METHODS

Subjects.

Female domestic piglets that were 1 day old and weighed 1.6–1.9 kg (mean 1.76 kg) were sedated by intramuscular injection of 8 mg azaperon, anesthetized by intravenous injection of 10 mg metomidate, and endotracheally intubated (tube diameter 3.5 mm). Anesthesia was maintained by continuous infusion of pentobarbital sodium as needed to suppress spontaneous movements and respiratory efforts, together with 5 ml/h electrolyte solution with 10% glucose. Body temperature was kept between 36 and 37°C by placing a heating pad under the animal and a radiator above it. The experiments were performed according to institutional guidelines and approved by the local committee for animal research.

Instrumentation.

The trachea was exposed and ligated proximal to the end of the endotracheal tube to exclude leaks. A cardiac catheter (5-Fr) with a piezoresistive pressure transducer on its tip (model PC 350, Millar Instruments, Houston, TX) was inserted into the trachea through a small hole near the end of the endotracheal tube—thereby not reducing the endotracheal tube’s lumen—and advanced ∼2 cm. The estimated position of the tip was ∼1 cm proximal of the carina. The insertion hole was sealed around the catheter with cyanoacrylate tissue glue. A Statham pressure transducer and a water manometer (accuracy 0.1 cmH₂O, compliance 0.09 ml/cmH₂O) were connected between the y piece of the ventilator tubing system and the endotracheal tube (see Fig. 1). The airway diameter at the connection site was 5 mm. The intratracheal and extratracheal pressure transducers were calibrated by opening the endotracheal tube to admit room air (zero pressure) and applying constant pressure steps with the oscillation switched off (continuous positive airway pressure), by using the water manometer as a calibration standard. The animals were apneic throughout. The right common carotid artery was cannulated for monitoring blood pressure and obtaining samples for blood-gas analyses. Oxygenation was monitored with a pulse oxymeter (Nellcor) attached to one lower leg, and inspired oxygen fraction was adjusted to maintain oxygen saturation >95%.

Measurements.

Measurements were performed with HFOV only, without additional conventional breaths, by using a SensorMedics 3100A (SensorMedics, Yorba Linda, CA) set to either 50% (SM₅₀) or to 30% inspiratory time (SM₃₀) (Ti/Te ratio 1:1 or 1:2.3, respectively); a Babylog 8000 (BL; HFOV software version 4.0, Drägerwerk, Lübeck, Germany); and an HFV-Infantstar (IS; software version 48, Nellcor Puritan Bennett, Carlsbad, CA). The Ti/Te ratio of the BL was ∼1:2 and was not changeable by the user. The IS had a fixed inspiratory time of 18 ms in the high-frequency mode; the expiratory time and the Ti/Te ratio were thus dependent on the oscillation frequency ( f ).

Before the start of the measurements, the MAP that produced the best possible oxygenation with the lowest possible inspired oxygen fraction without compromising arterial blood pressure in each animal was determined by using SM₃₀. This MAP was subsequently used throughout the experiment, and its values ranged from 6 to 12 cmH₂O in all animals. To use equally effective oscillation amplitudes in all ventilators, the oscillation amplitude was adjusted for each ventilator and animal to result in arterial Pco₂ (Pa_CO2) values between 35 and 45 Torr at f = 10 Hz. An equilibration period of at least 30 min was allowed, until two consecutive Pa_CO2 readings, at least 10 min apart, were within the target range.

To study intrapulmonary pressure in relation to MAP at the y piece, the endotracheal tube was cross-clamped for 3 s proximal to the pressure sensors (Fig. 1), and the pressure change distal of the clamp, a measure of the pressure difference between the y piece and the lung during oscillation, was read from the water manometer and simultaneously registered by the chart recorder. Leaks were excluded by observing a pressure plateau on the chart recorder after the clamp was closed. The animals were apneic throughout the experiments. Single clamping resulted in either a pressure rise or a pressure drop at the y piece, depending on the instant in the oscillatory cycle at which the clamping occurred (5). To obtain a mean pressure change, clamping was repeated 20 times per animal with each of SM₅₀, SM₃₀, BL, and IS in random order, and mean pressure changes for each animal and ventilator setting were calculated. From these individual means, mean values and SDs across all four animals were obtained and analyzed with Student’s t-test for unpaired observations.

Thereafter, the mean pressure difference between y piece and trachea was determined by using the proximal and intratracheal sensors. The oscillators were interchanged in random order, and the preset MAP and oscillation amplitudes were not changed from the first experiment. With each oscillator, measurements were carried out at the followingf values: 1, 3, 5, 7, 10, 12, 15, 17, and 20 Hz, as far as supported by the respective oscillator, always starting with the lowest frequency and increasing stepwise. To test whether the pressure in the trachea differed significantly from that at the y piece when the animals were ventilated with SM₅₀, SM₃₀, BL, and IS, the areas (integrals) between the curves of intratracheal pressure vs. f and the pressure at the y piece vs. f in the frequency range between 5 and 15 Hz were calculated for each animal by using the trapezoid rule. Median values were obtained for SM₅₀, SM₃₀, BL, and IS and were tested with the Wilcoxon signed-rank test for deviation from zero (P < 0.05).

Finally, a pneumotachograph, linear up to 15 l/min, with a resistance of 1.1 kPa ⋅ s ⋅ l⁻¹ at 5 l/min (23) was inserted between the y piece and the endotracheal tube (see Fig. 1) for simultaneous registration of flow and pressure waveforms of each oscillator. The Ti/Te ratio was determined by measuring the duration of inspiratory and expiratory flows and calculating the quotient. At the end, the animal was euthanized with a pentobarbital sodium overdose. All measurements were recorded with a multigraph (Hellige, Freiburg, Germany).

Calculation of the Bernoulli effect.

The measurement of the mean pressure at the y piece by a lateral-pressure tap may be distorted by the Bernoulli effect, because only the static component (Pst) of the total pressure (Pt) is being measured. According to the Bernoulli formula, Pst underestimates Pt by the velocity-dependent dynamic pressure 1/2 ρv² (3)

RESULTS

Pa_CO2 values between 35 and 45 Torr were easily achieved only with SM₅₀ and SM₃₀. Despite maximum amplitude settings, Pa_CO2 was >45 Torr in two of four animals with IS (range 35–51 Torr) and in all animals with BL (range 48–68 Torr). Clamping of the endotracheal tube resulted in a small but significant rise in mean pressure distal of the clamp with SM₅₀, but with SM₃₀, BL, and IS we observed a significant drop (P < 0.05; Fig. 2 and Table1).

Intratracheal pressure measurements at various f values, as shown in Fig. 3 and Table2, yielded 0.4–0.7 cmH₂O higher mean pressures than at the y piece with SM₅₀(P < 0.05). In contrast, with SM₃₀, BL, and IS, the intratracheal mean pressure was always lower than at the y piece by 0.4–0.9 cmH₂O (SM₃₀,P < 0.05), 0.3–3 cmH₂O (BL,P < 0.05), and 1–4.7 cmH₂O (IS,P < 0.05). Furthermore, the pressure difference between the trachea and the y piece was significantly greater with IS than with SM₃₀ and BL (P < 0.05, Table 2).

The effective Ti/Te ratio at the endotracheal tube connector, as determined from the durations of inspiratory and expiratory flow, was almost constant throughout the frequency range and close to the expected values with SM₅₀(expected 1:1.0, measured 1:1.0–1:1.1) and SM₃₀(expected 1:2.3, measured 1:2.1–1:2.4). The Ti/Te ratio of IS was always greater than expected from the 18-ms inspiratory pulse duration and increased with frequency (1:12.5–1:1.3). The Ti/Te ratio of BL also increased with frequency (1:2.2–1:1.3, Table 3).

Average flow rates during oscillation were between 23.5 and 84.5 ml/s. The Bernoulli effect (dynamic pressure) was <0.11 cmH₂O with all ventilators and settings (Table 3), and thus it did not significantly influence our results.

Flow and pressure waveforms at the y piece and intratracheal pressure waveforms of SM₅₀, SM₃₀, BL, and IS at 10 Hz are shown in Fig. 4. The shapes of the inspiratory and expiratory waveforms of SM₅₀ closely resembled each other, and so the oscillation was symmetrical. In SM₃₀, BL, and IS, the inspiratory peak was shorter and higher than the expiratory trough. The shape of each inspiratory pulse was distinctly different from the shape of each expiratory pulse; the oscillation was thus asymmetric. All waveforms were strongly dampened in passing through the endotracheal tube, and most device-specific patterns, such as the incisurae of the SM waveform or the short high peak of the IS waveform, were lost, leading to very similar intratracheal pressure waveforms with all devices.

DISCUSSION

Two different experiments were performed in this animal study to measure mean intrapulmonary pressure and intratracheal pressure during HFOV while comparing three high-frequency oscillators, one of them with two different Ti/Te ratios. First, the average intratracheal pressure change after repeated occlusion of the endotracheal tube enabled us to make an accurate estimate of the average intrapulmonary pressure during oscillation. This experiment did not require intratracheal sensors, and the pressure changes were measurable with a water manometer, which did not have to be calibrated. Depending on the phase in the oscillatory cycle in which the clamp was closed, pressure after clamping varied. To obtain representative mean values, clamping was repeated 20 times per animal and ventilator setting.

Second, intratracheal pressures over the complete frequency range provided by the devices were investigated with a catheter pressure transducer placed intratracheally. The direct measurement inside the trachea combined with the exceptional frequency response of the sensor (11) did not require correction of the signal as previously shown (4). The velocity-dependent Bernoulli effect may decrease the reading of the extratracheal pressure transducer during oscillation but not during clamping. Thus a pressure rise during clamping may be overestimated, and a pressure drop underestimated. In our setup, however, the Bernoulli effect should not have significantly altered the pressure measurements. Calculations of the Pdyn revealed values <0.11 cmH₂O, which were much smaller than in experiments performed on adult dogs (24), probably because our piglets required much smaller tidal volumes and flow rates.

Different techniques were used for the generation of the oscillation in the high-frequency oscillators investigated. In the SM, a membrane electrically driven by a linear motor generated the inspiratory and expiratory phase of the oscillation. Both phases were thus active. Square waves were used at all times. The f range was 5–15 Hz, the Ti/Te ratio was variable between 1:2.3 (SM₃₀) and 1:1 (SM₅₀). At the y piece, the effective Ti/Te ratio was always close to the theoretical value and independent of f(Table 3). This device was by far the most powerful one in our study and achieved the desired Pa_CO2 values easily. Designed as a pure oscillator, it lacks the option of conventional ventilation.

The BL with HFOV software version 4.0 used an electrically driven membrane in the expiration valve to generate oscillations and a Venturi system similar to the IS for an active expiratory phase, although the effect of the latter may be less obvious in the waveform recordings (see Fig. 4C). The f range was 5–20 Hz, the Ti/Te ratio, which could not be modified by the operator, was dependent on f and between 1:2.2 (at 5 Hz) and 1:1.3 (at 20 Hz, Table 3). The maximum-pressure amplitude was insufficient to produce the desired Pa_CO2values in our animals, but the ventilator performed well in our preterm infants who weighed less (data not shown).

In the IS, the opening of a high-pressure valve for short periods generated the inspiratory pulses. After each pulse, a Venturi system sucked air out of the ventilator circuit, resulting in an active expiratory phase. The f range was 1–22 Hz. As the pulse-on time was fixed to 18 ms, the Ti/Te ratio was dependent onf and could not be adjusted independently. At 10 Hz, for instance, the theoretical Ti/Teratio was 1:4.6, at 20 Hz it was 1:1.8. Because of dampening, the inspiratory pulse became broader on its way through the heater and the tubing system, resulting in Ti/Teratios between 1:12.5 (at 1 Hz), 1:2.3 (at 10 Hz), and 1:1.3 (at 20 Hz) at the y piece. Thus the effective Ti/Te ratio was only slightly smaller than that of SM₃₀ and BL at most frequencies (see Table 3). The new software version 83 with enhanced oscillation amplitude, which is now being built into the IS device, still uses a fixed pulse-on time of 18 ms; therefore, the results we obtained with software version 48 most likely apply to the software version 83 as well. The newer software may provide more efficient CO₂removal.

Despite these technical differences, intratracheal pressure waveforms were similar. The square waves of SM and the sharp and short pressure peaks of IS, which contained more noisy overtones in their frequency spectrum, were subject to stronger low-pass filtering (dampening) in passing through the endotracheal tube than the smooth oscillations of BL (Fig. 4). The resulting intratracheal pressure waveforms, stripped of the overtones, lacked most device-specific patterns and looked very similar (see below). At 10 Hz, inspiration and expiration were incomplete in all devices, as shown by the constantly remaining pressure gradient between the y piece and the trachea at the end of both phases and by the steep angles at the intersections of the flow curve and the zero line.

With SM₃₀, BL, and IS, the inspiratory time was shorter than the expiratory time, requiring a higher inspiratory peak than expiratory trough to compensate. This can be easily recognized if one draws a horizontal line, representing the mean pressure, through the extratracheal pressure curves in Fig. 4, B–D, so that the area between the line and the pressure curve is equal above and below it. The resulting inspiratory and expiratory waveforms were distinctly different; the oscillation was thus asymmetrical. Under these conditions, a significant pressure drop in the airways was observed after clamping the endotracheal tube, indicating that the lungs contained less air during oscillation than they would have had under continuous positive airway pressure of the same mean value at the y piece. This view is supported further by the direct measurements of the mean intratracheal pressure, which yielded significantly lower values than at the y piece. The difference was significantly greater with IS than with SM₃₀ or BL, probably because the IS had a smaller Ti/Te ratio than SM₃₀and BL at most frequencies (Table 3). The peculiar course of the curve of IS in Fig. 3 with a local minimum at 5 Hz and the minima of BL and IS at 20 Hz are difficult to explain. We speculate that the local minimum at 5 Hz may be caused by the extreme Ti/Te ratio (1:3.7) of the IS at 5 Hz. Furthermore, the efficiency of the Venturi system generating the active expiration of BL and IS may be dependent on frequency.

BL had to be used with a slightly less effective oscillation-amplitude setting than the other devices, because higher amplitudes were not available. We speculate that the pressure difference between the Y piece and the trachea might have been higher if the ventilator had had higher oscillation amplitudes available.

Intratracheal and alveolar pressures lower than at the y piece were also found in adult rabbits (15) and juvenile pigs (weight 10–16 kg) (29). Our findings of a similar pressure gradient with two technically totally different oscillators (SM₃₀ and BL), which had only the Ti/Te ratio (1:2) in common, an even higher pressure gradient with the third device (IS), which had a more extreme Ti/Te ratio, plus the reports in the literature cited above lead to the hypothesis that it is the asymmetry of the oscillation with a shorter inspiration than expiration that is responsible for the lower intratracheal pressure, and not some other ventilator-specific effects.

The lower intratracheal pressure may be explained by the interaction of the strongly low-pass-filtering properties of the endotracheal tube (13) with the asymmetry of the oscillation. In analogy to electronics, the endotracheal tube and the lung can be viewed as a resistor and a capacitor, respectively, connected in series. This circuit is known as a low-pass filter. The endotracheal tube (resistor) limits flow (current), and thus delays filling and pressure (voltage) rise inside the lung (capacitor), resulting in a delayed reaction of the intratracheal pressure to extratracheal pressure changes. The faster the pressure change, the greater the discrepancy between the extratracheal and intratracheal pressure waveforms.

Any cyclic curve can be viewed as a summation of sine waves with various frequencies. The proportion of higher frequencies, also called overtones, is greater in waveforms containing square waves or short, high peaks. The steeper the rising and falling edges of a square wave and the higher and shorter a peak, the greater the high-frequency component contained in it. When such a signal is subjected to low-pass filtering, the high-frequency components are removed first. Incisurae appear in square waves (as in Fig. 4, A and B), because the high-frequency components filling the incisurae have been removed. High, short peaks disappear, steep edges become slanted, but longer peaks and troughs are not equally affected.

In asymmetric oscillation, the shorter inspiratory pressure and flow pulses have a greater proportion of higher frequencies (overtones) than the longer expiratory pulses, and, therefore, are more strongly affected by low-pass filtering. The stronger filtering of the inspiratory pulses, combined with the better transmission of the expiratory pulses, results in a lower intratracheal pressure in comparison to the y piece. Therefore, asymmetrical oscillation with a shorter inspiration than expiration may result in a lower lung volume than a constant positive airway pressure of equal mean value at the y piece.

With SM₅₀, the inspiratory and expiratory waveforms were similar, the oscillation was thus symmetrical. Under this condition, air trapping occurred to a small degree, as shown by the small but significant pressure rise in the airway after clamping the endotracheal tube and the intratracheal pressure measurements, which yielded slightly but significantly higher mean pressures in the trachea than at the y piece. Similarly, an increase of lung volume by oscillation with symmetrical sinusoidal waves generated by different devices was demonstrated in human adults (22), isolated dog lungs (8), and neonates with respiratory disease (18), in the latter indirectly, with a jacket plethysmograph. Occlusion pressures in normal and lung-lavaged rabbits were higher than the respective mean pressures at the y piece, when low MAP values were used (5). In healthy adult rabbits ventilated with SM₅₀ (15), as well as in adult mongrel dogs (24) and excised dog lungs (1) ventilated with symmetrical sinusoidal waveforms, intratracheal and alveolar pressures exceeded the pressure at the y piece.

As lung hyperinflation has been noted with different HFOV generators and different waveforms (sinusoidal and square), and considering the almost-perfect symmetry of the SM₅₀ waveform with equal inspiratory and expiratory amplitude and duration recorded in our experiments, the air trapping is unlikely to be caused by residual asymmetries in the SM₅₀ waveform. Other mechanisms, such as differences between inspiratory and expiratory airway impedance, must be involved (8). It is known from spontaneous breathing as well as from conventional ventilation that bronchial elasticity gives rise to a higher resistance during expiration than during inspiration (21, 28). The extent of this effect in HFOV is not known. The abrupt airway diameter change at the end of the endotracheal tube may be another source of inspiratory-to-expiratory impedance differences in higher frequencies. According to an analysis of tube aerodynamics conducted by Bush et al. (7), flow from the smaller to the larger diameter (inspiration) faces a smaller impedance than vice versa (expiration), resulting in less inhibition of inspiratory flow than expiratory flow and, consequently, leading to air trapping. As the total diameter of the bronchial tree increases with increasing distance from the glottis, the same effect may also play a part in more distal airways. These effects can be compensated for or overcome, as discussed above, by asymmetrical waveforms with shorter inspiration than expiration times.

Air trapping may not affect all parts of the lung equally. Marked differences in alveolar pressures were noted between different lung lobes of adult rabbits (15), isolated fresh and dried pig lungs (27), and a lung model consisting of airbags sealed to isolated bronchial trees of dried pig lungs (27) and attributed to endobronchial aerodynamic effects. A photographic study of the pleural surface of excised dog lungs revealed increasingly asynchronous and nonuniform expansion under HFOV with increasing frequency; phase lags up to 180° between different lobes were noted at frequencies up to 30 Hz (19). Regional pressure differences between different parts of the piglets’ lungs were beyond the scope of our study and, probably, less important because we used lower f values.

We have presented the first direct comparison of the intratracheal pressures produced by different high-frequency oscillators and different Ti/Te ratios in the same animal model. From our data and evidence reported in literature, we conclude the following. In asymmetrical oscillation with shorter inspiration, intratracheal pressure is lower than at the y piece. Symmetrical oscillation promotes air trapping but obviously less than high-frequency jet ventilation (2). The different techniques employed in the ventilators to generate the positive- and negative-pressure swings of asymmetrical oscillations do not seem to play an important part with respect to air trapping, because all differences between the devices can be explained by different Ti/Te ratios. Although air trapping was not observed to a clinically important extent in our study (<1 cmH₂O pressure difference), some concern remains. For maximum safety and to avoid overinflation of the lung and compensate for airway-impedance differences between inspiration and expiration (7,21, 28), we suggest that asymmetrical oscillation with a shorter inspiration than expiration should be preferred when HFOV is used.