Tidal volume significantly affects oxygenation in healthy pigs during high-frequency oscillatory ventilation compared to conventional ventilation

The role of high-frequency oscillatory ventilation (HFOV) has long been debated. Numerous studies documented its benefits, whereas several more recent studies did not prove superiority of HFOV over protective conventional mechanical ventilation (CV). One of the accepted explanations is that CV and HFOV act differently, including gas exchange. To investigate a different level of coupling or decoupling between oxygenation and carbon dioxide elimination during CV and HFOV, we conducted a prospective crossover animal study in 11 healthy pigs. In each animal, we found a normocapnic tidal volume (VT) after the lung recruitment maneuver. Then, VT was repeatedly changed over a wide range while keeping constant the levels of PEEP during CV and mean airway pressure during HFOV. Arterial partial pressures of oxygen (PaO2) and carbon dioxide (PaCO2) were recorded. The same procedure was repeated for CV and HFOV in random order. Changes in PaCO2 intentionally induced by adjustment of VT affected oxygenation more significantly during HFOV than during CV. Increasing VT above its normocapnic value during HFOV caused a significant improvement in oxygenation, whereas improvement in oxygenation during CV hyperventilation was limited. Any decrease in VT during HFOV caused a rapid worsening of oxygenation compared to CV. A change in PaCO2 induced by the manipulation of tidal volume inevitably brings with it a change in oxygenation, while this effect on oxygenation is significantly greater in HFOV compared to CV.


Results
A total of 11 healthy animals underwent the whole protocol of the study and corresponding data were used for evaluation. One additional animal died due to a cardiac arrest during the experiment. The animals were ventilated in the CV phase and in the HFOV phase. In each phase, we varied V T over a wide range and recorded arterial partial pressures of oxygen (P a O 2 ) and carbon dioxide (P a CO 2 ). All data were aggregated into two groups corresponding to CV and HFOV. In total, the aggregated data contained 177 individual data entries (93 CV and 84 HFOV) containing V T , P a CO 2 and P a O 2 values.
The effect of tidal volume changes on carbon dioxide elimination during both conventional and high-frequency ventilation is presented in Fig. 1. The power function y = ax b + c was applied to the HFOV data (R 2 = 0.968) and to the CV data (R 2 = 0.943). The results document a statistically significant (P < 0.001) stronger effect of tidal volume changes on CO 2 elimination during HFOV compared to CV. The same relative change in tidal volume from the tidal volume that corresponds to normocapnia (P a CO 2 = 40 mmHg) and is individual for each animal, affects the average level of P a CO 2 more during HFOV than CV. For example, a decrease in tidal volume by 10% from normocapnia causes P a CO 2 to increase by 5 mmHg (13%) on average during CV, but by 10 mmHg (24%) on average during HFOV. The increase in tidal volume by 10% from normocapnia reduces P a CO 2 by 4 mmHg (10%) on average during CV, but by 7 mmHg (17%) on average during HFOV. As the ventilatory frequency was kept constant during the experiment, the minute ventilation was directly proportional to the delivered tidal volume; therefore, minute ventilation had a stronger effect on CO 2 elimination in HFOV compared to CV.
The effect of changes in tidal volume on oxygenation is presented in Fig. 2. The same relative change in tidal volume, expressed as a ratio of the current and the normocapnic tidal volume, generates different changes in oxygenation during HFOV and CV. In this figure, the change in arterial partial pressure of oxygen (ΔP a O 2 ) values represent differences in current P a O 2 from the reference normocapnic P a O 2 value determined for each animal separately. The power function y = ax b + c was applied to both the HFOV data (R 2 = 0.963) and the CV data (R 2 = 0.869). The function expresses the predicted average change in P a O 2 in response to the change in tidal volume relative to the tidal volume that was titrated to achieve normocapnia. This effect of changes in tidal volume on the change of P a O 2 is significantly stronger during HFOV (P < 0.001). Increasing tidal volume above its normocapnic value during HFOV represents a method to significantly improve oxygenation, whereas the improvement in oxygenation during CV hyperventilation is limited. Likewise, any decrease in tidal volume causes a rapid decrease in oxygenation during HFOV, whereas this effect is much smaller during CV. For example, the decrease in tidal volume by 10% from normocapnia causes an average reduction of P a O 2 by 7 mmHg during CV, but by 17 mmHg during HFOV. The increase in tidal volume by 10% from normocapnia causes an average increase of P a O 2 by 5 mmHg during CV, but 12 mmHg during HFOV. Additionally, as the ventilatory frequency was kept constant during the experiment, the same findings as those for tidal volume could be applied to minute ventilation. Graph presented in Fig. 3 summarizes the previous two graphs (Figs. 1 and 2) and shows the different degree of interconnectedness of the changes in P a CO 2 with the changes in P a O 2 during CV and HFOV. The linear functions were fitted to the data (R 2 = 0.980 for HFOV and R 2 = 0.885 for CV). The graph documents that the change in P a CO 2 induced by the manipulation of tidal volume inevitably brings with it a change in oxygenation (i.e., P a O 2 ), while this effect on oxygenation is significantly greater in HFOV compared to CV (P < 0.001). As a result, for example, correction of P a CO 2 by a change in V T causes a significantly greater change in P a O 2 in HFOV than in CV. The change in oxygenation in HFOV is more closely related to the change in P a CO 2 than in CV.

Discussion
The most important finding of this study is that changes in P a CO 2 induced by adjustment of tidal volume during HFOV affect oxygenation significantly compared to the same P a CO 2 changes during CV. Furthermore, increasing tidal volume above its normocapnic value during HFOV represents a way of a significant improvement in oxygenation, whereas improvement in oxygenation during CV hyperventilation is limited.
The limitation of improving oxygenation while increasing tidal volumes in CV is well known [2,18]. This means that increasing tidal volume is not considered as a tool to improve oxygenation during CV and the results of the current study are consistent with this principle. On the other hand, we did not find such a limitation in oxygenation improvement during HFOV. At the same time, any decrease in tidal volume causes a rapid worsening of oxygenation during HFOV, whereas this effect is much smaller during CV.
In order to decrease the tidal volume during CV, permissive hypercapnia may be introduced [15][16][17]. The reduction in oxygenation as an effect of the introduction of permissive hypercapnia during CV is unanticipated and has not been reported yet. The current study shows a stronger relationship between oxygenation and CO 2 elimination caused by changes in tidal volume during HFOV. Thus, the use of permissive hypercapnia during HFOV may further deteriorate oxygenation. In practice, for example, there are cases when it is not possible to achieve normocapnia despite the use of the largest possible tidal volume, which is limited by setting the maximum possible value of the parameter 'power' determining the magnitude of pressure oscillations and thus the maximum achieved tidal volume. If hypercapnia is allowed in this case, worsened oxygenation can be expected as a consequence of the significant effect of the suboptimal tidal volume not only on the elimination of carbon dioxide but also on oxygenation. Permissive hypercapnia may also be instituted via intentional reduction of the pressure oscillations amplitude (by setting lower 'power') in order to strengthen the protectivity of HFOV to the lung tissue [19]. Intentional reduction of tidal volume in this scenario may negatively affect oxygenation as well. The authors speculate that in some cases the deterioration in oxygenation during HFOV with permissive hypercapnia may be falsely interpreted as HFOV inefficiency. These results may help to explain the HFOV-related controversies after two large HFOV clinical studies OSCAR [20] and OSCILLATE [21] as well as the consequential analyses and review articles were published. Recent studies demonstrated that HFOV requires an individual approach when setting the ventilator [22,23] and the lack of the individual approach to HFOV settings could have affected the outcome of the earlier large studies [24,25]. Both in pediatric [26] and adult patients [27] HFOV needs to be applied on a case-by-case basis for which a better understanding of HFOV is required. The care of patients with ARDS is being further optimized, now also due to COVID-19 ARDS. Clinicians are studying new techniques and approaches. Effect of extracorporeal membrane oxygenation [28,29], different levels of PEEP [30], prone positioning [31,32], respiratory system impedance analysis [33], and different times of weaning [34] has been studied. For the use of HFOV in the treatment of ARDS, there is a need to understand the place of HFOV in the spectrum of diagnostic and therapeutic techniques for severe ARDS care and to define the optimal HFOV strategy [35]. For example, electrical impedance tomography seems to be a promising diagnostic tool for individualized settings of both HFOV [36] and CV [37].
During the study, we kept all the ventilatory parameters constant, except for the intentionally changed tidal volume (set indirectly using pressure amplitude ΔP during HFOV). If the intentional change in ΔP caused any change in other parameters, the setting of ventilator 3100B was immediately corrected. We especially observed mean airway pressure, as this parameter has a direct effect on oxygenation and therefore could affect the results of our study.
Several studies documented that mean alveolar pressure may differ from the mean airway pressure measured in the airway opening and measured by 3100B ventilator used. Pillow et al. [38] investigated, that I:E ratio is a crucial determinant of whether such a difference is present. Nevertheless, in their study, they found no evidence of significant air-trapping (i.e., the difference between mean alveolar pressure and mean airway pressure) at an I:E ratio of 1:1, which is the I:E ratio used in our study. Furthermore, they did not observe a significant development of air-trapping at I:E ratio of 1:1 across a wide range of ΔP from a minimum of 10 cmH 2 O to a maximum of 90 cmH 2 O. Several previous studies in dogs [39], rabbits [40], and adult human subjects [41,42] have shown that air-trapping elevated alveolar pressure above mean airway pressure when HFOV was employed at an I:E ratio of 1:1. A number of mechanisms have been postulated to account for this effect, and each depends on the presence of asymmetry between inspiratory and expiratory resistance [38]. In our study, we did not have an opportunity to assess mean alveolar pressure and to confirm or disprove the presence of an air-trapping that could influence oxygenation. However, we do not consider this condition to be a disadvantage of our protocol as mean airway pressure measured at the airway opening is so far the only parameter used for setting mean airway pressure during HFOV in all the published studies.
Several reasons led us to choose a group of healthy animals for the study. As we planned to evaluate the effects of tidal volume changes during the course of each protocol phase, a model stable for several hours was required so that the recorded changes within a phase would be driven by V T and would not be affected by changing lung condition and function. All ARDS models evolve in time and therefore the lung parameters change. The most common ARDS model induced by normal saline lung lavage does not represent a typical ARDS lung as ARDS in patients is rarely, if ever, solely a result of surfactant deficiency. Nevertheless, the effect of HFOV hyperventilation observed in this study in healthy animals corresponds to the results of an HFOV hyperventilation study conducted in a porcine ARDS model induced by saline lavage [43]. Hyperventilation with V T increased to 135% of its normocapnic value increased P a O 2 by 31 mmHg in the current study; whereas V T increased to 146% caused an increase in P a O 2 by 28 mmHg in the saline lavage ARDS model in pigs. The intention of our study design was to approach, as much as possible, the way in which both the CV and HFOV are used in clinical practice. The ventilation modes differ in the mechanism of gas exchange, the way of control and the significance and typical values of their parameters. Thus, in attempting a fair comparison, we did not seek a formal agreement on some parameter of the ventilator setup such as the same initial tidal volume. Instead, we adopted the opposite approach, individualizing the ventilation settings to have the same resulting effect on the internal environment, i.e., normocapnia. The aim of the study was to show that the change of ventilation, that is used to adjust P a CO 2 , affects the P a CO 2 and P a O 2 differently during the CV and HFOV under the condition of unaltered PEEP and mean airway pressure, respectively. The role of the mean airway pressure in the ventilation control in HFOV can be viewed in a manner similar to the PEEP level in CV [44]. When a change in oxygenation is unnecessary, the parameters remain usually unchanged. Our study can be understood as a sort of sensitivity analysis, where we investigate how the relative change in tidal volume to an individually set normocapnic tidal volume affects the change in blood gases in HFOV compared to CV. Our results suggest that whereas during CV the changes in oxygenation due to changes in ventilation are more or less negligible, the relatively same percentual changes in ventilation during HFOV may affect oxygenation significantly and should be taken into account. During HFOV, ventilation and oxygenation are coupled more than it has been assumed so far. This difference in coupling of oxygenation and CO 2 elimination is explicitly expressed by the different slopes of the CV and HFOV lines in Fig. 3, where P a O 2 changes more rapidly for the same change in P a CO 2 for HFOV than for CV.

Conclusions
The relative change in tidal volume occurring during titration of a target P a CO 2 during HFOV reasonably affects oxygenation during HFOV compared to CV. Increasing tidal volume improves oxygenation, whereas a decrease in tidal volume from its normocapnic value causes a rapid decrease in oxygenation during HFOV compared to CV. A change in P a CO 2 induced by the manipulation of tidal volume inevitably brings with it a change in oxygenation, while this effect on oxygenation is significantly greater in HFOV compared to CV.

Methods
The study protocol was approved by the Institutional Animal Care and Use Committee of the First Faculty of Medicine, Charles University. The study was performed in an accredited animal laboratory of the Institute of Physiology, First Faculty of Medicine, Charles University, in accordance with Act No. 246/1992 Coll., on the protection of animals against the cruelty that is harmonized with EU legislation. Twelve crossbred (Landrace × Large White) female pigs (Sus scrofa domestica) 4-5 months old with an average body weight of 44 kg (42-54 kg range), were involved in the study.

Anesthesia and preparation
Animals were premedicated by azaperone (2 mg/kg IM), followed by anesthesia with ketamine hydrochloride (20 mg/kg IM) and atropine sulphate (0.02 mg/kg IM). When placed on the operating table, initial boluses of propofol (2 mg/kg IV) and morphine (0.1 mg/kg IV) were administered. Animals were orotracheally intubated with a cuffed endotracheal tube (I.D. 7.5 mm) and connected to a conventional ventilator Hamilton G5 (Hamilton Medical AG, Bonaduz, Switzerland) in INTELLiVENT-ASV mode. An ear vein was cannulated and a continuous infusion of propofol (8 to 10 mg/kg/h IV) combined with morphine (0.1 to 0.2 mg/kg/h IV) and heparin (40 to 50 U/kg/h IV) was initiated to maintain anesthesia. The drugs doses were adjusted to keep the animals in the stage of surgical anesthesia throughout the procedure. The depth of anesthesia was checked according to reflex responses and brain monitoring of bispectral index (BIS). Initial rapid infusion of 1000 mL of normal saline was given intravenously, followed by a continuous IV infusion of 200 mL/h to 500 mL/h to reach and maintain central venous pressure of 5 to 7 mmHg. Femoral venous and arterial cannulation were performed to obtain central venous pressure (CVP, v. femoralis) and arterial blood pressure (ABP, a. femoralis). Cardiac output (CO), mixed venous blood oxygen saturation (SvO 2 ) and pulmonary artery pressure (PAP) were measured continuously by a Swan-Ganz catheter inserted via the right femoral vein and connected to a hemodynamic monitor (Vigilance I, Edwards Lifesciences, Irvine, CA, USA). Arterial blood gases were monitored continuously using a continuous blood parameter monitoring system CDI 500 (Terumo, Tokyo, Japan), via a sensor placed in the extracorporeal arteriovenous shunt (between a. femoralis and v. femoralis). The muscle relaxant pipecuronium bromide (4 mg boluses every 45 min) was administered during the experimental phase to suppress spontaneous breathing.

Study design
The study was designed as a prospective interventional crossover study. The study protocol is summarized in Fig. 4. After the preparation, instrumentation, and muscle relaxation, the animals were randomized either to conventional ventilation (CV phase) or to high-frequency oscillatory ventilation (HFOV phase). The randomization was conducted using a table of random numbers. After the completion of the designated phase, the other phase was initiated immediately.
During the CV phase, pigs were ventilated using the conventional ventilator Hamilton G5 in the volume-controlled ventilation mode (synchronized controlled mandatory ventilation, (S)CMV). At the beginning of the phase, a recruitment maneuver was performed using the P/V Tool feature of the Hamilton G5 ventilator (increasing airway pressure from 5 to 35 cmH 2 O at the speed of 3 cmH 2 O/s and followed by an inspiratory hold for 5 s) in order to minimize effects of the lung volume history and conditions among the animals. The ventilator settings were: respiratory rate 18 breaths per minute, FiO 2 21%, I:E 1:1 and a decelerating flow pattern was selected. The positive end-expiratory pressure was set 2 cmH 2 O above the lower inflection point (LIP) of the static pressure-volume curve recorded using the P/V Tool provided by the ventilator [44,45]. When PEEP predicted by the P/V Tool was lower than 5 cmH 2 O or the lower inflection point could not be identified, PEEP was set to 7 cmH 2 O. This ventilator setting, including FiO 2 , was kept constant during the whole CV phase. The initial tidal volume was set at 8.5 mL/kg of the actual body weight and was titrated to reach normocapnia (P a CO 2 = 40 ± 1 mmHg) while maintaining a respiratory rate of 18 breaths per minute. At the beginning of the HFOV phase, a recruitment maneuver was performed by inflating the lungs to a pressure of 40 cmH 2 O for 15 s [24] with no oscillations in order to minimize the effects of the lung volume history and conditions among the animals. A high-frequency oscillatory ventilator SensorMedics 3100B (SensorMedics, Yorba Linda, CA, USA) was used in the HFOV phase. The ventilator settings were: FiO 2 21%, bias flow 35 L/min, frequency of oscillations 5 Hz and I:E 1:1. The ventilator setting was kept constant during the whole HFOV phase. The HFOV mean airway pressure was initially set 5 cmH 2 O above the mean airway pressure during preceding conventional ventilation (CV phase or the preparatory phase when the animal was randomized to HFOV phase first) [5,25]. Then, the mean airway pressure was titrated to reach the best oxygenation without compromising circulation. The increase of mean airway pressure was stopped when the heart rate reached 160 bpm, mean arterial blood pressure was equal to or less than 50 mmHg, or mean pulmonary artery pressure was equal to or greater than 35 mmHg. The initial tidal volume was set to 2.5 mL/kg of the actual body weight by adjusting ΔP on the ventilator and was titrated to reach normocapnia (P a CO 2 = 40 ± 1 mmHg). There were no ETT cuff leaks used during HFOV.
The rest of both CV and HFOV phases were identical. The titrated normocapnic tidal volume V T titr found at the beginning of each phase was used for determining the range of tidal volumes to be tested and was not used later for any adjustment or calculation. The maximum tidal volume V T max was calculated as 135% of the normocapnic V T titr whereas the minimum tidal volume V T min was calculated as 80% of the normocapnic V T titr . First, V T was set to the calculated V T max and this setting was kept for at least 30 min. Arterial blood gases were recorded and their stability was confirmed after additional 5 min. Then, the tidal volume was reduced stepwise by 20 mL in CV phase or 5 mL in HFOV phase until V T min was reached. After each change of V T , the ventilatory parameters were checked immediately. When the change of V T induced a change in other parameters (e.g., mean airway pressure was slightly affected during HFOV), a correction was made to restore the original values recorded prior to the V T change. Then, all the ventilatory parameters were kept constant for 15 min and arterial blood gases were recorded. After another 5-min period, the measurement was repeated. If the change in any of the arterial blood gases was not greater than 1 mmHg, the stepwise V T reduction continued. Otherwise, another 5-min period for stabilization was introduced followed by another recording of arterial blood gases. The final values of blood gases at a given V T were calculated as an average of the last two measurements. In some pigs it was not possible to achieve the calculated V T max due to reaching maximum ΔP of the ventilator or to use the calculated V T min due to severe hypoventilation of the animal. Blood oxygen Page 11 of 14 Roubík et al. BioMedical Engineering OnLine (2022) 21:14 saturation was continuously monitored, and the reduction of tidal volume was stopped if SpO 2 fell below 85%. In such situations the range of set V T was shortened accordingly. Arterial blood gases (P a O 2 , P a CO 2 , and pH) were measured using the CDI 500 continuous blood parameter monitoring system. The accuracy of the system was verified prior to the experiment and then every 1 h by analysis of an arterial blood sample using a blood gas analyzer AVL Compact 3 (Radiometer, Copenhagen, Denmark). Heart rate, ABP, CVP, PAP, SpO 2 , ECG, body temperature, and BIS were monitored using the MU-631 RK (Nihon Kohden, Tokyo, Japan) patient monitor. CO and SvO 2 were measured by the Vigilance I monitor.
As the SensorMedics 3100B HFOV ventilator does not measure tidal volume, the delivered tidal volume was measured using a Florian (Acutronic, Hirzel, Switzerland) respiratory function monitor. The same device was used to measure delivered tidal volume during CV in order to eliminate a systematic error. The Florian monitor was calibrated using a 1-L calibrating syringe (Hans Rudolph, Shawnee, KS, USA) and the accuracy of the monitor was also verified prior to the experiment using a VT Plus FH (Fluke Biomedical, Everett, WA, USA) gas flow analyzer during both CV and HFOV ventilatory modes.

Data analysis and statistics
After the completion of each phase, characterized by the stepwise reduction of tidal volume from V T max to V T min , a value of tidal volume V T norm corresponding to normocapnia (P a CO 2 = 40 mmHg) was determined and used for further calculations, regardless the titrated normocapnic tidal volume V T titr found at the beginning of each phase. When the exact normocapnic P a CO 2 level was not directly recorded in the measured data, V T norm was estimated by linear interpolation between the two nearest tidal volumes. Then, relative tidal volumes, expressed as a percentage of V T norm , were calculated as V T /V T norm . The ventilation settings parameters (pressures and volumes) including the calculated V T norm are summarized in Table 1.
The data obtained from all the animals were aggregated into two groups corresponding to CV and HFOV phases. Three relationships among the pooled data were examined: (1) The recorded P a CO 2 levels related to the relative tidal volumes were analyzed. (2) The change in arterial partial pressure of oxygen (ΔP a O 2 ) was calculated as the difference between P a O 2 values and the normocapnic P a O 2 (i.e., the level of P a O 2 corresponding to V T norm ). These ΔP a O 2 values were analyzed with respect to relative tidal volumes. (3) Finally, the mutual relation between ΔP a O 2 and P a CO 2 as a result of changes in tidal volume was examined. A power function in the form of y = ax b + c was used to fit each of the data groups (CV and HFOV) in the figures P a CO 2 vs. V T /V T norm and ΔP a O 2 vs. V T /V T norm . The linear function was used to fit the data groups in the ΔP a O 2 vs. P a CO 2 figure. For the fitted curves, 95% functional prediction intervals were calculated. The curve fitting was performed in MATLAB Curve Fitting Toolbox (MathWorks Inc., Natick, MA, USA). The coefficient of determination (R 2 ) was calculated for each of the fitted curves to evaluate the goodness of fit. For a pair of fitted curves in each of the figures, a version of the F-test applied to nonlinear models was used to determine whether the HFOV and CV trends differ significantly [46,47]. P values less than 0.05 were considered as statistically significant.