Recruitment Maneuvers in Three Experimental Models of Acute Lung Injury Effect on Lung Volume and Gas Exchange

Transcription

1 Recruitment Maneuvers in Three Experimental Models of Acute Lung Injury Effect on Lung Volume and Gas Exchange THOMAS E. VAN der KLOOT, LLUIS BLANCH, A. MELYNNE YOUNGBLOOD, CRAIG WEINERT, ALEX B. ADAMS, JOHN J. MARINI, ROBERT S. SHAPIRO, and AVI NAHUM Department of Pulmonary and Critical Care Medicine, University of Minnesota, Regions Hospital, St. Paul, Minnesota; and Servei de Medicina Intensiva, Hospital de Sabadell, Corporacio Hospitalaria del Parc Tauli, Sabadell, Spain Recruitment maneuvers (RM), consisting of sustained inflations at high airway pressures, have been advocated as an adjunct to mechanical ventilation in acute respiratory distress syndrome (ARDS). We studied the effect of baseline ventilatory strategy and RM on end-expiratory lung volume (EELV) and oxygenation in 18 dogs, using three models of acute lung injury (ALI; n 6 in each group): saline lavage (LAV), oleic acid injury (OAI), and intratracheal instillation of Escherichia coli (pneumonia; PNM). All three models exhibited similar degrees of lung injury. The PNM model was less responsive to positive end-expiratory pressure (PEEP) than was the LAV or OAI model. Only the LAV model showed an oxygenation response to increasing tidal volume (VT). After RM, there were transient increases in Pa O2 and EELV when ventilating with PEEP 10 cm H 2 O. At PEEP cm H 2 O the lungs were probably fully recruited, since the plateau airway pressures were relatively high ( 45 cm H 2 O) and the oxygenation was similar to preinjury values, thus making the system unresponsive to RM. Sustained improvement in oxygenation after RM was seen in the LAV model when ventilating with PEEP 10 cm H 2 O and VT 15 ml/kg. Changes in EELV correlated with changes in Pa O2 only in the OAI model with PEEP 10 cm H 2 O. We conclude that responses to PEEP, VT, and RM differ among these models of ALI. RM may have a role in some patients with ARDS who are ventilated with low PEEP and low VT. Extensive animal research suggests that parenchymal lung injury and an inflammatory response may be caused and perpetuated if mechanical ventilation results in alveolar overdistention (1, 2) and allows cyclic collapse and reinflation of alveolar units with tidal breathing (3 8). On the basis of this work, pressure-limited, lung-protective mechanical ventilatory strategies have been proposed for acute respiratory distress syndrome (ARDS) (9 11), emphasizing the need to open the lung and keep it open (12) while avoiding alveolar overdistention. Amato and colleagues (13), in a randomized controlled trial, recently found reduced mortality in ARDS patients managed with such an approach. Recruitment maneuvers (RM) consisting of sustained inflations (SI) have been advocated as adjuncts to pressure-limited, lung-protective ventilatory strategies. In the protective strategy of Amato and colleagues (13), RM consisting of continuous positive airway pressure (CPAP) of 35 to 40 cm H 2 O were applied for 40 s at the airway opening frequently... especially after disconnections from the ventilator. The recent American European Consensus Conference on ARDS statement suggests that It (Received in original form September 4, 1998 and in revised form October 21, 1999) Supported by the Regions Foundation and SCOR grant HL from the National Institutes of Health. Dr. Blanch was supported by grant BAE 97/5478 from the FIS (Spain). Correspondence and requests for reprints should be addressed to Avi Nahum, M.D., Ph.D., Regions Hospital, Pulmonary and Critical Care, 640 Jackson Street, St. Paul, MN avi.x.nahum@healthpartners.com Am J Respir Crit Care Med Vol 161. pp , 00 Internet address: may be advisable to periodically use larger volume, higher pressure breaths of longer inspiratory duration to forestall atelectasis when very small tidal volumes and/or low PEEP are used (14). The rationale for these RM is that collapsed alveolar units require relatively high pressures for reexpansion (15). The use of smaller tidal volumes (VT) may limit the opening of these units during tidal ventilation, promoting atelectasis and thereby leading to progressive derecruitment. Rothen and colleagues (16) have shown in normal human subjects undergoing general anesthesia that RM result in a sustained increase in lung volume. In surfactant-depletion models of acute lung injury (ALI), RM have been shown to be capable of reopening collapsed alveolar units during apneic oxygenation (17) and high frequency ventilation (HFV) (18 21). Once recruited, these units require less pressure to maintain their patency, especially in surfactant-depleted lungs (12). Nevertheless, it is crucial to apply sufficient levels of PEEP to prevent their collapse and avoid ventilator-induced lung injury (VILI). Previous studies with animal models of ALI (18, 21) and with humans (22) have not shown any benefit to oxygenation, lung mechanics, and/or lung volume of RM during conventional mechanical ventilation (CMV) at high VT. Limited data exist on the safety and clinical role of RM during low VT CMV. In saline-lavaged rabbits, Bond and colleagues (21) showed that oxygenation improved after RM during CMV only when ventilating with a low VT (7 ml/kg) and zero end-expiratory pressure (ZEEP). Furthermore, with one exception (18), published studies of the effects of RM used a surfactant-depletion model, which may not be fully representative of all ARDS. Recent clinical studies have suggested that there may exist subgroups of patients with ARDS, based on the etiology of lung injury, that behaves differently in terms of lung mechanics and response to mechanical ventilatory strategies (23, 24). The response of patients with ARDS to different ventilatory strategies may thus depend on the nature of the lung insult, with consolidative processes being least responsive to recruitment strategies (25). The intention of our study was to determine the effect of RM on oxygenation and end-expiratory lung volume (EELV), and on the evolution of these latter two parameters over time, in three models of ALI (saline lavage [LAV], oleic acid injury [OAI], and intratracheal instillation of bacteria pneumonia [PNM]). We hypothesized that the response to RM would vary among the models and would depend on the baseline ventilatory strategy (i.e., VT and PEEP) before and after the application of RM. We elected to study the effect of RM applied during CMV at PEEP levels below and above the lower inflection point (Plip) of the pressure volume (PV) curve of the respiratory system. METHODS Animal Preparation and Measurements We studied 18 supine, adult, mixed-breed dogs of either sex, weighing kg (mean SD), according to a protocol approved by the

2 1486 AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE VOL Animal Care and Use Committee of the University of Minnesota. Each dog was anesthetized with 30 mg/kg sodium pentobarbital and orally intubated with an 8-mm I.D. Hi-Lo Jet endotracheal tube (Mallinckrodt, Argyle, NY). Mechanical ventilation was initiated in the constant-flow, volume-cycled mode (70a; Puritan Bennett, Carlsbad, CA), with a VT of ml/kg, a frequency of 12 breaths/min, a PEEP of 3 cm H 2 O, an inspiratory time fraction (TI/Ttot, where TI inspiratory time and Ttot total respiratory cycle time) of 0.33, and an inspired oxygen fraction (FI O2 ) of The dogs were immobilized by intravenous injection of pancuronium (0.1 mg/kg). Muscle relaxation and anesthesia were maintained by continuous infusion of pentobarbital (2 to 4 mg/kg/h) and pancuronium (0.1 mg/kg/h) throughout the experiment. Femoral artery and vein catheters were inserted via an incision, and were sutured in place. Systemic mean arterial pressure (MAP) and heart rate (HR) were recorded from the femoral artery catheter. Arterial ph, PCO 2, and PO 2 were measured continuously with an intravascular blood gas monitoring system (Paratrend 7; Biomedical Sensors, Malvern, PA) inserted via the femoral artery catheter. A pulmonary artery (PA) catheter was introduced via the right external jugular vein, with its tip positioned in the pulmonary artery for measurement of PA and balloon-occlusion (Ppao) pressures. Pressure tracings were used to verify the positions of the proximal and distal ports of the PA catheter, which were in the right atrium and PA, respectively. Cardiac output ( Q. ) was measured with bolus injections of cold 5% dextrose delivered during expiration (0 C; Model 95A; Edwards, Irvine, CA). All intravascular pressures were measured with low-displacement transducers (Model PX600F; Baxter, Irvine, CA) referenced to the midchest level. Arterial and mixed venous blood gases were analyzed at 37 C (Corning 178 blood gas analyzer; Corning Glass, Medfield, MA) and were corrected to core temperature as measured with the thermistor at the distal end of the PA catheter. Airway opening pressure (Pao) was measured from a lateral pressure tap attached to the proximal end of the endotracheal tube (ETT), using a differential pressure transducer (MP cm H 2 O; Validyne Corp., Northridge, CA). Total PEEP (set PEEP plus auto-peep) was measured by occlusion of the airway at end-expiration, using a Braschi value, as described previously (26). We estimated pleural pressure by introducing a balloon catheter to measure the esophageal pressure (Pes) (MP cm H 2 O; Validyne Corp.). The position of the esophageal balloon was verified by the airway occlusion method before muscle relaxation (27). A pneumotachograph (No. 3719; Hans Rudolph, Kansas City, MO) was positioned in the common ventilator circuit at the airway opening, to measure inspiratory and expiratory flows. Expired CO 2 was measured continuously with an on-line infrared capnograph (Capnogard; Novametrics, Wallingford, CT). Lung volumes were measured continuously with dogs in the supine position, using a volume-displacement plethysmograph equipped with a Krogh spirometer (J. H. Emerson, Cambridge, MA) and a linear transducer (Schaevitz Engineering, Pennsauken, NJ). FRC measurements were made with the helium dilution technique at ZEEP (28). PV curves were obtained before (preinjury) and after induction and stabilization of lung injury (postinjury), using the supersyringe method (29). For each PV curve, stepwise volume changes of 50 ml were made over the inspiratory and expiratory capacity range (static airway pressure: 0 to 40 cm H 2 O). Static Pao and Pes were recorded after a 3-s pause, and were used to calculate transpulmonary pressure (PL Pao Pes). Lung volume history was standardized prior to determinations of FRC and recording of PV curves, by applying a CPAP of 30 cm H 2 O at the airway opening for s, followed by s of apnea at ZEEP. Models of Lung Injury After recording baseline hemodynamic, gas exchange, and mechanics measurements, we induced lung injury according to one of the models described subsequently. Six dogs were randomly assigned to each model of ALI. In all models, fluid management was standardized, using a continuous infusion of 0.9% NaCl at a rate of 100 ml/h, with the following exceptions: (1) if MAP was 60 mm Hg or Ppao was 4 mm Hg, an additional 100 ml of 0.9% NaCl was infused over 6 min; and (2) if Ppao was 8 mm Hg, the continuous infusion was stopped until Ppao returned to within the goal range of 4 to 8 mm Hg. LAV. With the dog in the supine position, the ETT was disconnected from the ventilator and warmed sterile saline (1,500 ml) was instilled, us- ing gravity, via the ETT until a fluid meniscus level was seen in the ETT. After 1 min of apnea, the fluid was gravity-drained and the recovered volume was measured. Ventilation (VT ml/kg, PEEP 3 cm H 2 O) was resumed for 10 min after each lavage while Pa O2 was continuously followed to assess lung injury. The LAV process was repeated until adequate injury was evident (defined as Pa O2 80 mm Hg after 10 min of ventilation following lavage) or until eight successive lavages had been performed. Subsequently, ventilation was continued with a VT of 15 ml/ kg and a PEEP of 3 cm H 2 O for a stabilization period of 90 min before the experimental protocol was begun. This period was chosen on the basis of published data (30) as well as experience in our laboratory. OAI. With the dog in the supine position, oleic acid (0.09 ml/kg) was injected into the right atrium via the proximal port of the PA catheter. This was followed by a 90-min period of initiation of injury and of stabilization at baseline ventilation (PEEP 3 cm H 2 O, VT 15 ml/kg) before the experimental protocol was begun. Previous work has shown that OAI pulmonary edema develops and stabilizes by 60 to 90 min after administration of oleic acid (31). Intratracheal instillation of Escherichia coli. Escherichia coli (collection strain 25922; American Type Culture Collection [ATCC], Rockville, MD) was prepared as a solution with a concentration of 10 8 cfu/ml as previously described (32). Quantitative cultures were done of a portion of the solution to determine the number of viable bacteria given to each dog. Subsequently, 1 ml of the solution was suspended in 49 ml of 0.9% NaCl and, with PEEP increased to 10 cm H 2 O, was introduced above the carina via the insufflation lumen of the ETT. The bacterial solution was divided and instilled in four equal doses, rotating the animal at random through the left lateral, right lateral, prone, and supine positions. The animal was held for 5 min in each position after bacterial instillation, and was returned to the supine position after completing the bacterial administration. PEEP was subsequently returned to 3 cm H 2 O and the animal was ventilated at baseline settings for a period of 6 h before the experimental protocol was begun. The 6-h initiation and stabilization period was determined from previous experience in our laboratory (32). Experimental Protocol The experimental protocol used in the study is outlined as a timeline in Figure 1. Before initiation of injury in each model, measurements were made of hemodynamics, gas exchange, and lung mechanics (including a PV curve and FRC at ZEEP). Subsequently, we injured the lungs and allowed the injury to stabilize, as defined by a 5% change in the measured Pa O2 over a 15-min period. Once the injury was stable, we obtained a set of measurements (at a PEEP of 3 cm H 2 O), and then increased PEEP to 10 cm H 2 O while keeping VT constant at 15 ml/kg. After 15 min at these settings, we recorded arterial and mixed venous blood gas tensions and conducted an RM. The time point immediately before an RM was defined as t 3 min. Each RM consisted of three successive SI, as follows: (1) SI-1: CPAP at 40 cm H 2 O for 30 s, followed by 30 s of ventilation at pre-rm settings; (2) SI-2: CPAP at 60 cm H 2 O for 30 s, followed by 30 s of ventilation at pre- RM settings; and (3) SI-3; CPAP at 60 cm H 2 O for 30 s, followed by return to pre-rm ventilation. The time-point at 30 s after SI-3 was defined as t 0 min. Arterial and mixed venous blood gases were measured at times t 3, 0, 2, 4, 6, 10, and 15 min. Lung volume measurements were recorded continuously throughout the protocol. VT was then increased to ml/kg (with PEEP 10 cm H 2 O), and the animal was ventilated for 15 min before undergoing another measurement phase followed by an RM. PEEP was then increased to cm H 2 O (with VT ml/kg), and after 15 min, a measurement phase and an RM were repeated. Our choice of VT (15 and ml/kg) was based on our previous experience and on published data indicating that a relatively large VT was required to avoid excessively high levels of Pa CO2 in dogs with ALI (32). In order to assess EELV above FRC, we imposed apneic pauses at ZEEP ( 30 s) at each ventilator setting at three time points: (1) initial : at 60 s after PEEP and VT were set; (2) pre-rm : immediately prior to before an RM (i.e., t 3 min); and (3) post-rm : at 15 min after an RM. At each time point, EELV above FRC was determined by the difference between plethysmographic EELV at PEEP and plethysmographic EELV during apnea at ZEEP. The change in

3 Van der Kloot, Blanch, Youngblood, et al.: Recruitment Maneuvers in Models of ALI 1487 Figure 1. Schematic diagram of experimental protocol used in study. The axis is relative and not to scale. Each stage consisted of hemodynamic and gas-exchange measurements followed by an RM, with arterial and mixed-venous blood gas measurements at 0, 2, 4, 6, 10, and 15 min after the RM. Gravimetric indices of pulmonary edema were obtained at the end of the experimental protocol. PEEP positive end-expiratory pressure. VT tidal volume. Open ovals represent changes in ventilatory strategy. EELV above FRC between the initial and pre-rm points may be considered cumulative tidal recruitment (or derecruitment ), and the change in volume between the pre- and post-rm points may be considered RM recruitment (or derecruitment ), provided that FRC at ZEEP at each time point remains the same. After completion of the protocol, the animal was given intravenous bolus injects of pentobarbital (10 mg/kg) and heparin (5,000 U intravenously) and rapidly exsanguinated via an external jugular venous catheter. The lungs were removed and weighed to obtain their wet weight (WW), and were then desiccated in a vacuum oven at 55 C until a constant dry weight (DW) was obtained (after 4 to 5 d). We report the normalized data as wet weight-to-body weight (WW/BW) and WW/DW ratios. Data Acquisition and Analysis Flow rates, Pao, Pes, expired CO 2, and volume displacement-based lung volume were simultaneously recorded on a chart recorder (Model 95000; Astro-Med, West Warwick, RI) and stored on digital tape (RD-111T; TEAC, Tokyo, Japan). Static respiratory system compliance (Crs), venous admixture ( Q. V/ Q. T), pulmonary vascular resistance (PVR), and systemic vascular resistance (Rsv) were calculated as described previously (32). Chest wall compliance (Cw) was calculated by dividing VT by the difference between end-inspiratory Pes and end-expiratory Pes. Lung compliance (C) was calculated with the formula C (Cw Crs)/(Cw Crs). Three randomly chosen, consecutive breaths were used for data analysis in each experimental stage. Plip and the upper deflection point of the respiratory PV curve (Pudp) were determined visually as described previously (29). Data analysis was done with a data acquisition and analysis software system (LabVIEW; National Instruments, Austin, TX). We examined the correlation between changes in EELV and arterial oxygenation in two ways. The change in EELV above ZEEP-FRC (obtained by measuring volume changes during declines from specific PEEP levels to ZEEP) was compared with changes in Pa O2 before and TABLE 1 HEMODYNAMIC PARAMETERS BEFORE AND AFTER INJURY AT EACH VENTILATOR SETTING Before Injury After Injury PEEP (cm H 2 O) VT (ml/kg) MAP, mm Hg LAV OAI PNM CO, L/min LAV OAI * PNM * Rpv, dyne/s cm 5 LAV OAI PNM Rsv, dyne/s cm 5 LAV 1, , , , , * OAI 2,218 1,382 1, , , , * PNM 1, , , , ,132 1,239 Definition of abbreviations: ANOVA analysis of variance; CO cardiac output; LAV saline lavage; MAP mean aterial blood pressure; OAI oleic acid injury; PEEP positive end-expiratory pressure; PNM pneumonia; Rpv pulmonary vascular resistance; Rsv systemic vascular resistance; VT tidal volume. Values are mean SD. * p versus postinjury with VT 15 ml/kg. p versus PEEP 10 cm H 2 O and 3 cm H 2 O. p 0.05 versus OAI and PNM. p 0.05 versus OAI and LAV.

4 1488 AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE VOL min after an RM. Second, immediately after an RM, changes in lung volume as measured by body plethysmography were compared with changes in Pa O2 at specified time points (as described earlier). Statistical Analysis All values are reported as mean SD unless otherwise specified. Separate one-way analysis of variance (ANOVA) procedures were used to test for differences in pulmonary mechanics and hemodynamic variables across the three lung injury models at equivalent ventilatory settings. Because we compared the data from three lung injury models for 14 variables, statistical tests with borderline significance levels should be interpreted with caution. A repeated-measures ANOVA was used to determine the effect of the four postinjury ventilatory strategies (PEEP 3 cm H 2 O/VT 15 ml/kg; PEEP 10 cm H 2 O/VT 15 ml/kg; PEEP 10 cm H 2 O/VT ml/kg; PEEP cm H 2 O/VT ml/kg) within a single category of lung injury. Duncan s post hoc analysis (pairwise comparisons were considered significant if p 0.05) was performed if the global F test indicated statistical significance (p 0.05). The changes in Pa O2 and Q. V/ Q. T over time after an RM (at a chosen PEEP/VT setting and in a specific lung injury model) were analyzed through repeated-measures ANOVA and Tukey s post hoc analysis to compare the values immediately before the RM (t 3 min) with those at each of the subsequent seven time points. A similar analysis was performed for each injury-peep/vt category, to compare the volume above FRC at three time points: initial (60 s after setting PEEP), immediately before an RM, and 15 min after the RM. The change in ZEEP-FRC after each method of lung injury, as compared with preinjury values, was tested with a paired t test. RESULTS For LAV, the volume of saline instilled was 9,358 1,990 ml. The volume of saline returned was 9,037 1,938 ml, with a positive balance of ml. The total number of viable E. coli actually administered to the PNM animals was cfu. All of the animals survived the experimental protocol. Hemodynamics Hemodynamic data before and after injury at various ventilator settings are summarized in Table 1. There were no significant differences in hemodynamic parameters among the models, except for an increased pulmonary vascular resistance (Rpv) in the PNM as compared with the LAV and OAI models at both VT settings with PEEP 10 cm H 2 O. MAP trended downward with increasing PEEP in all models. Cardiac output (CO) decreased significantly with increased VT, from 15 to ml/kg, in the OAI and PNM groups (p ). CO was lower in all models at a PEEP of cm H 2 O than at lower PEEP levels, and a PEEP of cm H 2 O was associated with significantly higher Rpv than were lower PEEP levels in the LAV and PNM models (p 0.007). Lung Mechanics In all animals, no Plip was observed in the preinjury PV curve. After injury, the mean Pao values corresponding to Plip were , , and cm H 2 O for the LAV, OAI, and PNM models, respectively, whereas the mean Pao values corresponding to Pudp were , , and cm H 2 O, respectively, for the three models. There were no statistically significant differences in either postinjury Plip or postinjury Pudp among the models. Table 2 summarizes the lung mechanics prior to injury and at each ventilator setting (before an RM) after injury. In all models, Crs decreased significantly after lung injury (p ). In all models, Crs decreased further when PEEP was increased to cm H 2 O, but this reached statistical significance only in the LAV and PNM models (p 0.007). With PEEP 10 cm H 2 O and VT ml/kg, Crs was significantly higher in the LAV than in the other models (p 0.05). Cw did not change significantly either with injury or with changes in PEEP and VT. Conse- TABLE 2 LUNG MECHANICS BEFORE AND AFTER INJURY AT EACH VENTILATOR SETTING Before Injury After Injury PEEP (cm H 2 O) VT (ml/kg) Peak Pao, cm H 2 O LAV * 45 5 OAI PNM * 48 6 Plateau Pao, cm H 2 O LAV * 41 4 OAI PNM Mean Pao, cm H 2 O LAV OAI PNM Crs, ml/cm H 2 O LAV OAI PNM Cw, ml/cm H 2 O LAV OAI PNM C, ml/cm H 2 O LAV OAI PNM Definition of abbreviations: ANOVA analysis of variance; C lung compliance; Crs respiratory system compliance; Cw chest wall compliance; LAV saline lavage; OAI oleic acid injury; Pao airway opening pressure; PEEP positive end-expiratory pressure; PNM pneumonia; VT tidal volume. Values are mean SD. * p versus postinjury with VT 15 ml/kg. p versus PEEP 10 cm H 2 O and 3 cm H 2 O. p 0.05 versus OAI and PNM.

5 Van der Kloot, Blanch, Youngblood, et al.: Recruitment Maneuvers in Models of ALI 1489 quently, most of the observed changes in Crs could be attributed to changes in C. Gas Exchange Comparison of different ALI models at baseline. Pre- and postinjury Pa O2 and Q V/ Q T at PEEP 3 cm H 2 O are shown in Figures 2A and 2B. There were no significant differences in Pa O2 or Q V/ Q T among the models at PEEP 3 cm H 2 O, either before or after injury. In all models, Pa O2 decreased significantly and Q V/ Q T increased significantly after lung injury (p ). Effect of PEEP and VT on gas exchange after injury. In LAV and OAI animals, Pa O2 (measured immediately before the RM in each ventilatory strategy) improved significantly as PEEP was increased (p ; Figure 2A). The trend toward improved oxygenation with increasing PEEP in PNM animals was not statistically significant. Only in LAV animals did Pa O2 improve significantly when VT was increased from 15 to ml/kg at PEEP 10 cm H 2 O (p ). With a VT of both 15 ml/kg and of mg/kg, increasing PEEP from 3 to 10 cm H 2 O significantly reduced Q V/ Q T (Figure 2B) in OAI animals (p ), whereas this reduction reached statistical significance in LAV animals only at the higher VT ( ml/kg; p ). Q V/ Q T decreased further at PEEP cm H 2 O in all models, although this change did not reach significance in the PNM model. At PEEP cm H 2 O, LAV animals had a significantly higher Pa O2 and lower Q V/ Q T than did PNM animals (p 0.04). In all three models, a PEEP of cm H 2 O normalized Q V/ Q T to values similar to those observed at a preinjury PEEP of 3 cm H 2 O. At PEEP 10 cm H 2 O and VT 15 ml/kg, Pa CO2 (measured immediately before the RM in each ventilatory strategy) was 69 mm Hg, and decreased significantly (p 0.01) to mm Hg when the VT was increased to ml/kg. Increasing PEEP to cm H 2 O increased Pa CO2 significantly (p 0.01), to mm Hg. There were no statistically significantly differences in Pa CO2 among the models at each ventilator setting. Effect of RM on gas exchange. The oxygenation responses of the models to an RM are depicted in Figures 3A and 3B. With PEEP 10 cm H 2 O, there was a transient increase in Pa O2 at both values of VT, with a corresponding decrease in Q V/ Q T after an RM in all three models. The improvement in oxygenation was sustained for 15 min after an RM only in the LAV model with PEEP 10 cm H 2 O and VT 15 ml/kg (p ). At PEEP cm H 2 O, an RM changed neither Pa O2 nor Q V/ Q T significantly at any time point in any of the Figure 2. (A) Pre- and postinjury Pa O2 values ( SEM) before RM in different models of ALI. LAV (solid bar), OAI (dotted bar), PNM (cross hatched bar), VT, and PEEP used at each stage were as specified (*p 0.05 versus OAI and PNM at same settings; p 0.05 versus PNM at same settings; p 0.05 versus preinjury; p 0.05 versus PEEP 10 cm H 2 O, VT 15 ml/kg; p 0.05 versus PEEP 10 cm H 2 O, VT ml/kg). (B) Pre- and postinjury venous admixture ( Q V/ Q T) values ( SEM) before RM in different models of ALI. LAV (solid bar), OAI (dotted bar), PNM (crosshatched bar), VT, and PEEP used at each stage were as specified (*p 0.05 versus preinjury in same model; p versus PEEP 3 cm H 2 O, VT 15 ml/kg in same model; p versus all other postinjury settings in same model; p 0.04 versus LAV and OAI at same setting).

6 1490 AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE VOL Figure 3. Pa O2 (A) and Q V/ Q T (B) values before and after an RM in each model of ALI. LAV saline lavage, OAI oleic acid injury, PNM pneumonia, PEEP positive end-expiratory pressure, VT tidal volume. The RM was completed 30 s before time 0 min. Bars represent SEM (*p 0.05 versus t 3 min). models. RM were associated with transient decreases in MAP, limited to the duration of administration of CPAP. Lung Volume Effect of ALI on FRC. In all animals, the postinjury PV curve was shifted down and to the right (i.e., decreased FRC and compliance) as compared with the preinjury PV curve. ZEEP- FRC did not differ among the models either before (LAV: ml; OAI: ml; PNM: ml) or after injury (LAV: ml; OAI: ml; PNM: ml). In each model, ZEEP-FRC decreased significantly with lung injury (p 0.002). Effect of sequential SI on lung volume. Figure 4A shows the change in EELV with each SI for each ventilatory strategy. The recruited volume for each SI was determined as the change in the plethysmographic EELV from immediately before application of SI to 30 s after establishing tidal ventilation with PEEP. At PEEP 10 cm H 2 O and a VT of either 15 ml/ kg or ml/kg, there was a progressive decline in recruited volume with successive SI; the first SI produced the largest increment in lung volume. The lower VT strategy produced the greatest total recruited volume after an RM. At PEEP cm H 2 O, SI-2 recruited more volume than did either SI-1 or SI-3, but these differences did not reach statistical significance. Figure 4B shows the recruited volume with each SI in each model (all ventilatory patterns). In each model, recruited volume tended to decrease with successive SI. In all models, SI-3 recruited less volume than did either SI-1 or SI-2, with the differences reaching statistical significance in the OAI and PNM models (p 0.001). Effect of VT and RM on lung volume. Figure 5 depicts EELV above ZEEP-FRC at the initiation of each ventilator strategy before 15 min of ventilation), immediately (before an RM, and again, 15 min after an RM in each model and at each ventilator setting. In the LAV model with PEEP 10 cm H 2 O at tidal volumes of both 15 and ml/kg, there was a significant increase in EELV over 15 min of ventilation before an RM (p 0.002). The increase in EELV seen in the OAI and PNM models at these settings did not reach statistical significance. At PEEP 10 cm H 2 O, the increase in EELV at 15 min after an RM was significant in the LAV and PNM models (at both tidal volumes), but was not significant in the OAI model. At PEEP cm H 2 O there was no change in EELV with ventilation before or after an RM. Relationship between lung volume and oxygenation. The relationship between changes in total lung volume with respect to FRC at PEEP (as measured with body plethysmography) and changes in arterial oxygenation was examined for the 15- min period following an RM for each ventilatory strategy (time points: 3, 0, 2, 4, 6, 10, and 15 min). In the LAV and PNM models there were no correlations between changes in EELV and Pa O2 with any ventilatory strategy, whereas in the OAI model at PEEP 10 cm H 2 O (at both tidal volumes), there was a significant positive linear correlation (p 0.001). In the OAI model at PEEP cm H 2 O, no correlation was evident in at either value of VT. Gravimetric Indices of Lung Injury There were no significant differences in WW/DW (LAV: ; OAI: ; PNM: ) or WW/BW (LAV: ; OAI: ; PNM: ) among the three models.

7 Van der Kloot, Blanch, Youngblood, et al.: Recruitment Maneuvers in Models of ALI 1491 Figure 4. (A) Change in lung volume after each SI in all ALI models. PEEP positive end-expiratory pressure; VT tidal volume; PEEP 10 cm H 2 O, VT 15 ml/kg (dotted bars); PEEP 10 cm H 2 O, VT ml/kg (gray bars); PEEP cm H 2 O, VT ml/kg (striped bars). Bars represent SEM (*p 0.04, S1 versus S2; p 0.04 S2 versus S3; p 0.04 S1 versus S3). (B) Change in lung volume after each SI at all ventilatory settings. PEEP positive end-expiratory pressure, VT tidal volume. LAV (solid bars); OAI (dotted bars); PNM (crosshatched bars) ( p 0.001, SI 3 versus SI 1, and SI 3 versus SI 2). DISCUSSION Comparison of Models of ALI All three experimental models of ALI used in this study were associated with a significant reduction in FRC and Crs (related to a decreased in C) and a deterioration in gas exchange, without any significant differences among the models in terms of gas exchange (at a PEEP and VT of 3 cm H 2 O and 15 ml/kg, respectively), lung mechanics, hemodynamic parameters, or gravimetric indices of lung injury. By these criteria, the severity of lung injury did not differ among the LAV, OAI, and PNM models. There were, however, differences among the models in several respects. Oxygenation in the PNM model was less responsive to increasing levels of PEEP than in the LAV and OAI models. Presumably, PNM caused a greater degree of inflammatory air space consolidation, as opposed to the prominent pulmonary edema and atelectasis present in the other models, thus increasing the pressure required to open collapsed alveolar units. However, the similar values of Plip observed on the PV curves of each model suggests that initial alveolar recruitment begins at similar levels of PEEP in each model (33). Our data also indicate that when injured lungs are ventilated with combinations of relatively low PEEP and VT, either an RM or a larger VT may recruit lung volume. However, the response in terms of oxygenation may not parallel the changes in lung volume. LAV animals showed a significant improvement in gas exchange and EELV when VT was increased from 15 to ml/kg at PEEP 10 cm H 2 O. This finding supports previously reported data showing improved gas exchange and evidence for alveolar recruitment with increased VT in earlystage ARDS (34). Interestingly, a PEEP of 10 cm H 2 O and VT of 15 ml/kg in LAV animals was the only situation in which there was a sustained (at 15 min) improvement in oxygenation after an RM. These findings imply that the LAV, as opposed to the OAI and PNM models, may involve collapsed airways that can be recruited and maintained by tidal ventilation. An RM may accelerate this process by opening and stabilizing additional lung units. The surfactant-depletion model may behave differently than models primarily involving pulmonary edema or air space inflammation with secondary surfactant dysfunction. In fact, Mink and colleagues reported that at Day 3 in a canine model of lobar pneumococcal pneumonia, the area of consolidation did not inflate when the lungs were inflated to TLC (35). In previous work with LAV and OAI in rabbits, a single SI was shown to improve Pa O2 and lung volume during HFV but not during CMV with a high VT ( 15 ml/kg) (18, 21). These observations emphasize the importance of VT recruitment during CMV (33). Under certain conditions, ventilating with a large VT may eliminate the need for an RM. Composition of an RM The literature has a paucity of data on various methods of performing an RM in animal or human subjects. Most investiga- Figure 5. Effects of cumulative tidal recruitment and RM on lung volume above FRC at ZEEP (ZEEP- FRC) in three; LAV, OAI, and PNM models of ALI. PEEP positive end-expiratory pressure, VT tidal volume (*p versus initial value; p versus before RM).

8 1492 AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE VOL tors have utilized a single SI consisting of a CPAP of 30 to 40 cm H 2 O applied for 15 s (17, 21). Byford and colleagues (36) showed that small fluctuations in VT superimposed on CPAP (i.e., during HFV) were more effective in recruiting lung volume than was CPAP alone in lung-lavaged rabbits. Clearly, neither the magnitude nor the duration of the CPAP during an SI that achieves maximal recruitment are known, and both most likely depend on the nature and severity of the underlying lung injury as well as on the mechanics of the chest wall. Conceivably, a lower pressure applied for a longer period and superimposed on excursions in VT may be superior to high pressure applied for a short period in terms of lung volume recruitment while avoiding development of VILI. We chose to perform three sequential SI with different CPAP levels, in part to examine the effect of different pressures and repetitive applications of pressure on lung volume and gas exchange. Neither animal nor human data are available to guide the choice of a safe level and duration of CPAP during an SI. For that matter, the safety of applying sequential SI cannot be determined from our data. Our data, derived from three models of ALI, suggest that an SI with 40 cm H 2 O applied for 30 s is insufficient to maximally recruit lung volume. Increasing the CPAP level to 60 cm H 2 O during the second SI consistently recruited further lung volume (Figure 4B). In all models of ALI, the third SI failed to raise EELV significantly, and did not increase Pa O2 (as monitored with continuous Pa O2 measurement) beyond that achieved by the second SI. Although we did not specifically examine the question, it is probable that a single SI at a higher pressure ( 60 cm H 2 O) would have been sufficient to inflate the majority of recruitable lung units. We did not experience any adverse sequelae from application of an RM, but given the potential for hemodynamic compromise as well as barotrauma related to SI, our data would suggest that repeated application of SI may involve more potential risk than benefit. Under certain conditions, repeated SI may also perpetuate VILI and worsen gas exchange, since in some dogs (in all models) we observed a decrement in both Pa O2 and lung volume after the completion of an RM. Effect of an RM at Higher PEEP The lack of response of an RM, in terms of both gas exchange and lung volume, at the higher PEEP level used in our study ( cm H 2 O) indicates that alveolar recruitment was maximized with PEEP and VT alone, and that little benefit could be obtained with an RM. This suggestion is supported by the lack of significant differences in gas exchange parameters from before to after injury with a PEEP of cm H 2 O in any of the models we studied. The question thereby raised is what role an RM may have in the open lung approach, in which PEEP and VT are set at or above Plip in an attempt to maximally recruit the lung and maintain alveolar patency. Our data, although suggesting that an RM may be superfluous when ventilating at high PEEP levels (above Plip), cannot be generalized to the open lung approach, since the tidal volumes used in our study were higher than those advocated in a pressure-limited, lung-protective ventilatory strategy (13). It is possible that with a lower VT, an RM may prove beneficial even when applied atop high levels of PEEP. With longer periods of ventilation, progressive derecruitment may occur (particularly at low VT) even at PEEP levels above Plip, with reemergence of the oxygenation and/or lung volume response to an RM. Nevertheless, it is also plausible that when a VT and PEEP combination maintains the patency of all recruitable lung units, an RM may induce and/or perpetuate lung injury through alveolar overdistention. Relationship between Lung Volume and Gas Exchange We observed a significant correlation between changes in lung volume and oxygenation only in the OAI model at a PEEP of 10 cm H 2 O and at both tidal volumes tested. Interestingly, under other conditions, even when an RM or tidal ventilation increased EELV, oxygenation failed to improve in proportion to changes in lung volume. These findings have several possible explanations. It is possible that an RM adversely affects ZEEP-FRC, and that consequently, changes in EELV above FRC do not represent changes in total lung volume. To avoid the possibility of the FRC measurement influencing the stability of the experimental model, we did not repeat FRC measurements before and after each RM. An increase in the observed lung volume above FRC without a corresponding improvement in oxygenation may therefore have been caused by a decrease in FRC after an RM, without any change in total lung volume. This explanation is unlikely, however, since it presumes that the RM simultaneously decreased FRC and increased lung compliance (i.e., the EELV inflated by the level of PEEP used). Moreover, collapse of lung units in similar models of lung injury (OAI, LAV, and endotoxin infusion) in pigs has been shown to occur within 4 s (37). Consequently, a - to 30-s period of apnea should standardize the volume history of the lungs and allow the lungs to reach similar volumes at ZEEP, irrespective of prior interventions (i.e., an RM). In other words, ZEEP-FRC should remain essentially constant, provided the lung injury model remains stable over time. Another explanation for the lack of correlation between changes in lung volume and oxygenation is that lung volume increased preferentially through the expansion of compliant alveolar units rather than from the recruitment of collapsed units, with diversion of blood flow to diseased units. In such a case, lung volume may increase with no improvement in or even a worsening of oxygenation parameters. Our choice of FI O2 (0.80) may also have influenced our results. In normal humans undergoing general anesthesia, Rothen, and associates (16) demonstrated that an RM decreased intrapulmonary shunt and increased areas of low ventilation/perfusion ( V / Q ), which resulted in a small improvement in arterial oxygenation. Consequently, if we had performed our study with 100% oxygen instead of 80% oxygen, we might have observed a greater improvement in oxygenation with an RM. With an FI O2 of 0.80, an RM may increase lung volume by opening collapsed units and thus reducing shunt while increasing areas of low V / Q, thereby resulting in only a minor improvement in oxygenation. This may explain in part the lack of correlation between changes in lung volume and Pa O2 in our study. The extent of this process may differ among the models we used, explaining the lack of improvement in oxygenation in response to an RM in the PNM model despite the observed increase in lung volume. In the OAI model, transient improvements in EELV correlated with increases in Pa O2 after an RM, suggesting that in this model, increasing EELV improved V / Q relationships. However, this response was short-lived, with a rapid simultaneous loss of both lung volume and Pa O2. At higher PEEP levels in the OAI and possibly also in the PNM and LAV models, V / Q relationships were nearly normalized and were not affected by an RM. Another consideration is that the frequency response of the continuous Pa O2 measuring device used in our study may have influenced our results. This possibility seems unlikely, since the 90% response time for this intravascular blood gas monitoring system is reported to be 70 s (38). Moreover, we frequently collected arterial blood samples and compared the analyzed results with those recorded by the intravascular monitor at the time of collection, with similar results.

9 Van der Kloot, Blanch, Youngblood, et al.: Recruitment Maneuvers in Models of ALI 1493 Clinical Application of RM Our data were obtained in animal models of ALI, and should therefore not be directly extrapolated to the clinical setting. However, certain insights into clinical practice can be obtained from our findings. Our results suggest that when a low PEEP/low VT ventilatory pattern was used, the LAV model was more responsive to an RM than were the other two models of ALI. The prominent airway collapse caused by surfactant depletion (30) in the LAV model is likely to be more characteristic of neonatal respiratory distress syndrome as well as early ALI in adults, before the development of parenchymal remodeling and/or fibrosis. In addition, recent human data (24) suggest that the etiology of ARDS may influence the response to mechanical ventilation, with ARDS resulting from an extrapulmonary source being more responsive to PEEP. It is therefore likely that an RM will be less effective in pulmonary ARDS, as well as in late ARDS (i.e., in patients in whom the lung injury causing ARDS is associated with more consolidation or fibrosis). In such cases, an RM may in fact be detrimental, leading to stretch injury of more compliant lung units, with redirection of blood flow to injured areas, as well as increasing the risk of barotrauma. Our data also suggest that an RM may be of no benefit if the lung has already been optimally or nearly-optimally recruited by PEEP and/or VT. However, we did not specifically study the effect of an RM when ventilating with PEEP above Plip at a lower VT. It is possible that a decrease in tidal recruitment when ventilating with such settings will lead to progressive derecruitment despite the level of PEEP, and will thereby make an RM beneficial. We have also not addressed the role of RM as an adjunct to other forms of mechanical ventilation, such as HFV. Our data suggest that the response to an RM found previously in surfactant-depletion models with HFV (18 21) may not be the same in OAI or PNM models of ALI. Moreover, opening collapsed lung units with an RM without providing sufficient PEEP to maintain their patency during ventilation may expose those units to cyclic shear forces, predisposing them to VILI. In summary, our findings show that each of the models of ALI examined in our study behaves differently in response to an RM superimposed on a chosen mechanical ventilatory strategy. This finding has important implications in the interpretation of animal studies in ALI. RM appears to have a beneficial role in the setting of surfactant depletion when ventilating with PEEP below Plip and at a relatively low VT. When using a higher VT or PEEP above Plip, and in the OAI and PNM models, we did not observe any benefit from an RM. When an RM is used, it would appear that a brief (30 s), single application of a high Pao may be optimal and well tolerated. Our findings also show a lack of correlation between absolute lung volume and oxygenation after an RM, highlighting the heterogeneous distribution of injured areas that contribute to intrapulmonary shunt in some models of ALI. Acknowledgment : The authors wish to gratefully acknowledge Cynthia Gross, Ph.D., Associate Professor, University of Minnesota College of Pharmacy and School of Nursing, and Scott Bunner, Research Assistant, University of Minnesota Department of Internal Medicine, for their extensive review and revision of statistical analyses. References 1. Dreyfuss, D., P. Soler, G. Basset, and G. Saumon High inflation pressure pulmonary edema: respective effects of high airway pressure, high tidal volume, and positive end-expiratory pressure. Am. Rev. Respir. Dis. 137: Bowton, D. L., and D. L. Kong High tidal volume ventilation produces increased lung water in oleic acid-injured rabbit lungs. Crit. Care Med. 17: Webb, H. H., and D. F. Tierney Experimental pulmonary edema due to intermittent positive pressure ventilation with high inflation pressures: protection by positive end-expiratory pressure. Am. Rev. Respir. Dis. 110: Argiras, E. P., C. R. Blakely, M. S. Dunnill, S. Otremski, and M. K. Sykes High PEEP decreases hyaline membrane formation in surfactant deficient lungs. Br. J. Anaesth. 59: Sandhar, B. K., D. J. Niblett, E. P. Arigas, M. S. Dunnill, and M. K. Sykes Effects of positive end-expiratory pressure on hyaline membrane formation in a rabbit model of the neonatal respiratory distress syndrome. Intensive Care Med. 14: Parker, J. C., L. A. Hernandez, and K. J. Peevy Mechanisms of ventilatory-induced lung injury. Crit. Care Med. 21: Tremblay, L., F. Valenza, S. P. Ribeiro, J. Li, and A. S. Slutsky Injurious ventilatory strategies increase cytokines and c-fos m-rna expression in an isolated rat lung model. J. Clin. Invest. 99: Dreyfuss, D., and G. Saumon Ventilator-induced lung injury: lessons from experimental studies. Am. J. Respir. Crit. Care Med. 157: Snyder, J. V., and A. Froese The open lung approach: concept and application. In J. V. Snyder and M. R. Pinsky, editors. Oxygen Transport in the Critically Ill. Year Book Medical Publishers, Chicago Amato, M. B. P., C. S. V. Barbas, D. M. Medeiros, G. D. P. P. Schettino, G. L. Filho, R. A. Kairalla, D. Deheinzelin, C. Morais, E. D. O. Fernandes, T. Y. Takagaki, and C. R. R. D. Carvalho Beneficial effects of the open lung approach with low distending pressures in acute respiratory distress syndrome. Am. J. Respir. Crit. Care Med. 152: Stewart, T. E., M. O. Meade, D. J. Cook, J. T. Granton, R. V. Hodder, S. E. Lapinsky, C. D. Mazer, R. F. McLean, T. S. Rogevein, B. D. Schouten, R. J. Todd, and A. S. Slutsky Evaluation of a ventilation strategy to prevent barotrauma in patients at high risk for acute respiratory distress syndrome. N. Engl. J. Med. 338: Lachmann, B Open the lung and keep it open. Intensive Care Med. 18: Amato, M. B. P., C. S. V. Barbas, D. M. Medeiros, R. B. Magaldi, G. D. P. P. Schettino, G. L. Filho, R. A. Kairalla, D. Deheinzelin, C. Munoz, R. Oliveira, T. Y. Takagaki, and C. R. R. D. Carvalho Effect of a protective-ventilation strategy on mortality in the acute respiratory distress syndrome. N. Engl. J. Med. 338: Artigas, A., G. R. Bernard, J. Carlet, D. Dreyfuss, L. Gattinoni, L. Hudson, M. Lamy, J. J. Marini, M. A. Matthay, M. R. Pinsky, R. Spragg, P. M. Suter, and the Consensus Committee The American- European Consensus Conference on ARDS, Part 2. Am. J. Respir. Crit. Care Med. 157: Rothen, H. U., B. Sporre, G. Engberg, G. Wegenius, and G. Hedenstierna Reexpansion of atelectasis during anesthesia: a computed tomographic study. Br. J. Anaesth. 71: Rothen, H. U., B. Sporre, G. Engberg, G. Wegenius, and G. Hedenstierna Reexpansion of atelectasis during general anesthesia may have a prolonged effect. Acta Anaesthesiol. Scand. 39: Dorrington, K. L., and F. M. Radcliffe Effect of a single inflation of the lungs on oxygenation during total extracorporeal carbon dioxide removal in experimental respiratory distress syndrome. Intensive Care Med. 17: Kolton, M., C. B. Cattran, G. Kent, G. Volgyesi, A. B. Froese, and A. C. Bryan Oxygenation during high-frequency ventilation compared with conventional mechanical ventilation in two models of lung injury. Anesth. Analg. 61: Bond, D. M., and A. B. Froese Volume recruitment maneuvers are less deleterious than persistent low lung volumes in the atelectasisprone rabbit lung during high-frequency oscillation. Crit. Care Med. 21: Froese, A. B., P. R. McCulloch, M. Sugiura, S. Vaclavik, F. Possmayer, and F. Moller Optimizing alveolar expansion prolongs the effectiveness of exogenous surfactant therapy in the adult rabbit. Am. Rev. Respir. Dis. 148: Bond, D. M., J. McAloon, and A. B. Froese Sustained inflations improve respiratory compliance during high-frequency oscillatory ventilation but not during large tidal volume positive-pressure ventilation in rabbits. Crit. Care Med. 22: Novak, R. A., L. Shumaker, J. V. Snyder, and M. R. Pinsky Do periodic hyperinflations improve gas exchange in patients with hypoxemic respiratory failure? Crit. Care Med. 15: Ranieri, V. M., N. Brienza, S. Santostasi, F. Puntillo, L. Mascia, N. Vitale, R. Giuliana, V. Memeo, F. Bruno, T. Fiore, A. Brienza, and A. S.