared with k_d.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

A primary finding of this study is that tidal flow harmonics in VLBW infants follow a power law functional form similar to that described in older and/or larger infants (11). Also, tidal flow spectra are essentially similar, as are, consequently, their corresponding s and k_d indexes, if derived from direct face mask P-tach flows measured at the airway opening or indirectly via distal RIP (Figs. 2 and 3).

The averaged s (mean s_P-tach: 3.73%) and k_d (mean k_d,P-tach: 38.8%) values of VLBW babies were similar to those reported (11) in older and/or larger babies with CLD (CLD: 3.84 and 44.3%, respectively). Yet the s and k_d of VLBW babies differ appreciably from the corresponding values (4.24 and 26.2%, respectively) reported in healthy 1-mo-old infants (11), despite being of closer age and size.

TPTEF/TE, a simple, noninvasive tidal flow index derived from spontaneous breathing time domain expiratory flow data, has been shown to convey information regarding airway mechanics in infants (2, 3, 6, 7, 20) and has been also linked to inspiratory drive (13, 25) and postinspiratory respiratory muscle drive (26). In our VLBW infants, TPTEF/TE values were noticeably greater than those reported in older infants, whether healthy or diseased. Indeed, when combined with the data of Frey et al. (11), TPTEF/TE data indicate 1) a systematic decrease in this parameter up to ~6 mo postnatal age, and 2) little or no effect of CLD and reactive airway disease on this parameter (Fig. 4). This is rather in contrast to s and k_d values that seemed to differ more with health status, as opposed to maturation. Hence it appears that the time-domain expiratory phase parameter, TPTEF/TE, may provide different and possibly complementary information to that available from the two spectral (or frequency-domain) indexes, s and k_d.

View larger version (18K): [in this window] [in a new window]   Fig. 4.   Mean (±95% confidence interval) time to peak tidal expiratory flow-to-expiratory time ratio (TPTEF/TE; top), s (middle), and kd (bottom) of 5 infant groups of varying size [left; weight (Wt) in kilograms] and age [right; postconception age (PCA) in weeks]. All data points except for VLBW (solid triangles) were based on values reported by Frey et al. (11). Open circles, averaged longitudinal data from 10 healthy infants at 1, 6, and 12 mo after birth (an average PCA of 40 wk at birth was assumed for these babies); shaded squares, averaged data from 10 chronic lung disease infants (postnatal age: 8.1 ± 7.2 mo) who were premature at birth (gestational age: 27.7 ± 2.4 wk).

It has been suggested that significantly lower breath-to-breath s and k_d variability relative to that of TPTEF/TE indicated increased robustness of these spectral indexes, perhaps making them more useful as clinical parameters (11). Our results in preterm infants did not fully duplicate their within-subject RCV% findings. We, instead, found that 1) VLBW babies exhibited greater TPTEF/TE, s, and k_d variability compared with older infants; and 2) RCV% was lowest for s and was similar for k_d and TPTEF/TE. A variety of factors can potentially influence RCV% of TFW indexes. The possibility that the greater RCV% in the VLBW babies data that we present here is a characteristic of prematurity cannot be discounted, particularly given the variations in the underlying disease severity (see Table 1). However, the limited study population and study design do not allow for definitive conclusions.

A notable measurement-related difference between our study and that by Frey et al. (11) is that our infants were studied during quiet sleep but without sedation. The potential effects of sedation on the derived parameters are unknown to us, but such effects are possible, particularly as these indexes vary with sedation-related breathing pattern changes. Breathing pattern may also be modified by each infant's response to the unavoidable manipulation associated with face mask flow measurements. Tolerance of such manipulation may differ in sedated compared with nonsedated babies. These artificial effects attest to the potential benefits of distal or chest wall flow measurements, which our data indicated will provide an equivalent, yet nonintrusive, surrogate to face mask flow data.

Overall, the breathing pattern that we measured in VLBW infants was characterized by a somewhat high breathing rate (f_br) and low VT (in ml/kg). Also, both f_br and VT showed significant within- and between-patient variations (Table 1). As discussed above, these values may have been partly due to the intrusive nature of face mask measurements and may have been tolerated differently among subjects. We examined the effects of varying f_br and VT on s and k_d and found no systematic effects of VT on either index. Alternatively, we found a significant trend of increasing s at higher f_br (r² = 0.36; P < 0.001), a trend similar for mouth and chest wall flow measurements (Fig. 5, top). Then, expectedly, k_d tended to decrease with increasing f_br, albeit less steeply (r² = 0.13; P < 0.001; Fig. 5, bottom). We found no evidence that any of the TFW indexes exhibited age and/or size dependence for the range of values (see Table 1) in the studied VLBW baby population.

View larger version (31K): [in this window] [in a new window]   Fig. 5.   Breathing rate dependence of tidal flow spectral indexes s (top) and kd (bottom). This rate dependence was similar for indexes derived from RIP (open triangles) and P-tach (solid circles) flow data. Linear regression lines (thick solid lines) with 95% confidence (dashed lines) and prediction range (thin solid lines) correspond to all data in graph.

While not reported in their paper (11), Frey and colleagues did not find systematic changes in s and k_d estimates with breathing pattern (personal communication). Importantly, their data also did not include high respiratory rates (>60 breaths/min) and were collected in sedated babies who are less likely to alter their breathing pattern during the measurements. Alternatively, such rates are rather common in preterm infants in respiratory distress, particularly when removed from their NCPAP support and with the placement of the face mask for measurements. Therefore, for respiratory rates characteristic of VLBW babies, changes in breathing pattern may contribute to observed variations in spectral indexes and hence should be considered when interpreting tidal flow s and k_d data in this population.

A potential limiting or confounding factor in our analysis is the fact that the measurements were done in VLBW babies shortly after discontinuing their NCPAP support, with or without supplemental FI_O2 (see Table 1). Moreover, the levels of distending pressure and oxygen support could have varied among infants, as did the time between support interruption and the actual analyzed breath data. Possible consequences of these factors are that measurements may have been done at differing 1) effective lung volumes and 2) oxygenation levels among the different babies. In fact, even for the same subject, the rate of lung volume derecruitment and possible changes in oxygenation after discontinuing NCPAP could have varied, depending on the extent of underlying RDS, level of support, and duration of face mask measurements. Note that babies were not allowed to drop their oxygen saturation <90 at any time during the measurements. One expects that varying lung volume and oxygenation will alter respiratory control (and hence the tidal flow characteristics) through the stretch- and chemoreceptor feedback mechanisms feeding into the respiratory oscillator (4, 13, 16, 19).

An important result of our study is that we found that accurate flow measurements with essentially equivalent flow spectral information are possible by distal flow measurements obtained at the chest wall via inductance plethysmography. To illustrate, this approach would allow characterization of tidal flows in its clinically relevant form, i.e., while NCPAP and FI_O2 support is maintained and airway opening conditions are not artificially altered. Infant breathing is also known to exhibit long-range correlations (5, 10, 14, 21), which means that very long time series of tidal flow indexes such as s and k_d are needed to accurately and more fully characterize their underlying control of breathing. This too is greatly facilitated by the fact that s and k_d can be equivalently derived from nonintrusive chest wall measurements that are less likely to modify breathing pattern, interrupt ventilatory support, or introduce additional and/or altered imposed work of breathing load.

In the smallest VLBW babies, the immature chest wall may lead to excessive chest wall distortion and possibly paradoxical breathing. In such instances, use of RIP flows (particularly if combining RC and Abd signals) to assess TFW spectral characteristics may be problematic. It is unclear whether using RC or Abd movement alone would yield satisfactory results in such situations and should be investigated in future studies. In this study, we did not encounter cases of excessive chest wall distortion, which may have partly been due to the fact that babies were on NCPAP support until shortly before the measurements, as well as because of the relative short duration of data collection.

In conclusion, our study reports for the first time that the harmonic content of tidal flows of spontaneously breathing VLBW infants is characterized by a power law functional form power spectrum. Moreover, we showed that essentially equivalent flow spectra and corresponding shape indexes (s and k_d) may be obtained from direct proximal flows measured at the mouth or indirect distal flows measured at the chest wall. This finding is important, given the desirable, nonintrusive nature of the distal measurements via RIP, which 1) avoid artificial and air leak effects of face mask measurements, 2) facilitate longer measurements, and 3) do not interfere with treatment delivery such as NCPAP and inspired oxygen. When considering our findings together with those of Frey et al. (11) in older and/or larger healthy and diseased infants (Fig. 4), one might speculate that frequency domain tidal flow indexes (s and k_d) potentially provide complementary information to that available from TPTEF/TE. Specifically, s and k_d in VLBW babies with RDS were similar to those reported in older and/or larger CLD infants, whereas they differed appreciably in healthy values (11). Alternatively, TPTEF/TE seemed less sensitive to disease and was decreased systematically up to ~6 mo after birth, which is indicative of being sensitive to maturational changes. Finally, pending further evaluation, s and k_d determined from spontaneous breathing flow data in VLBW preterm infants may provide a useful and fairly practical (if RIP derived) clinical tool to possibly assess changes in their underlying lung mechanics and/or control of breathing. Note that, at least for the VLBW babies in this study, we also found that s and k_d are breathing rate dependent, and future studies should hence determine whether a similar (or modified) dependence is present in healthy and diseased older infants, as well as explore its effects on interpreting s and k_d.