INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

THE MECHANICAL PROPERTIES of the respiratory system are conveniently encapsulated in terms of its mechanical impedance (Z), a complex function of frequency that accounts for both the conservative and dissipative properties of the system. Z is usually determined by the so-called forced oscillation technique in which an external generator, such as a loudspeaker or piston pump, generates a broad-band flow into the lungs via the mouth while pressure is measured at the mouth simultaneously. Input impedance (Zin) is then obtained as the Fourier-domain ratio of pressure to flow (16). Alternatively, a transfer impedance of the respiratory system may be obtained from the Fourier-domain relationship between pressure oscillations applied to the body surface and flow measured at the mouth or vice versa (15). It is also possible to obtain regional lung impedance if the applied pressures and flows can be confined to some local region of the lung, such as the alveolar input impedance provided by the alveolar capsule oscillator technique in dogs (3). Zin, transfer impedance, and alveolar input impedance are not equivalent quantities, but rather give complementary information about respiratory mechanics, each with its own particular advantages.

Thus, in general terms, some Z relating to the mechanical properties of the respiratory system can be obtained from any oscillatory source capable of generating corresponding pairs of pressure and flow signals whose relationships are somehow affected by respiratory system mechanics. To date, virtually all endeavors of this nature have utilized external power sources to generate the necessary oscillations in pressure and flow. This allows the spectral contents of the applied signals to be optimized for the application at hand but also requires a certain amount of instrumentation that may, in some cases, be cumbersome and impractical.

In the present study, we utilize the fact that the beating heart naturally oscillates the lung because of the close juxtaposition between the organs. This results in so-called cardiogenic oscillations (CO) in flow that can be measured at the mouth when the glottis is open. Correspondingly, if the mouth is occluded, the same phenomenon leads to CO in pressure that can be measured just distal to the site of occlusion. The Fourier-domain relationship between these flow and pressure oscillations constitutes a cardiogenic output impedance (Zc), according to the classic notion of a Thevinin equivalent circuit in which an idealized pressure source (due to the heart) acts on a series source output impedance. Exactly what Zc corresponds to physiologically is unknown, but it has the potential advantage of not requiring an external oscillator to produce perturbations in flow. In contrast, whereas Zin does require an external oscillator, its physiological interpretation is well understood. Therefore, the purpose of the present study was to elucidate the physiological interpretation of Zc by investigating how it relates to Zin in normal human volunteers.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

We studied five healthy human volunteers (2 women, 3 men) with no history of lung disease. The study was approved by our Institutional Review Board, and written, informed consent was obtained from each subject.

Resting measurements of Zc and Zin were made first (see below). The subjects then exercised on a cycle ergometer in 3-min stages and from 10 to 50 W (depending on the subject). At the end of each stage, the subject got off the bicycle and immediately sat down at the measurement apparatus, and Zc and Zin recordings were again taken.

Measurement of Zc. Subjects wore a nose clip and sat upright in a straight-backed chair while breathing through a mouthpiece (Fig. 1A). Measurements were made at functional residual capacity (FRC) over a period of up to 10 s during which subjects attempted to relax all respiratory muscles while supporting the cheeks with the hands and keeping the glottis open. During the first half of the measurement period, cardiogenic flow (c) at the mouth was measured with a pneumotachograph (Fleisch no. 1) connected to the mouthpiece. A piezoresistive differential pressure transducer (SensorTechnics HCXPM002D6V) was connected by the shortest possible length of flexible tubing to the two ports of the pneumotachograph. During the final half of the measurement period, the outflow port of the pneumotachograph was occluded, and cardiogenic pressure (Pc) was measured with a gauge differential pressure transducer (Fujikura, FPM-02PG) positioned just proximal to the mouthpiece. c and Pc were low-pass filtered at 50 Hz with 6-pole Bessel antialiasing filters and were sampled at 128 Hz with a 12-bit analog-digital converter (DT EZ-01, Data Translation, Marlborough, MA) before being stored on a PC computer using the LABDAT data acquisition software (RHT-InfoDat, Montreal).

View larger version (12K): [in this window] [in a new window]   Fig. 1.   Experimental setup for cardiogenic oscillation recordings (A) and forced oscillation technique (B). Pressure at the airway opening (P) was measured via a lateral tap near the mouthpiece. Flow () was measured with a pneumotachograph.

Measurement of conventional Zin. Conventional forced oscillatory measurements were made using a 50-ml piston oscillator (Fig. 1B), which generated an 8-s broad-band volume perturbation consisting of the superposition of seven discrete sinusoidal components having mutually prime frequency between 0.5 and 10.25 Hz. The amplitude (A) of each sinusoidal component was chosen according to the formula
where a = 0.1 and b = 0.2. This formula gives roughly equal power to each frequency in applied flow, with somewhat proportionately greater power at high frequencies, in an attempt to optimize the signal-to-noise ratio in the measurements. Piston position was recorded with a linear variable differential transducer (Trans-Tek, model 0244-0000) calibrated in units of volume displacement (ml). The peak-to-peak amplitude of the piston stroke was set to 10 ml so as to generate flows of comparable amplitude to c. Subjects were connected to the oscillator through the same mouthpiece as used for the CO measurements (see above). During the application of the 8-s forced oscillations, the subject remained relaxed and apneic with open glottis at FRC.

Data processing. Data analysis was carried out by using the Matlab 5.3 mathematical software (The Mathworks, Natick, MA). To calculate Zc, we first used the heart sound signal to divide c and Pc into individual heart cycles. Figure 2 shows typical recordings c and Pc, for a subject at rest, together with the heart sound signal that identifies each cycle. Next, c and Pc were divided into individual cycles, and each cycle was resampled by use of linear interpolation to have exactly an integer power of two data points (usually 128). The individual c and Pc cycles were then ensemble-averaged (Fig. 3). Finally, Zc was calculated by taking the ratio of the Fourier transform of the averaged Pc to the transform of the averaged c at those frequencies having power equal to 2% or more of the maximum. This assumes that the Pc and c signals were stationary, i.e., the only difference between successive cycles was random noise. To test the validity of this assumption, we calculated Zc for subject 1 using each possible pair of individual Pc and c cycles. Figure 4 shows the mean and a standard deviation of the individual real parts (Rc) and imaginary parts (Xc) of Zc, and demonstrates that for most frequencies the variation in individual Zc is small, which supports our assumption of stationarity for c and Pc.

View larger version (31K): [in this window] [in a new window]   Fig. 2.   Representative examples of flow and pressure measured at the mouth of a subject sitting quietly at functional residual capacity with open glottis, together with heart sounds recorded over the chest. Arrows indicate time of closure of the valve. Heart sounds units are arbitrary.

View larger version (34K): [in this window] [in a new window]   Fig. 3.   Representative recording of cardiogenic oscillations at rest: heart sounds (A) and mouth flow (B) before valve closure, each overlayed cycle by cycle, and cycle-by-cycle heart sounds (C) and mouth pressure (D) after valve closure.

View larger version (19K): [in this window] [in a new window]   Fig. 4.   Mean ± SD cardiogenic impedances (Zc) determined by using all combinations of heart cycles of cardiogenic flow (c) and cardiogenic pressure (Pc) for subject 1 at rest. Rc is the real part of Zc (resistance), and Xc is the imaginary part of Zc (reactance).

P

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Figure 5 shows Zc and Zin for all subjects studied at rest. In two (subjects 3 and 5), Rc and Rin agree well over most of the frequency range, whereas in the three remaining subjects Rc is markedly lower than Rin. The reactances are quite different in four of the subjects, with Xin being considerably lower than Xc at all frequencies. Only in subject 5 are Xin and Xc similar. The most striking finding, however, is that Xc is essentially zero at all frequencies in all subjects.

View larger version (22K): [in this window] [in a new window]   Fig. 5.   Zc (circles) and input impedance (diamonds) for subjects 1 to 5 (A to E, respectively) studied at rest. Closed symbols are resistance; open symbols are reactance.

Figure 6 shows the effects of exercise, and hence of increased heart rate (which reached levels 20-40% greater than baseline), on Rc. Values of Rc at rest and two exercise levels are shown for all subjects, normalized to the value at rest at 2 Hz. None of the values at any frequency or exercise level are significantly different (ANOVA, P < 0.05). The mean values of Xc at the same frequencies, for all exercise levels, are given in Table 1. None was significantly different from zero (one sample t-distribution, P < 0.05). These results show that increasing heart rate had no effect on the estimated Zc.

View larger version (14K): [in this window] [in a new window]   Fig. 6.   Rc at rest and at low and moderate levels of exercise (mean ± SD for all subjects). Values are normalized to the resting values at 2 Hz.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Each beat of the heart produces an asymmetric deformation of the lung that causes the fluctuations in airway flow (or pressure) known as CO. The amplitude of these oscillations increases with increases in cardiac output and filling pressure (10), both of which increase the mechanical deformation applied to the lung. Conversely, CO in flow are diminished or absent when airway resistance is increased (4). These observations lead to the notion that CO may harbor useful information about heart and/or lung function. Despite this, relatively little has been done to try and extract any such information. A few studies (7, 11, 19, 20) have investigated CO as a means of tracking changes in stroke volume. CO have also been used as a qualitative index for differentiating between obstructive and central apneas during sleep (1, 9, 12). However, in most situations, CO are regarded as a nuisance and attempts are frequently made to eliminate them from recordings of airway opening pressure and flow and esophageal pressure (17, 18). CO have also been associated with troublesome autotriggering of mechanical ventilators (10).

Our goal was to investigate the use of CO in airway flow and pressure as means for determining a mechanical output impedance related to lung function. This involved first obtaining and processing Pc and c in an appropriate way and then interpreting the resulting Zc in physiological terms. The data acquisition and processing steps involved a number of assumptions. First, we assumed that the oscillations in c measured before airway occlusion corresponded to the oscillations in Pc measured afterward and that each bore the same phase relationship to the heart sounds measured throughout the recording period. This assumption seems reasonable because we do not expect that occluding the airway of a relaxed apneic subject will have any immediate effect on the activity of the heart. Another issue is that a period of apnea would be expected to cause a progressive change in cardiac activity, and possibly lung mechanics, as a result of neural effects and changes in blood gases. There was a slight decrease in heart rate throughout the breath-hold maneuver, but it was small enough to be considered negligible, and the individual cycles of Pc and c were highly reproducible (Fig. 3). Furthermore, when Zc was calculated with different combinations of the individual cycles in Pc and c, the resulting Zc showed good reproducibility (Fig. 4).

CO have interested physiologists for a long time, particularly in terms of their manifestations in expired gas concentrations (4). CO in flow measured at the mouth are probably due to a direct action of the heart on the lung parenchyma (4), although exactly how is complicated and still somewhat obscure. The heart has an irregular shape and contracts with a twisting action, producing localized transient inflations and deflations of the lung. This produces transient redistribution of gas throughout the lung (4), so there is no simple relationship between movement of the heart and the size of the CO in flow seen at the mouth. Also, the total heart volume varies very little throughout its beat (8), so the CO in flow presumably have little to do with stroke volume. It seems inevitable that CO in both pressure and flow should be modulated by anatomic or physiological factors (5, 11, 19), such as exercise, lung volume, and body posture. In the five subjects we studied, the peak-peak swings in Pc at rest were 0.377 ± 0.073 cmH₂O. During the two levels of exercise, these swings were 0.423 ± 0.227 and 0.492 ± 0.297 cmH₂O, respectively. The corresponding swings in c were 0.0350 ± 0.0278 ml/s at rest and 0.0346 ± 0.0376 and 0.0314 ± 0.0287 ml/s, respectively, at the two levels of exercise. The magnitudes of the CO in both pressure and flow were, thus, perhaps surprisingly, essentially unaffected by the exercise even though there was considerable variation between subjects. We also observed differences in the morphology of the oscillations in the same subject when repeated measurements were made on different days.

An important aspect of our study was thus to ascertain the extent to which Zc is reproducible with variations in the nature of the heartbeat. We investigated this issue by determining Zc at rest and after low and moderate exercise (we did not attempt severe exercise because the subjects had to remain relaxed and apneic for 8 s immediately afterward for the measurements to be made). The subjects were all young and healthy, so minimal exercise-induced bronchodilation and changes in respiratory mechanics were expected. We found that Zc was highly reproducible regardless of the changes in heart rate and stroke volume produced by the exercise (Fig. 4), which supports the notion that Zc does indeed reflect a well-defined quantity.

To elucidate the physiological interpretation of Zc, we compared it to Zin measured by the forced oscillation technique. Obviously, we could not determine Zin at exactly the same frequency as the heart rate using the same amplitude of the flow as the CO because the very presence of the CO would have interfered with the determination of Zin. Therefore, we identified Zin over a range of frequencies bracketing the heart rate and its first few harmonics. We also used flow amplitudes that were similar to those of the CO in an attempt to make Zc and Zin comparable, because Zin is known to depend to a certain extent on flow amplitude (16). However, this resulted in a poorer ² than would have been obtained with a larger amplitude flow perturbation. We attempted to achieve a compromise between requiring an adequate ² and retaining sufficient data to make an effective comparison between Zin and Zc by setting our acceptance cutoff at ²  0.90.

The most striking aspect of the comparison between Zin and Zc (Fig. 5) is that Xc is essentially zero at all frequencies whereas Xin is mostly negative and increases with frequency. Some of the subjects (subjects 3 and 4) exhibited quasi-hyperbolic Xin as described in numerous previous studies (16) whereas the remaining subjects had Xin that varied less with frequency. These differences may be due to differences in the degree of relaxation among the subjects, because any respiratory muscle tone would be expected to have a large effect on respiratory tissue stiffness and hence on Xin. Nevertheless, the absence of any appreciable Xc indicates that Zc was essentially purely resistive. This conclusion is further supported by the observation that Xc remained essentially zero despite increases in heart rate (Table 1).

The agreement between Rc and Rin was good in subjects 3 and 5 (Fig. 5), whereas in the remaining three subjects Rin was uniformly higher than Rc. One possibility for such discrepancies is that the spectral content of the applied flow oscillations used to obtain Zin was not the same as that producing Zc. In particular, the harmonics of c fell off quickly with increasing frequency, whereas the components of flow used to produce Zin did not. Because respiratory impedance depends on many factors, including flow amplitude (6, 14, 20), it is possible that imperfect matching of measurement conditions could have accounted for the discrepancies between Rc and Rin. However, it is also possible that Rc and Rin reflect different quantities. Rin is a measure of the resistance of the total respiratory system and, even at high frequencies, contains components from both the airways and the chest wall tissues (2). The fact that Zc is purely real suggests that Rc is a measure of an airway flow resistance only, without any contribution from the respiratory tissues (which are viscoelastic and would give rise to a significant reactive component to Zc). Therefore, we speculate that Rc is a measure of the resistance of the central and upper airways. In other words, when the mouth and glottis are open, the beating heart acts as a flow generator by compressing the lung tissue at the base of the airway tree. This produces c at the mouth. When the mouth is occluded, the pressure oscillations producing c are transmitted to the mouth to produce Pc. The ratio of Pc to c then yields a measure of the output impedance of the respiratory system, which in this case is the flow resistance of the conduit between the site of origin of the oscillations and the measurement point at the mouth.

It is perhaps somewhat curious that there is no appreciable imaginary part to Zc. Given that the movement of the heart is known to produce local redistribution of flow in the lungs (4), one might expect there to be some effect of a local shunt compliance. In such a situation, the imaginary part of Zc would reflect this compliance. It is possible that the shunt compliance was small and not discernable above the noise in our measurements. Interestingly, however, the real parts of Zc tend to decrease with frequency (Figs. 4 and 5). The only way this can happen, barring nonlinear effects, is if there is a reactive component to the system. Some of the Zc imaginary parts in Fig. 5 have a tendency to increase very slightly with frequency. Although these trends are not significant, they may hint at a small finite reactive component hidden in the noise.

As a further test of this interpretation of Zc, we repeated its measurement in two of our subjects before and after the addition of an external resistance between the mouth and the pressure and flow transducers. The resistor had a value (1.5 cmH₂O · s · ml¹) comparable to that of normal human airway resistance. Four measurements were made under resting conditions in each configuration. In one subject, the real part of Zc at 2 Hz with the added resistance had a mean value of 7.29 ± 0.89 cmH₂O · s · ml¹, whereas without the added resistance the mean value was 5.32 ± 1.13 cmH₂O · s · ml¹. In the other subject, the mean value was 4.94 ± 0.84 cmH₂O · s · ml¹ with the added resistance and 3.85 ± 1.29 cmH₂O · s · ml¹ without the added resistance. The added resistance thus made a difference in the real part of Zc for the two subjects of 1.97 and 1.09 cmH₂O · s · ml¹, respectively. These differences are not precisely equal to the added resistance of 1.5 cmH₂O · s · ml¹, but this could easily be due to slight differences in measurement conditions (e.g., lung volume, glottic aperture) between the measurements made with and without the added resistor. However, the values of resistance obtained with the added resistance bracket that of the added resistance itself, which supports the notion that Zc gives a measure of the resistance due to flow of gas through airways.

To be able to measure CO in Pc and c reliably, it was necessary for our subjects to remain relaxed with an open glottis throughout the measurement period. This is not an easy thing to do and requires a considerable degree of subject cooperation. Untrained subjects either tend to close the glottis shortly after they suspend breathing or cannot discern whether the glottis is open or closed (6, 13). This problem is almost certainly exacerbated with dyspnea after exercise. Our subjects became quite practiced at sitting relaxed with open glottis, and the measurements of Zin and Zc were made as close together in time as possible. We also observed that the signals measured during the first 4 s of the forced oscillation measurements were similar to those obtained during the second 4 s, which suggests that glottic aperture was consistent during these measurements. Nevertheless, it is still possible that our comparisons of Zin and Zc were affected by differences in glottic aperture.

In summary, we calculated Zc between ~1.5 and 10 Hz using CO in airway pressure and flow measured in relaxed normal subjects at FRC with open glottis. Zc was insensitive to heart rate changes induced by exercise and had an imaginary part close to zero at all frequencies investigated. Rc was similar to or smaller than Rin determined by the forced oscillation technique. We speculate that Zc provides a measure of the flow resistance of the central and upper airways of the lung. Zc may thus be useful as a means of obtaining information about lung mechanics without the need for an external source of flow perturbations.