INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

A mean airway pressure (MAP) sufficiently high to achieve adequate recruitment of the infant's lung, called the "high lung volume strategy," has been shown to be crucial for the achievement of optimal results in ventilating animals with high-frequency oscillatory ventilation (HFO): The MAP is an important determinant of oxygenation in the HFO of dogs (1), rabbits (2), and preterm infants (3) with respiratory failure. Pulmonary function and compliance were improved, lung damage reduced and the benefit of exogenous surfactant prolonged if the lung volume was optimized in surfactant-deficient rabbits (4, 5) and premature baboons (6). Likewise, the use of HFO led to improved results and fewer complications than conventional mechanical ventilation (CMV) only in those clinical trials in which the ventilatory strategy placed a priority on optimizing lung volume (7). Such benefits were not seen in trials that used a strategy favoring lower mean pressures (11, 12). Lung hyperinflation, however, can lead to hemodynamic compromise secondary to increased pulmonary vascular resistance (13) and pulmonary interstitial emphysema. Determination of the best MAP for adequate recruitment of each patient's lung, however, is generally based on indirect signs such as oxygenation, chest radiograph (radiolucency and diaphragm position), and the absence of hemodynamic compromise. Direct measurement of mean lung volume (MLV) has not as yet been available during HFO, because gas transport during HFO occurs more by dispersion and diffusion (14) than by bulk flow as during CMV, and such transport cannot be readily measured. Following an idea of Karna and coworkers (15) and Wood and coworkers (16), we have developed a method to measure MLV during HFO using the sulfur hexafluoride (SF₆) washout method. In this study, we demonstrate the feasibility of MLV measurements during HFO and, for the first time, present data concerning the MLV of preterm infants ventilated with HFO according to the high lung volume strategy. We compare the results to known percentiles of functional residual capacity (FRC) in healthy preterm infants (17) in order to establish whether our strategy in fact results in the desired high lung volume. We have further investigated the effect of MAP changes on MLV and the time needed for stabilization of MLV on a new level after modification of MAP.

We feel that the term "functional residual capacity (FRC)," which denotes the end-expiratory lung volume, is not appropriate during HFO and should only be used in CMV or spontaneous breathing, when there is discrete expiration with a definite end. Therefore, we use "mean lung volume (MLV)" in this report for measurements during HFO.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Patients

The study was approved by the University of Ulm institutional review board as part of an ongoing multicenter HFO trial. MLV was measured in 13 premature infants ventilated by HFO according to the high lung volume strategy because of respiratory distress (for patient data, see Table 1).

All infants had received surfactant during their initial treatment, according to our surfactant criteria (see below). Each infant was monitored with a heart rate monitor, a transcutaneous Pa_CO2 sensor, and a pulse oximeter. The infants were not paralyzed, but in quiet sleep during the measurements. Informed consent was obtained from the parents of each infant. SF₆ was supplied by Linde (Munich, Germany), as a 1:1 mixture with nitrogen. This mixture has been approved by the Bavarian Government Health Authority for diagnostic use in human subjects.

Ventilation

Infants requiring intubation in the delivery room received CMV until admission to the neonatal intensive care unit (less than 0.5 h). Afterwards, all infants in this study were ventilated using the HFO-type ventilator HFV-Infant-Star with HFV software version 83 (Nellcor Puritan Bennett, Carlsbad, CA). Oscillation frequency was 10 Hz throughout; additional conventional breaths were not used. Oscillation amplitude was adjusted to obtain Pa_CO2 values between 35 and 50 mm Hg. MAP was adjusted according to our guidelines for the high lung volume approach, which were as follows: MAP, as measured with the built-in pressure transducer, was initially set 2 cm H₂O higher than during the previous CMV, usually 10 to 12 cm H₂O, and increased in 1-cm-H₂O steps until no further improvement of oxygenation was observed or the diaphragm position on an anterior-posterior chest radiograph reached the posterior part of the ninth rib. MAP was reduced to avoid hyperinflation if the diaphragm position was below the ninth rib. After tracheal suctioning, MAP was increased by 5 cm H₂O for 15 s. Surfactant was administered when oxygenation was insufficient (fraction of inspired oxygen [FI_O2]  0.30 for an arterial saturation of  90%) and either the diaphragm position was at the posterior part of the ninth rib on chest radiograph, or the MAP exceeded the following pressure limits: 12 cm H₂O for gestational age (GA) < 26 wk, 13 cm H₂O for GA 26 to 27 + 6/7 wk, 14 cm H₂O for GA  28 wk.

Mean Lung Volume Measurement

MLV was measured by washout of SF₆, an insoluble, nontoxic tracer gas (18, 19). The setup of the measurement apparatus, the measurement process, and the algorithms for calculation of lung volume have been previously described in detail (20). In brief, a pneumotachograph and a fast mainstream infrared SF₆ analyzer (Siemens-Elema, Solna, Sweden) were inserted between the Y-piece of the ventilator circuit and the endotracheal tube. The pneumotachograph (21) had a deadspace of 0.9 ml and a resistance of 1.1 kPa · s/L at 5 L/min, and the cuvette of the SF₆ analyzer had a deadspace of 1 ml and a resistance of 0.26 kPa · s/L at 5 L/min. The oscillation amplitude was increased by about 30% to prevent an increase in Pa_CO2 resulting from the increased airway impedance and deadspace. SF₆ was added to the ventilator bias flow to a concentration of 0.8 to 1.1%, until the airway SF₆ concentration was constant. The MLV was determined during washout of the tracer by continuous registration of respiratory flow and SF₆ concentration and calculation of the total volume of expired SF₆. Division of the latter by the initial SF₆ concentration yielded the lung volume at the transition from wash-in to washout. Any changes of ventilation pattern or lung volume during the washout would not change the total amount of tracer gas to be washed out and thus would not influence the measurements. Therefore, using CMV for the washout instead of HFO did not distort the results (15) and was used to ensure unidirectional gas flow during the washout, when exact tidal volume recordings were mandatory, to avoid the complex rebreathing occurring during HFO and to reduce diffusion and dispersion (14). After equilibration with the tracer gas during HFO, the transition from HFO to CMV for washout required two switches to be set on the HFV-Infant-Star. First, HFO was switched off (continuous positive airway pressure [CPAP]). This was detected by the computer controlling the MLV measurement apparatus. SF₆ admixture to the bias flow was stopped, its concentration in the airway stored for later MLV calculation, and registration of the expired SF₆ was started. Second, the HFV-Infant-Star was switched from CPAP to intermittent mandatory ventilation (IMV), and tracer gas washout began. The previous MAP became the positive end-expiratory pressure (PEEP) of the new CMV. If the PEEP was above 6 cm H₂O, it was lowered during the following breaths to allow for a sufficiently large tidal volume for a quick washout. The other CMV settings were: peak inspiratory pressure, 20 to 22 cm H₂O; inspiratory time, 0.25 s; rate, 60 breaths /min. Correct synchrony between the change of the ventilation mode and the start of the expired tracer gas registration and volume calculation was ensured by the software of the measurement device. After completion of the washout (duration approximately 1 min), the ventilation was switched back to HFO.

The accuracy and reproducibility of MLV measurements carried out with our apparatus have been tested in CMV and HFO. Measurements of a conventionally ventilated dummy lung with a known, adjustable FRC resulted in a difference of 0.7 ± 3.2% (mean ± SD). The coefficient of variation across 20 FRC determinations in five adult rabbits was 1.7% during CPAP and 1.98% during CMV (20). Measurements of a dummy lung with a known volume (50 ml), using HFO for wash-in and CMV for washout with the same method as described in this work resulted in a difference of 0.4 ± 2.3% (U. Thome, unpublished).

Sequence of Lung Volume Measurements

Each change of MAP was followed by serial MLV measurements until no further trend was detectable in three consecutive measurements. This was assumed if MLV readings varied by less than 5% between measurements and in both directions, i.e., upward and downward. The steady-state MLV was then obtained as the mean value of three consecutive measurements. First, MLV was measured at least three times while ventilating the infants with the MAP chosen by the clinician. Next, MAP was changed in steps of 1 cm H₂O as follows: If the initial MLV was above 25 ml/kg, MAP was first increased until MLV changed by less than 20% of the preceding measurement or MLV was above 42.8 ml/kg, which was found to be the 95th percentile in a meta-analysis of FRC values in healthy preterm and term infants (17). Then, MAP was returned to the initial value and, after stabilization of MLV, decreased stepwise further until MLV changed by less than 20% per 1 cm H₂O step or was below 18.3 ml/kg, the fifth percentile (17). If the initial MLV was below 25 ml/kg, we started with the reduction of MAP, followed by the increase. Thus, we always worked our way to the closer endpoint first, in order to keep the experiment as short as possible. Finally, MAP was set to result in an MLV of 30 to 35 ml/kg, dependent on the results obtained during the previous measurements. When MLV had stabilized, the measurement apparatus was removed. Any manipulation detrimental to the infant, e.g., resulting in a proportionately large rise of Pa_CO2 or FI_O2 requirement, was abandoned (see example in Figure 1). The process was time-consuming because of the necessary stabilization time between modifications of MAP, and it was not possible for it to be continued to the end with each infant because of interference with other procedures of patient care. Endotracheal tube leaks were excluded by continuous comparison between inspiratory and expiratory tidal volume during the tracer washout with CMV (one infant had to be excluded because of leaks; in the remaining infants, no leak was detectable). No suctioning, hand-ventilation, or sustained inflations were done during the study. Spontaneous respiratory activity was suppressed by providing sufficient oscillation amplitude for Pa_CO2 to remain in the normal range.

View larger version (11K): [in this window] [in a new window]   Figure 1.   Patient L.R.: After reduction of MAP from 8 to 7 cm H2O in the 53rd min, MLV declined slowly for 17 min, followed by a rapid drop, indicating delayed alveolar collapse.

Data Analysis

For standardization, all MLV values were related to body weight. Linear regressions of steady-state MLV, as determined by three consecutive measurements, versus MAP (MLV = slope factor · MAP + offset) were calculated separately for each patient from measurements made before and after modifications of MAP. From the resulting values for slope factor and offset, median values across all patients were obtained. The Wilcoxon rank sum test was used to determine whether the slope factors of all patients were significantly greater than zero (p < 0.05).

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

MLV was measured 9 to 32 times (median, 22 times) in the 13 infants listed in Table 1, and the effects of 2 to 7 (median, 4) modifications of MAP were observed. The variation coefficients of repeated MLV measurements under steady-state conditions ranged from 0.02 to 21%, with a median value of 4.9%. Initial MLV, corresponding to the clinically chosen MAP, covered a wide range between 23.3 and 41.9 ml/kg (median, 33.5 ml/kg; Table 1). There was no correlation between the initial MAP and the initial MLV, indicating a wide range of pulmonary pathology. In all patients, the initial MLV, corresponding to the clinically selected MAP, was above the fifth percentile (18.3 ml/kg) of FRC values in healthy preterm infants (17), but in four patients below the 50th percentile (27.5 ml/kg), despite using a high lung volume strategy. No initial MLVs above the 95th percentile (42.8 ml/kg) were noted. Chest radiographs were taken within 1 to 17 h of MLV measurement (median, 5.3 h), and yielded diaphragm levels between 8.8 and 11 ribs (median, 10 ribs; Table 1). Initial MLV, initial FI_O2, and the diaphragm level were not correlated with each other (r ² values < 0.05).

Under systematic variation of MAP, MLV in steady state ranged between 14.7 and 45.4 ml/kg (median, 29.1 ml/kg). When MAP was reduced, MLV fell below the 50th percentile in 11 infants, and below the fifth in four of them. When MAP was increased, MLV exceeded the 50th percentile in 11 infants, and in one of these, it exceeded the 95th percentile.

MLV was strongly dependent on MAP in each patient (examples in Figures 1-3). Stepwise reduction of MAP resulted in a decline of MLV in Patient B.T., followed by a rapid restoration of the previous MLV after setting MAP to the initial value (Figure 2). In Patient L.R., MLV declined gradually for 17 min after reduction of MAP from 8 to 7 cm H₂O but dropped by 53% (total) thereafter (Figure 1). Oxygenation deteriorated as well, so we did not wait for the endpoint of the decline. In Patient K.P., stepwise increases of MAP up to 12 cm H₂O failed to have a clear-cut effect on MLV, but after increasing MAP to 13 cm H₂O, a large increase of MLV was observed. Further increments of MAP resulted in smaller increases of MLV (Figure 3).

View larger version (12K): [in this window] [in a new window]   Figure 2.   Patient B.T.: Stepwise reduction of MAP resulted in a concomitant decline of MLV. Return of MAP to the initial value led to rapid restoration of the initial MLV.

View larger version (11K): [in this window] [in a new window]   Figure 3.   Patient K.P.: Increasing MAP up to 12 cm H2O had little influence on MLV, but the increase from 12 to 13 cm H2O resulted in a marked increase of MLV. Further increases of MAP had less effect.

Overall correlation between MLV and MAP was poor. The studied patients showed a wide variety of pulmonary pathology and therefore required different MAP settings to maintain lung inflation. Linear regressions of the interrelation of MLV and MAP were therefore calculated individually for each infant and resulted in slope factors between 1.0 and 6.9 ml/cm H₂O/kg (median, 2.83 ml/cm H₂O/kg), and significantly greater than zero (p < 0.002), with correlation coefficients between 0.77 and 0.99 (median, 0.94). Data points and regression lines are shown in Figure 4, and the slope and offset values of each patient are in Table 2.

View larger version (19K): [in this window] [in a new window]   Figure 4.   Steady-state MLV in relation to MAP of all 13 patients. Individual linear regressions were calculated for each infant and included in the figure (thin lines). Median values were obtained from the slope and offset values of all regression lines. The dashed bold line corresponds to the median values for slope and offset (MLV = 2.83 · MAP + 3.70).

The time lag between modification of MAP and final stabilization of MLV on a new level varied between 2 min (minimum time owing to the duration of the measurement process) and 25 min (median, 9 min; Figure 5).

View larger version (43K): [in this window] [in a new window]   Figure 5.   The time lag between modification of MAP and stabilization of MLV.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

No direct measurements of MLV in HFO-ventilated preterm infants have been published so far. Two obstacles had to be overcome: First, gas transport during HFO occurs mainly by convective dispersion and augmented diffusion (14), which cannot be measured by a pneumotachograph, in contrast to the bulk flow in CMV. Second, the SF₆ sensor was not fast enough to accurately measure the SF₆ waveform at a ventilation frequency of 10 Hz. For both reasons, calculation of tracer gas volumes during the washout was impossible during HFO, but switching to CMV for the washout allowed us to measure the amount of tracer that had been washed in during HFO. This amount of tracer would remain the same, even if the lung volume changed during washout as a result of the alternated ventilation mode, and thus accurately reflects the MLV during HFO.

However, our method may be, like any other tracer gas method, susceptible to endotracheal tube leaks. If tracer gas escapes, MLV would be underestimated. The presence of leaks was always excluded by continuous comparison of the inspiratory and expiratory tidal volumes during CMV.

Using SF₆ as tracer gas enabled us, in contrast to other methods, to perform rapid serial MLV measurements regardless of the inspired O₂ concentration, because an SF₆ concentration of about 1% was sufficient. In several publications, FRC measurements obtained by SF₆ washout did not show a systematic difference compared with those obtained by body plethysmography (22), nitrogen washout (23), and helium dilution in healthy adult humans (22, 26) as well as in adult rabbits (27).

Two factors may have decreased the accuracy of our results: First, when the HFV-Infant-Star was switched to CPAP, MAP may have changed slightly, and so lung volume may have been affected before starting the washout of the tracer. This would have affected all measurements of all infants in this study equally and thus would not have changed the relationships of the values obtained. The change in MAP was always less than 1 cm H₂O. According to the observed tidal volumes of about 6 to 10 ml/kg during the CMV used for washout with an inspiratory-expiratory pressure difference of 15 cm H₂O, this small MAP change at the transition from HFO to CPAP could lead to a maximum error in the MLV measurement of 1 ml/kg, which is in accordance with the accuracy of the dummy lung measurements (see METHODS). Second, the interposed phases of CMV for tracer gas washout could have led to a prolonged change of MLV. In this case, we would expect to see significant changes in the MLV during the initial measurements, before the first modification of MAP, and this we did not observe. The tracer washout with CMV was probably short enough (approximately 1 min) to be completed before the collapse of a significant number of alveoli occurred, and thus had no sustained effect on MLV.

Based mainly on findings in animal experiments (4), it has been suggested that the lung volume should be optimized during HFO (high lung volume strategy) by appropriate selection of MAP and sustained inflations (28, 29), in order to achieve the best possible result. Overinflation, however, may decrease the cardiac output and increase the likelihood of air leaks or intracranial bleeding, and should therefore be avoided. Because we followed this strategy in our patients, as shown by the diaphragm levels on chest X-rays taken within 1 to 17 h of the MLV measurements, we expected the MLV to be higher than the end-expiratory lung volumes in CMV, which are known as FRC. Unexpectedly, MLV was below the 50th percentile of FRC values (17) with the clinically selected MAP in four of our infants. In none of them was it above the 95th percentile. These findings show that using our strategy, we were able to avoid overinflation, but we did not obtain a high lung volume in all cases.

As shown for the relationship between PEEP and FRC (30), the relationship of MAP versus MLV was, at least in some patients, not linear. This can be easily recognized in Patients L.R. (Figure 1) and K.P. (Figure 3). In Patient L.R., a reduction of MAP from 8 to 7 cm H₂O led to a drop in MLV to about 50%, a larger change than after all other MAP steps, indicating that an increasing number of alveoli collapsed after their closing pressure was no longer reached. A similar observation, but in the opposite direction, was observed in Patient K.P. after increasing MAP from 12 to 13 cm H₂O, suggesting that the opening pressure of a larger number of alveoli had been exceeded. The course of other patients, (e.g., Patient B.T., Figure 2), suggested a more linear relationship.

Despite the obvious nonlinearities in some patients, we used a linear model as a first-order approximation. The database was too narrow and the severity of pulmonary disease too heterogeneous to generate a mathematical model of the theoretically sigmoidal pressure-volume curve. With all patients, good linear correlations were obtained, with slope factors significantly greater than zero, indicating that MLV increases with MAP, as expected, but the wide range of slope factors and offset values makes it impossible to predict MLV solely based on MAP.

The median slope factor of 2.83 ml/cm H₂O/kg fitted well with results obtained on the relationship between PEEP and FRC in CMV: The median slope factor of linear correlations between FRC and PEEP, as measured by our method, was 2.94 ml/cm H₂O/kg (31). In another study of conventionally ventilated, surfactant-treated preterm infants using the helium dilution technique (30), FRC increased significantly by 1.3 ml/ kg (when PEEP was increased from 2 to 3 cm H₂O), 2.9 ml/kg (when PEEP was increased from 3 to 4 cm H₂O), and 3.6 ml/ kg (when PEEP was increased from 4 to 5 cm H₂O). Comparisons of FRC determinations during CMV with MLV determinations during HFO must be interpreted with caution, however, because during CMV, FRC is solely dependent on PEEP and not on MAP (31), while during HFO, MLV is dependent on MAP, as shown here. The time lags between modifications of MAP and stabilization of MLV on a new level varied from 2 min (the minimum time owing to the duration of the measurement process) to 25 min. In most patients, it was completed within 10 min. No other ventilation parameter was correlated to the time lag, so it could not be predicted. After any modification of MAP, the clinician therefore should wait at least half an hour before the full effect of the MAP change can be evaluated. Sustained inflations may shorten the time lag when increasing MAP (32).

In this study, we present the first MLV measurements in HFO-ventilated preterm infants. We have shown that, in our hands, a clinically instituted high lung volume strategy never produced MLV values above the 95th percentile of the FRC values of healthy preterm infants (17), but resulted in unexpectedly low MLV (below the 50th percentile) in some infants. The influence of MAP on MLV depended on the individual properties of the lung, and therefore prediction of MLV solely from MAP was not possible. Stabilization of MLV after modification of MAP lasted no more than 10 min in most cases, but took up to 25 min in some.

Although our method has some limitations, we have shown that MLV measurements in HFO-ventilated preterm infants are possible with sufficient accuracy and reproducibility, and this may be further improved by an HFO ventilator specifically designed to minimize pressure changes at the transition from HFO to CPAP. Remote control of the ventilator by the measurement apparatus could automatize the complete process of measurement and further improve accuracy and reproducibility. Whether MLV-guided MAP settings would improve the outcome in HFO-ventilated preterm infants remains to be studied.