AJRCCM Articles in Press. Published on October 3, 2002 as doi:10.1164/rccm.200207-717OC

Effects of Respiratory Rate, Plateau Pressure, and PEEP on PaO2 Oscillations after Saline Lavage James E. Baumgardner,1,2 Klaus Markstaller,3,4 Birgit Pfeiffer,5 Marcus Doebrich,4 and Cynthia M. Otto6,7 1 2 3 4 5 6 7

Department of Anesthesia, University of Pennsylvania , Philadelphia, PA 19104 SpectruMedix LLC, State College, PA 16803 Department of Anesthesiology, Johannes Gutenberg-University, 55131 Mainz, Germany Department of Radiology, Johannes Gutenberg-University, 55131 Mainz, Germany Department of Anesthesiology and Intensive Care Medicine, Ernst-MoritzArndt-University, 17489 Greifswald, Germany Department of Clinical Studies-Philadelphia, School of Veterinary Medicine, University of Pennsylvania, Philadelphia, PA 19104 Center for Sleep and Respiratory Neurobiology, University of Pennsylvania, Philadelphia, PA 19104

Address for Correspondence: James E. Baumgardner, MD, PhD Department of Anesthesia Hospital of the University of Pennsylvania 3400 Spruce St. Philadelphia, PA 19104-4283 [email protected] 215-662-4272 215-349-5078 (fax) Sources of Support: Supported in part by NIH GM59274; NIH HL59052; AHA 0151528U; DFG MA2398/2; DFG PF424/1; SpectruMedix LLC, State College PA; the Departments of Anesthesia at the University of Pennsylvania, the University of Mainz, and the University of Greifswald; and the Emergency Service at the School of Veterinary Medicine, University of Pennsylvania. Running Head: PaO2 Oscillations, Respiratory Rate Subject Category: 1 Word Count for Body of Manuscript: 4040 This article has an online data supplement, which is accessible from this issue’s table of content online at www.atsjournals.org.

Copyright (C) 2002 by the American Thoracic Society.

1

Abstract One of the proposed mechanisms of ventilator-associated lung injury is cyclical recruitment of atelectasis. Collapse of dependent lung regions with every breath should lead to large oscillations in arterial PO2 (PaO2) as shunt varies throughout the respiratory cycle. We placed a fluorescent-quenching PO2 probe in the brachiocephalic artery of 6 anesthetized rabbits after saline lavage. Using pressure-controlled ventilation with oxygen, ventilator settings were varied in random order over 3 levels of PEEP, respiratory rate (RR), and plateau pressure minus PEEP (delta). Dependence of the amplitude of PaO2 oscillations on PEEP, RR, and delta was modeled by multiple linear regression. Before lavage, arterial PO2 oscillations varied from 3 to 22 Torr. After lavage, arterial PO2 oscillations varied from 5 to 439 Torr. Response surfaces showed markedly non-linear dependence of amplitude on PEEP, RR, and delta. The large PaO2 oscillations observed provide evidence for cyclical recruitment in this model of lung injury. The important effect of respiratory rate on the magnitude of PaO2 oscillations suggests that the static behavior of atelectasis cannot be accurately extrapolated to predict dynamic behavior at realistic breathing frequencies.

Number of words in abstract: 181

Keywords: Respiratory Distress Syndrome, Adult; atelectasis; cyclical recruitment; ventilator-associated lung injury

2

Introduction Ventilator-associated lung injury (VALI) in the acute respiratory distress syndrome (ARDS) is thought to damage the alveolar epithelium and capillary endothelium primarily by stretch injury. Overdistention of normal but functionally small parts of the lung with normal tidal volumes can result in mechanical injury and an inflammatory response that further damages the acutely injured lung (1-3). Recently a large multi-center trial demonstrated a significant reduction in mortality in ARDS by use of protective ventilation strategies that included reduced tidal volumes to minimize overdistention (4). An additional mechanism of VALI is thought to be the stretch injury associated with repetitive collapse and re-expansion of atelectasis with each breath, often called cyclical recruitment. Protective ventilation strategies seek to minimize cyclical recruitment by use of high levels of extrinsic PEEP to keep the lung maximally recruited throughout the respiratory cycle (1-3). Several experimental studies have suggested that increased extrinsic PEEP can reduce VALI (5-9). In contrast, preliminary results of the ARDS Network ALVEOLI study (http://hedwig.mgh.harvard.edu/ardsnet/ards04.html) showed no benefit to increased PEEP. Details of this study required for critical evaluation, however, have not been published to date. Recently, de Durante et al. showed that the higher respiratory rate associated with the low tidal volume ventilation strategy in the ARDS Network study (4) elevates intrinsic PEEP, and suggested this may have been a factor in reducing VALI by reducing cyclical recruitment (10). The effects of extrinsic PEEP in reducing atelectasis and cyclical recruitment in ARDS at very low respiratory rates have been extensively

3

investigated (11-14). The effects of respiratory rate specifically on cyclical recruitment, however, have not been reported. Williams et al. recently showed in saline-lavaged dogs that arterial oxygen tension oscillated at a frequency matching the respiratory rate, with large amplitude at certain ventilator settings. The large magnitude of these oscillations could only be explained by changes in shunt fraction within the respiratory cycle, a phenomenon consistent with cyclical recruitment (15). Their probe was not fast enough to accurately record the amplitude of PO2 oscillations at high respiratory rate, and there have been no reports to date of the effects of respiratory rate on PaO2 oscillations. The aim of our study was to investigate the effects of respiratory rate, extrinsic PEEP, and plateau pressure minus PEEP on the amplitude of PaO2 oscillations. We placed a fast PO2 probe in rabbits after saline lavage to accurately measure PaO2 oscillations over a range of respiratory rates. We specifically tested the hypothesis that respiratory rate has a significant effect on PaO2 oscillations. The study of the effects of several predictor variables is most efficiently approached by choice of several levels of each of the predictor variables and analysis of the response variable with multiple linear regression. Although this multivariable approach to experimental design has not been applied frequently in the physiology literature, it is well established in engineering analysis (16). Some of the results of these studies have been previously reported in the form of an abstract (17).

4

Methods Number of Words in Methods Section: 490 We chose an experimental design of 3 levels for each of the 3 independent variables: respiratory rate (RR); delta (Pplat – PEEP); and extrinsic PEEP. The levels of RR chosen were 10, 20, and 30 breaths per minute. The levels of delta chosen were 20, 30, and 40 cm H20. Pilot studies showed unacceptable hypoxemia at PEEP of zero after lavage, and unacceptable hemodynamic compromise at PEEP greater than 16 cm H2O. The levels of extrinsic PEEP chosen were therefore 4, 10, and 16 cm H2O. Pilot studies also showed an unacceptably high incidence of pneumothorax at high peak pressures, and the upper ranges of delta and PEEP were therefore jointly modified at the highest levels to keep maximum airway pressure less than 46 cm H2O. A detailed description of the 27 ventilator settings, and experimental design considerations that led to these choices, can be found in the online data supplement. The study protocol was approved by the Institutional Animal Care and Use Committee at the University of Pennsylvania. The experimental protocol was to induce general anesthesia, prepare the rabbits with tracheostomy, vascular catheters, and the PO2 probe, and calibrate the probe in vivo for each animal against standard blood gas analysis. PaO2 oscillations were then recorded for normal lungs at 2 ventilator settings, and mild to moderate lung injury was induced by saline lavage. Throughout the experimental study, the rabbits were ventilated (Servo 900C, Siemens, Germany) in pressure-controlled mode (PCV) with an FIO2 of 1.0, and I:E ratio of 1:1. After lavage, the ventilator was adjusted to each of the 27 pre-chosen settings in random order, and the PaO2, respiratory

5

mechanical, and hemodynamic data were recorded at each setting. Colloid boluses and IV epinephrine infusions were used as necessary for hemodynamic support. The oxygen probe used for these studies was a fiber optic, fluorescencequenching probe with an uncoated ruthenium complex at the probe tip (FOXY-AL300, Ocean Optics, Dunedin, FL). Details of the spectrometer, in vitro bench studies to test for possible signal artifacts, probe time response, and criteria for proper in vivo placement can be found in the online data supplement. After digital filtering to reduce high-frequency noise components, the peak-topeak amplitude of PaO2 was modeled as the predicted variable in a multiple linear regression with RR, delta, and extrinsic PEEP as predictor variables (16). Solutions to the normal equations, residual plots and analysis, and calculations of sums of squares were carried out in Mathcad 2000i (Mathsoft, Cambridge MA). Details of the development of the model and analysis of residuals can be found in the online supplement. The full model for the amplitude data was:

3

Y

2

2

2

β 0⋅ X 0 + β 1⋅ X 1 + β 2⋅ X 2 + β 3⋅ X 3 + β 11⋅ X 1 + β 22⋅ X 2 + β 33⋅ X 3 + β 12⋅ X 1⋅ X 2 + β 13⋅ X 1⋅ X 3 + β 23⋅ X 2⋅ X 3 + β 4⋅ X 4 + ε

where Y is the peak-to-peak amplitude, X0 is the constant term, X1 is RR, X2 is delta , X3 is PEEP, and X4 is order number. The predictive values of RR, delta, and PEEP were tested by partial F testing, comparing the full model to models with the terms for each of these predictors deleted.

6

Results Unless otherwise noted, amplitude of the arterial PaO2 oscillations refers to peakto-peak amplitude. Ventilator settings are abbreviated in the format of RR20 delta20 PEEP10 to denote a respiratory rate (RR) of 20 breaths/minute, plateau airway pressure minus extrinsic PEEP (delta) of 20 cm H20, and extrinsic PEEP of 10 cm H20. All data are presented in the format mean ± standard deviation. Four animals received 3 lavages and two animals received 4 lavages. The lavages (26 ml/Kg per lavage) induced a mild to moderate degree of acute lung injury, as demonstrated in Table 1. Systemic blood pressure and heart rate, at identical ventilator settings, were similar after lavage and the pulmonary artery pressure was mildly elevated. The reduction in PaO2 after lavage, on 10 cm H2O of PEEP, was significant but moderate. After lavage, dynamic compliance actually increased rather than decreased, consistent with operating at the flat upper end of the left-shifted pressure-volume curve (PV curve) of the normal rabbit versus the steeper normal part of the curve when applying 10 cm H20 of PEEP after lavage (18, 19). Average rectal temperature at the completion of the protocol was 37.8 ± 1.0 oC. Venous blood gas data at the completion of the protocol showed a hemoglobin of 7.9 ± 0.9 g/dL, and base deficit of 6.9 ± 6.0 mEq/L. Prior to lavage, the amplitude of PaO2 oscillations ranged from 3.3 to 21.7 Torr. The average amplitude of the PaO2 oscillations at ventilator settings of RR20 delta20 PEEP10 was 8.9 ± 5.8 Torr (n=6). After lavage the peak-to-peak amplitude of the PaO2 oscillations ranged from 5.3 Torr at a setting of RR30 delta20 PEEP16, to 439 Torr at a

7

setting of RR10 delta40 PEEP4. The average amplitude of the PaO2 oscillations at ventilator settings of RR20 delta20 PEEP10, after lavage, was 130.8 ± 62.2 Torr (n=6). Figures 1a and 1b illustrate two representative examples of the effect of respiratory rate on the amplitude of oscillation and on mean PaO2, after saline lavage. Decreasing respiratory rate from 20 to 10 caused a large increase in amplitude, and increasing the respiratory rate from 10 to 30 caused a large decrease in amplitude, in these examples. Figures 1a and 1b also illustrate the rapid establishment of a new steadystate in PaO2 oscillations after a change in the respiratory rate, with other ventilator settings held constant. Similar plots showing rapid establishment of a new steady-state after changes in PEEP and delta are presented in the online supplement. Figures 2a, b, and c illustrate a specific example of the effects of respiratory rate and PEEP on PaO2 oscillations. In this rabbit, similar mean PaO2 and amplitude were achieved by low respiratory rate and high PEEP in figure 2a, and lower PEEP and high RR in figure 2b. Figure 2c illustrates the effect of reducing PEEP at the lower RR. Average data for all of the rabbits is plotted in the three dimensional response surfaces showing the effects of PEEP and RR on amplitude, at three fixed delta values, in Figure 3. These surfaces are the plots of the complete regression model (Equation 1) that fits the data from all rabbits that completed the protocol (n=6). The response surfaces show highly nonlinear effects of both PEEP and RR on amplitude, with substantial interactions between PEEP and RR. Figure 4 presents the three dimensional response surfaces showing the effects of PEEP and delta on amplitude, for three fixed RR. These response surfaces also show a high degree of nonlinearity and of interaction between the predictor variables. For clarity, individual data points scattered around the response

8

surface plots have been omitted. Variability between individuals is illustrated in Figure 5, where the surface of Figure 3b has been expanded into three 2-dimensional plots, with variability between individuals represented by the error bars around the plotted points. Peak values of PaO2 for each ventilator setting are presented in the online supplement. Partial F testing for the amplitude model showed that the order of predictive value for the 3 independent variables is: RR>PEEP>>delta (F values 86.4, 84.1, and 12.1). These F values are associated with P values of