High-frequency oscillatory ventilation (HFOV) is frequently used for children with moderate to severe pediatric acute respiratory distress syndrome (PARDS) even though there is limited controlled evidence of improved outcomes (1–3). In theory, HFOV provides a lung protective strategy by maintaining alveolar ventilation at low intrapulmonary pressure amplitudes and tidal volumes (VT). This lung protective strategy may minimize ventilator-induced lung injury by avoiding atelectasis or overdistention of alveoli, the danger zones of mechanical ventilation (4–6).
Clinicians do not routinely measure the delivered VT or mean airway pressure (MAP) during HFOV. In the absence of direct measurements, clinicians depend on chest radiographs, observed chest wall vibration, and blood gas monitoring to guide HFOV strategy. Perhaps, this lack of monitoring and resulting variability of practice in ventilator management may contribute to some of the negative results in recent trials of HFOV in both adults and children, as the HFOV protocol and prescription may affect the degree of lung protection (7–10). Because routine monitoring of distal volume and pressure changes is not generally used or available, test lung models have played an important role in understanding the interactions between HFOV, endotracheal tube (ETT) size, and lung mechanics (11–16).
Delivered VT in neonatal and adult test lung models have been measured using a Florian hot-wire anemometer (CE0124; Acutronics Medical Systems, Barr, Switzerland) (13, 15, 17). However, a number of conditions limited the Florian’s accuracy in measuring delivered VT and therefore its clinical application. In neonatal test models, errors were in the ±20% range at frequencies less than 8 Hz and greater than 13 Hz (13). In adult lung models, the Florian was unreliable in measuring delivered VT at inspiratory-to-expiratory time (I:E) ratio of 1:2, frequencies 3–4 Hz, and amplitude of 30 cm H2O (17). In both studies, frequency-dependent flow integration algorithms were required to improve measurement accuracy.
The primary objective of this study is to determine the variables independently associated with delivered VT and MAP for a given ETT size during HFOV. For this purpose, we measured and calculated the delivered VT and MAP within in vitro test lungs representing a range of compliances encountered in PARDS. We chose to perform this study using the Sensormedics 3100A and 3100B oscillators (SensorMedics, Yorba Linda, CA) as they are the two most frequently used HFOV ventilators in the United States (18). The secondary objective was to construct multivariable models for each ETT size that quantify the association between independent variables and delivered VT, and construct separate multivariable models for each ETT size that quantify the association between independent variables and delivered MAP.
MATERIALS AND METHODS
We used an in vitro bench model of the intubated pediatric respiratory system during HFOV (Fig. 1). The model consisted of a high-frequency oscillatory ventilator connected to a static test lung via an ETT. Flow and proximal airway pressures were measured by a Florian hot-wire monitor (CE0124; Acutronics Medical Systems, Barr, Switzerland) between the Y-connector of the ventilator and ETT. The Florian has an analog output delay of less than or equal to 2 ms and was used to provide real-time visual feedback of flow and pressure within the system. Measurements obtained from the Florian were not used in data analysis or calculations. A calibrated, rapid response, Heise pressure transducer (901A; Ashcroft, Stratford, CT) simultaneously measured the delivered oscillatory waveform pressure within the test lung. The Heise pressure transducer (Ashcroft) has an analog output accuracy of 0.2% of span and response time of 3 ms. A data acquisition system (DI-155; DataQ, Akron, OH) sampled and stored all analog pressure and flow waveforms at a rate of 1,000 Hz.
Each test lung system was validated prior to the start of testing and immediately after testing. Volume calibration syringes (5510, 5540, and 5550 Series; Hans Rudolph, Shawnee, KS) were used to inject a known volume of air into the test lung, and the pressure waveform was recorded during the injection. The accuracy of each syringe is ±0.5 of 1% full scale. Based on known compliance of the test lung and known volume of air injected from the syringe, the resulting pressure was calculated. Calculated pressure was used to verify measured and sampled pressures from the Heise pressure transducer (Ashcroft) and data acquisition system. This process was repeated at the end of experimentation with each test lung.
Ventilators and ETTs.
The calibration and performance checks were completed on the 3100A and 3100B (SensorMedics) ventilators in accordance with the manufacturer’s instructions. Both ventilators used a nonhumidified source gas for calibration and testing. Consistent with clinical application, the Sensormedics 3100A (SensorMedics) was tested with ETT sizes 3.0–5.0 mm internal diameter (ID). The Sensormedics 3100B (SensorMedics) is recommended for patients weighing greater than 35 kg; therefore, it was tested with the 6.0 and 7.0 mm ID ETT.
We used cylindrical rigid test lungs partially filled with copper strands for this study. Test lung volume and compliance were precisely recorded. Each test lung was weighed before and after being filled with water (wet method). Volume was calculated by subtracting the empty test lung weight from the water-filled test lung weight. Compliance was calculated by injecting a known volume of air and measuring the pressure within the test lung (dry method).
Each lung was tested for absence of air leaks and the ability to hold pressure constant. The high heat capacity of the copper strands allowed the nonhumidified gas to remain at a constant temperature creating an isothermal condition. Test lung compliances were selected to represent moderate to severe stages of PARDS.
We designed the experimental protocol based on conditions relevant to the clinical setting (Fig. 2). All the experiments were conducted at ambient temperature, pressure, and humidity. Measurements were made of test lung pressure at every combination of frequency (f: 5, 8, 10, and 12 Hz), amplitude (ΔP: 30, 40, 60, and 80 cm H2O), and ventilator set MAP (Paw: 20, 30, and 40 cm H2O). Power setting was adjusted to achieve specific amplitudes. I:E ratio was held constant at 1:2 across all measurements, as this variable is rarely adjusted in the pediatric clinical setting. Bias flow was also held constant at 30 L/min on the 3100B ventilator (SensorMedics). However, in order to achieve a Paw of 40 cm H2O on the 3100A (SensorMedics), the bias flow was increased from the recommended 20 to 30 L/min. Test lung compliances were chosen for each ETT size that would approximate moderate (0.6 mL/cm H2O/kg), moderate-severe (0.3 mL/cm H2O/kg), and severe (0.15 mL/cm H2O/kg) PARDS.
Methods of Measurement
We collected 30 cycles of pressure and flow signals for each combination of variables tested. The start of each cycle was identified by the initiation of positive flow measured by the Florian HW anemometer (Acutronics Medical Systems). The waveforms were stored using data acquisition and playback software (Windaq; DataQ, Akron, OH).
Delivered VT was calculated using the ideal gas equation based on Boyle’s law (VT = C · ΔPTL), where C is the known compliance of the test lung and ΔPTL is the change from peak-to-trough pressures within the test lung averaged over 10 waveform cycles. Because inspiration and exhalation are active during HFOV, the total delivered VT during one duty cycle accounts for the trough-to-peak change in pressure within that cycle.
We calculated the delivered MAP using the stored test lung pressure waveform. The calculations were done by taking the average pressure within the test lung, using every data point within 10 duty cycles, sampled at a frequency of 1,000 Hz.
The study variables were analyzed using Statistica version 12 (StatSoft, Tulsa, OK). We sought to determine which variables are independently associated with delivered VT and delivered MAP using univariate regression analysis, stratified by ETT size. For each ETT size, any of the study variables that had a univariate relationship with delivered VT or MAP were considered for inclusion in a multivariable linear regression model predicting delivered VT and a separate model predicting delivered MAP. For variables that retained significance (p < 0.05) in the multivariable models, we created multiplicative interaction terms to identify effect modifications. Overall, model fit was assessed using adjusted R2 with appropriate regression diagnostics ensuring all assumptions of linear regression were satisfied.
Tidal Volume Delivery in Distal Test Lung
On univariate regression analysis, ΔP (p < 0.001), frequency (p < 0.001), and test lung compliance (p < 0.001) were independently associated with delivered VT for all ETT sizes. Both higher test lung compliance and higher set ΔP were associated with higher delivered VT, whereas a lower frequency was associated with higher delivered VT across the range of pediatric ETT sizes. Multiple regression analysis also showed the multiplicative interaction between decreasing frequency and increasing ΔP (Fig. 3A) magnifying delivered VT across all ETT sizes (p < 0.001) (Table 1).
The adjusted R2 values for all delivered VT multivariable models were greater than 0.85. Based on the multivariable models, decreasing the frequency as compared to increasing the ΔP is associated with a greater percent increase in delivered VT. For example, in a 3.0 mm ID ETT and lung compliance of 0.5 mL/cm H2O, an increase in ΔP from 40 to 44 cm H2O (10%) at a frequency of 10 Hz is associated with a 6.9% increase in delivered VT. Decreasing the frequency by 10%, from 10 to 9 Hz, while maintaining a ΔP of 40 cm H2O, is associated with a 9.5% increase in delivered VT. The same increase in ΔP, from 40 to 44 cm H2O, at a lower frequency, 9 Hz, is associated with a magnified increase in delivered VT by 17.2% (Table 2).
MAP Delivery in Distal Test Lung
On univariate regression analysis, ventilator set Paw (p < 0.001), ΔP (p < 0.001), and frequency (p < 0.05) were independently associated with delivered MAP for all ETT sizes. Test lung compliance (p < 0.001) was independently associated with delivered MAP for ETT sizes 3.0–5.0 mm ID (Table 3). Multiple regression analysis also showed the multiplicative interaction between frequency and ΔP was independently associated with delivered MAP for all ETT sizes except for the 5.0 mm ID ETT (Table 3).
The adjusted R2 values for all delivered MAP multivariable models were greater than 0.95. The multivariable models show that set Paw is the largest determinant of delivered MAP (p < 0.001); however, increasing the ΔP results in a lower delivered MAP (Fig. 3B). At low ΔP, delivered MAP approximated set Paw. But as ΔP is increased, there is a reduction in delivered MAP. For example, in a 6.0 mm ID ETT and lung compliance of 5 mL/cm H2O, with a frequency of 10 Hz and an amplitude of 70 cm H2O, the delivered MAP is 5.5 cm H2O lower than the set Paw of 30 cm H2O. Increasing the amplitude to 80 cm H2O (14% increase from 70 cm H2O) further decreased the delivered MAP 6.4 cm H2O lower than the set Paw of 30 cm H2O. In contrast, decreasing the frequency by the same 14%, from 10 to 8.6 Hz, with a ΔP of 70 cm H2O did not further decrease delivered MAP (Table 4).
Similar to other neonatal and adult studies, we have demonstrated an association between ΔP, frequency, and lung compliance with delivered VT (13–15, 19–21). However, unlike other studies, we have quantified the importance of the interaction between frequency and amplitude on delivered VT, as a function of ETT size. Furthermore, we are the first to quantify the degree to which distal lung MAP gets attenuated over set Paw as a function of increasing amplitude. These findings may be important as we consider optimal ways to maintain a lung protective strategy when managing HFOV.
In neonatal studies, Scalfaro et al (13) found that increasing ΔP or decreasing frequency was associated with larger delivered VT using a 2.5 mm ID ETT during HFOV. Pillow et al (14, 19) demonstrated, in an ETT size range of 2.5–4.0 mm ID, that increased lung compliances are associated with larger delivered VT. In a separate study, Pillow et al (20) described that increasing ΔP produced a concomitant decrease in delivered MAP in both animal and bench lung models using ETTs of 2.5–3.5 mm ID.
As applies to adult populations, Sedeek et al (21) (ETT sizes not specified) and Hager et al (15) (ETT of 6, 7, and 8 mm ID) found that increasing ΔP, decreasing frequency, or increasing test lung compliance was associated with larger delivered VT. Their data and graphs suggest, but did not quantify, a multiplicative interaction between frequency and ΔP on delivered VT. The association between ΔP and delivered MAP was not evaluated in either adult lung models or ventilated adult patients.
Our study verifies that the multiplicative interaction between frequency and ΔP is independently associated with delivered VT, not only in adult ETT sizes but also across the range of pediatric ETT sizes (3.0–7.0 mm ID). We also found that an increase in ΔP was independently associated with decreased delivered MAP across the full range of pediatric ETT sizes, not only in neonatal ETTs. For each ETT size, our multivariable models quantify the multiplicative interaction of frequency and ΔP, on delivered VT. These models also quantify the independent association between changes in ventilator set ΔP and delivered MAP.
Although our study was a bench model, our observations have important clinical implications. Management of HFOV is dependent on the goals of mechanical ventilation, which must evolve along with the progression of PARDS. If the goal is to minimize delivered VT (allowing permissive hypercapnia) and maintain MAP because of inadequate oxygenation, then perhaps the best lung protective strategy includes prioritizing lower ΔP and middle range values for frequency. If oxygenation is less compromised (and a lower delivered MAP can be tolerated), then perhaps the best lung protective strategy involves prioritizing increases in frequency, allowing for somewhat higher ΔP.
It is important to remember that adjustments to frequency will have a different effect on delivered VT depending on the set ΔP. Increases in frequency while on a higher ΔP will result in larger decreases to delivered VT. In contrast, the same increases in frequency while on a lower ΔP will have relatively less of an effect on delivered VT.
To our knowledge, we are the first to describe the relationship between ΔP and delivered MAP across the range of pediatric ETT sizes. We speculate that the lower delivered MAP with a high ΔP may be the result of more time spent in active exhalation when compared with active inspiration given a set I:E ratio of 1:2. Approximately, 66% of the duty cycle is spent in active exhalation which lowers measured peak and trough distal airway pressure directly resulting in a lowered MAP.
There are important limitations to our study. The rigid test lung allows for accurate and precise measurements and calculations. It was not designed to measure gas exchange, histologic markers, or biologic markers. It also does not provide information regarding cardiopulmonary interactions (e.g., hemodynamic data). In addition, the simple rigid test lung differs from the heterogeneous respiratory system in PARDS. It lacks the dynamic properties of patient lungs and does not take into account the interactions of the chest wall or diaphragm. Therefore, this model does not fully account for the nonlinear relationship between volume and pressure at extremes of respiratory compliance. The linear multivariable models generated adjusted R2 values of greater than 0.85 even though the relationship between variables is not perfectly linear. These limitations underscore the importance of recreating these experiments in humans, once appropriate sensors and methods are available to measure tidal volume and delivered MAP.
Currently, there are no consistent HFOV ventilator guidelines regarding setting adjustments. Protocols that describe HFOV strategy differ across the scientific literature (8, 9, 15, 22, 23). The complex interactions between independent variables make delivered VT and MAP predictions difficult. The multivariable models constructed in our study allow estimation of delivered VT and MAP under varying and relevant clinical conditions. The results obtained from this study may guide future HFOV strategies and clinical trial designs needed in the pediatric population (24). Perhaps, the future development of HFOV ventilators will provide delivered VT and MAP values to optimize lung protective strategies. Even prioritizing the effective changes in HFOV management may improve protocols created for clinical practice and research.
We thank CareFusion for use of their test lungs in conducting this research.
1. Rimensberger PC, Cheifetz IM; Pediatric Acute Lung Injury Consensus Conference Group: Ventilatory support in children with pediatric acute respiratory distress syndrome: Proceedings from the pediatric acute lung injury consensus conference. Pediatr Crit Care Med 2015; 16(5 Suppl 1):S51–S60.
2. Khemani RG, Smith LS, Zimmerman JJ, et al; Pediatric Acute Lung Injury Consensus Conference Group: Pediatric acute respiratory distress syndrome: Definition, incidence, and epidemiology: Proceedings from the pediatric acute lung injury consensus conference. Pediatr Crit Care Med 2015; 16(5 Suppl 1):S23–S40.
3. Rettig JS, Smallwood CD, Walsh BK, et al. High-frequency oscillatory ventilation in pediatric acute lung injury: A multicenter international experience. Crit Care Med 2015; 43:2660–2667.
4. International consensus conferences in intensive care medicine. Ventilator-associated lung injury in ARDS. American Thoracic Society, European Society of Intensive Care Medicine, Societe de Reanimation Langue Francaise. Intensive Care Med 1999; 25:1444–1452.
5. Arnold JH. High-frequency ventilation in the pediatric intensive care unit. Pediatr Crit Care Med 2000; 1:93–99.
6. Imai Y, Slutsky AS. High-frequency oscillatory ventilation and ventilator-induced lung injury. Crit Care Med 2005; 33(Suppl 3):S129–S134.
7. Bateman ST, Borasino S, Asaro LA, et al; RESTORE Study Investigators: Early high-frequency oscillatory ventilation in pediatric acute respiratory failure: A propensity score analysis. Am J Respir Crit Care Med 2016; 193:495–503.
8. Ferguson ND, Cook DJ, Guyatt GH, et al; OSCILLATE Trial Investigators; Canadian Critical Care Trials Group: High-frequency oscillation in early acute respiratory distress syndrome. N Engl J Med 2013; 368:795–805.
9. Young D, Lamb SE, Shah S, et al; OSCAR Study Group: High-frequency oscillation for acute respiratory distress syndrome. N Engl J Med 2013; 368:806–813.
10. Gupta P, Green JW, Tang X, et al. Comparison of high-frequency oscillatory ventilation and conventional mechanical ventilation in pediatric respiratory failure. JAMA Pediatr 2014; 168:243–249.
11. Fredberg JJ, Glass GM, Boynton BR, et al. Factors influencing mechanical performance of neonatal high-frequency ventilators. J Appl Physiol (1985) 1987; 62:2485–2490.
12. Jouvet P, Hubert P, Isabey D, et al. Assessment of high-frequency neonatal ventilator performances. Intensive Care Med 1997; 23:208–213.
13. Scalfaro P, Pillow JJ, Sly PD, et al. Reliable tidal volume estimates at the airway opening with an infant monitor during high-frequency oscillatory ventilation. Crit Care Med 2001; 29:1925–1930.
14. Pillow JJ, Wilkinson MH, Neil HL, et al. In vitro performance characteristics of high-frequency oscillatory ventilators. Am J Respir Crit Care Med 2001; 164:1019–1024.
15. Hager DN, Fessler HE, Kaczka DW, et al. Tidal volume delivery during high-frequency oscillatory ventilation in adults with acute respiratory distress syndrome. Crit Care Med 2007; 35:1522–1529.
16. Custer JW, Ahmed A, Kaczka DW, et al. In vitro performance comparison of the Sensormedics 3100A and B high-frequency oscillatory ventilators. Pediatr Crit Care Med 2011; 12:e176–e180.
17. Hager DN, Fuld M, Kaczka DW, et al. Four methods of measuring tidal volume during high-frequency oscillatory ventilation. Crit Care Med 2006; 34:751–757.
18. Kneyber MC, van Heerde M, Markhorst DG. Reflections on pediatric high-frequency oscillatory ventilation from a physiologic perspective. Respir Care 2012; 57:1496–1504.
19. Pillow JJ, Sly PD, Hantos Z, et al. Dependence of intrapulmonary pressure amplitudes on respiratory mechanics during high-frequency oscillatory ventilation in preterm lambs. Pediatr Res 2002; 52:538–544.
20. Pillow JJ, Neil H, Wilkinson MH, et al. Effect of I/E ratio on mean alveolar pressure during high-frequency oscillatory ventilation. J Appl Physiol (1985) 1999; 87:407–414.
21. Sedeek KA, Takeuchi M, Suchodolski K, et al. Determinants of tidal volume during high-frequency oscillation. Crit Care Med 2003; 31:227–231.
22. Arnold JH, Anas NG, Luckett P, et al. High-frequency oscillatory ventilation in pediatric respiratory failure: A multicenter experience. Crit Care Med 2000; 28:3913–3919.
23. Fessler HE, Derdak S, Ferguson ND, et al. A protocol for high-frequency oscillatory ventilation in adults: Results from a roundtable discussion. Crit Care Med 2007; 35:1649–1654.
24. Kneyber MC, van Heerde M, Markhorst DG. It is too early to declare early or late rescue high-frequency oscillatory ventilation dead. JAMA Pediatr 2014; 168:861
acute lung injury; high-frequency ventilation; in vitro techniques; mechanical ventilation; pediatric intensive care units©2017The Society of Critical Care Medicine and the World Federation of Pediatric Intensive and Critical Care Societies