Intensive Care Med (2002) 28:1151Ð1156 DOI 10.1007/s00134-002-1368-5 EXPERIMENTAL

Huibert R. van Genderingen Oxygenation index, an indicator Johannus A. van Vught Jos R. C. Jansen of optimal distending pressure during Elisabeth L. I. M. Duval Dick G. Markhorst high-frequency oscillatory ventilation? Adrian Versprille

D.G. Markhorst Received: 5 June 2001 tion) in series, followed by a similar Pediatric Intensive Care Unit, decrease of Paw (deflation). Accepted: 14 May 2002 Department of Pediatrics, Published online: 20 June 2002 Vrije Universiteit Medical Center, Measurements and results: At each © Springer-Verlag 2002 PO Box 7057, 1007 MB Amsterdam, Paw level, the OI and physiological The Netherlands shunt fraction were determined. The This study was supported by SensorMedics. A. Versprille OI reached a minimum of 6.2±1.4 at Pathophysiological Laboratory, Paw 30±4 cmH2O during inflation Department of Pulmonary Diseases, and a minimum of 2.4±0.3 at Paw Erasmus University Rotterdam, 13±2 cmH O during deflation. H.R. van Genderingen (✉) PO Box 1738, 3000 DR Rotterdam, 2 Department of Clinical The Netherlands Pawoptimal was 32±6 cmH2O on the Physics & Informatics, inflation limb and 14±2 cmH2O on Vrije Universiteit Medical Center, the deflation limb. The difference PO Box 7057,1007 MB Amsterdam, The Netherlands Abstract Objective: To test the hy- between the Paw at minimal OI and e-mail: [email protected] pothesis that, during high-frequency Pawoptimal was Ð1.9±4.2 cmH2O (NS) Tel.: +31-20-4444118 oscillatory ventilation (HFOV) of during inflation and Ð1.5±1.6 cmH2O Fax: +31-20-4444147 pigs with acute lung injury, the oxy- (p<0.05) during deflation. In 15 out of the 16 comparisons, the differ- J.A. van Vught á D.G. Markhorst genation index (OI = Paw*FIO2*100/ A. Versprille ence in Paw was within one step PaO2) is minimal at the lowest con- Pediatric Intensive Care Unit, tinuous distending pressure (Paw), (±3 cmH O). Conclusion: The mini- Department of Pediatrics, 2 Wilhelmina Children’s Hospital, where the physiological shunt frac- mal OI is indicative for the Paw where oxygenation is optimal during PO Box 80590, 3508 AB Utrecht, tion is below 0.1 (Pawoptimal). The Netherlands Design and setting: Prospective, ob- HFOV in surfactant-depleted pigs. J.R.C. Jansen servational study in a university re- Department of Intensive Care, search laboratory. Subjects: Eight Keywords Acute lung injury á Leiden University Medical Center, Yorkshire pigs weighing 12.0±0.5 kg, High-frequency oscillatory PO Box 9600, 2300 RC Leiden, with lung injury induced by lung la- ventilation á Oxygenation index á The Netherlands vage. Interventions: After initiation Gas exchange E.L.I.M. Duval of HFOV, the pigs were subjected to Pediatric Intensive Care Unit, Department of Pediatrics, a stepwise increase of Paw to obtain Paola Children’s Hospital, Lindendreef 1, under-inflation, optimal inflation and 2020 Antwerp, Belgium over-distension of the lungs (infla-

Introduction by a continuous distending pressure (Paw) and the super- imposed small oscillations of air flow provide gas ex- High-frequency oscillatory ventilation (HFOV) is a safe change. The level of Paw applied is a major determinant mode of respiratory support in the treatment of newborn of patient outcome [1, 2]. The optimal Paw is a compro- infants suffering from idiopathic respiratory distress syn- mise between maximal lung recruitment and minimal drome (IRDS) [1]. Atelectatic lung regions are opened over-distension. To determine lung recruitment, CT is 1152 the preferred diagnostic tool [3], but it is not available at mark). Inspiratory and mixed expiratory gases were analyzed the bedside. The gold standard for assessment of pulmo- by a mass spectrometer (Perkin-Elmer, MGA 1100, Pomona, Calif., USA). Q′S/Q′T and physiological dead space (VD/VT) in nary oxygen transfer is the physiological shunt fraction percentage were calculated according to standard formulas (Q′S/Q′T). Its determination, however, requires invasive [10]. The respiratory system compliance was derived according to access for mixed venous sampling not commonly used in the equation (Crs=VT/(Pplat-PEEP) where Pplat is the plateau infants. We investigated the use of the oxygenation index pressure. The oxygenation index (OI) was derived from (OI= Paw*FIO2*100/PaO2). Cardiac output (Q′T) was determined by (OI=Paw*FIO2*100/PaO2) to find the optimal Paw. The use of the thermodilution method. Four determinations were OI combines the goals to maximize arterial oxygenation equally spread over the ventilatory cycle and averaged [11]. Oxy- and minimize Paw and inspired oxygen concentration. gen delivery (DO2) was determined [10]. So far, use of the OI has been proposed to identify in- Next, FIO2 was set at 1.0 and PEEP was set at 2 cmH2O. The fants who require extra corporeal membrane oxygenation lungs were repeatedly lavaged with 35 ml/kg saline until PaO2 was less than 80 mmHg [12]. After lavage, the baseline measure- therapy (ECMO) [4, 5], to predict outcome of ECMO [6] ments were repeated (post-lavage measurements). Next, the ani- and to assess severity of illness in pediatric patients dur- mals were intubated with a 5.5 mm diameter endotracheal tube. ing HFOV [7]. The index has not been validated to opti- HFOV (Sensormedics 3100A, Yorba Linda, Calif.) was applied at mize ventilatory conditions. We hypothesized that the OI 8 Hz, 33% inspiratory time and bias flow 20 l/min. Paw was set 3 cmH2O higher than the during CMV. reaches a minimum at the Pawoptimal, defined as the low- Pressure amplitude was set (54±6 cmH2O) and maintained con- est Paw where Q′S/Q′T is below 0.1. This criterion is stant. After the experimental procedure described below, the tube adapted from Lachmann’s “open lung concept” [8], was removed, CMV was resumed and the baseline measurements where the optimal ventilatory condition during conven- repeated (end experiment measurements). tional (CMV) is defined as the lowest positive end-expiratory pressure (PEEP) to Experimental procedure achieve a fully recruited lung, i.e. a Q′S/Q′T below 0.1 [9]. In the present study, surfactant-deficient piglets were The continuous distending pressure was increased in steps of subjected to a stepwise increase of Paw to obtain under- 3 cmH2O, until PaO2 was more than 450 mmHg or the increase in PaO was less than 50% of the previous step. One more Paw step inflation, optimal inflation and over-distension of the 2 was added before Paw was set at 40 cmH2O (inflation phase). lungs in series, followed by a stepwise decrease of Paw. Then, it was lowered to the value at which the above-mentioned PaO2 criterion was reached. Subsequently, Paw was decreased in steps of 3 cmH2O until PaO2 was less than 80 mmHg (deflation phase). At each Paw level, the following were measured after Methods and materials a 10min stabilization period: blood gases, blood pressures and car- diac output. All experiments were performed according to the Guide for the Care and Use of Laboratory Animals, published by the National Institutes of Health (NIH publication 85Ð23, revised 1985) and ap- Data analysis proved by the Animal Care Committee of Erasmus University Rotterdam, The Netherlands. The data are presented as means ± SD. Data comparison was made using the Student’s t-test unless stated otherwise. Both during in- flation and deflation, Paw was determined at the minimal OI Surgical preparation (PawminOI) and the lowest Paw was determined where Q′S/Q′T was below 0.1 (Pawoptimal). PawminOI and Pawoptimal were compared us- Anesthesia of eight Yorkshire pigs (12.0±0.5 kg body weight) was ing the Wilcoxon signed rank test (p<0.05). induced with an intraperitoneal injection of sodium pentobarbital (30 mg/kg) and maintained by continuous infusion (8.5 mg/kg per h). After tracheostomy the pigs received a Y-shaped tracheal can- Results nula and were ventilated (CMV) with 10 breaths/min, an FIO2 of 0.21 and PEEP of 5 cmH2O. The inspiratory: expiratory ratio was 1:1, with an inspiratory pause of 0.6 s. Tidal volume Lavage caused a significant increase of Q′S/Q′T and (V =17.1±0.9 ml/kg) was adjusted to obtain normocapnia T VD/VT and decrease of Crs (Table 1), as can be expected (38Ð45 mmHg). Arterial blood pressure measurement and blood in IRDS. Comparison of the baseline measurements after sampling were performed using an arterial line in the aorta. A pul- monary arterial catheter was used to measure pulmonary arterial lavage and at the end of the experiment revealed only a pressure and blood temperature, and to sample mixed venous significant increase in PaCO2 (Table 1). Figure 1 shows a blood. Fluids (5 ml/kg per h saline) and anesthetics were adminis- typical example of the relationship between Paw and tered through a line in the superior vena cava. The bladder was PaO , Q′ /Q′ and OI. In all animals we found a similar catheterized to avoid urine retention. Spontaneous was 2 S T suppressed with pancuronium bromide (0.3 mg/kg per h). OI pattern: during inflation the OI decreased from 32.0±10.6 towards a minimum of 6.2±1.4 at PawminOI 30±4 cmH2O and subsequently increased to 7.6±1.0 at Measurements and lavage Paw 40 cmH2O; during deflation the OI decreased to a Prior to lavage, baseline measurements were performed (pre- minimum of 2.4±0.3 at PawminOI 13±2 cmH2O and then lavage measurements). Blood gas samples were analyzed with increased to 22.0±18.4 at Paw 9±2 cmH2O. Q′S/Q′T de- ABL3 and OSM2 oximeters (Radiometer, Copenhagen, Den- creased from 0.53±0.10 at Paw 17±2 cmH2O to 0.06±0.03 1153

Table 1 Settings and measurements during conventional mechani- cal ventilation before lavage, after lavage and at the end of the ex- periment. Data are presented as means (±SD). Values at end ex- periment were tested against post-lavage (FIO2 fractional inspired oxygen, PEEP positive end-expiratory pressure, PaO2 arterial par- tial pressure of oxygen, Q′S/Q′T physiological shunt fraction, PaCO2 arterial partial pressure of carbon dioxide, VD/VT, Crs res- piratory system compliance, Q′T cardiac output, DO2 oxygen de- livery)

Pre-lavage Post-lavage End experiment

FIO2 0.21 1.0 1.0 PEEP (cmH2O) 5 2 2 PaO2 (mmHg) 103 (10) 60 (11) 96 (61) Q′S/Q′T 0.01 (0.02) 0.46 (0.08) 0.39 0.14) PaCO2 (mmHg) 42 (2) 52 (5) 60 (8)* VD/VT (%) 40 (4) 61 (4) 64 (5) Crs (ml/cmH2O 1.20 (0.16) 0.54 (0.06) 0.55(0.07) per kg) Q′T (ml/s per kg) 1.9 (0.3) 1.9 (0.4) 1.9 (0.5) DO2 (ml/s per kg) 0.25 (0.04) 0.20 (0.04) 0.23 (0.08) *p<0.05

Table 2 Physiological variables during inflation and deflation at the continuous distending pressure (Paw) where the oxygenation index (OI) is minimal (PawminOI) and at the lowest Paw where Q′S/Q′T is less than 0.1 (Pawoptimal). (PaO2 arterial partial pressure of oxygen, Q′S/Q′T physiological shunt fraction) Data are present- ed as means (±SD)

At PawminOI At Pawoptimal Inflation

Paw (cmH2O) 30 (4) 32 (6) PaO2 (mmHg) 472 (51) 483 (46) OI 6.2 (1.4) 6.4 (1.7) Q′S/Q′T 0.09 (0.03) 0.08 (0.02) Deflation

Paw (cmH2O) 13(2) 14 (2) PaO2 (mmHg) 489 (47) 509 (40) OI 2.4 (0.3) 2.6 (0.2) Q′S/Q′T 0.09 (0.03) 0.07 (0.02)

at Paw 40 cmH2O during inflation and subsequently in- creased to 0.61±0.20 at Paw 9±2 cmH2O during deflation (Table2). The shunt criterion to determine optimal Paw was reached in 15 out of 16 cases. In one animal Q′S/Q′T did not decrease below 0.1 during inflation, therefore we took the Paw at the minimum Q′S/Q′T in this case. This yielded Paw of 32±6 cmH O during inflation and Fig. 1 Arterial partial pressure of oxygen (PaO2), oxygenation in- optimal 2 dex (OI) and physiological shunt fraction (Q′S/Q′T) as a function Pawoptimal of 14±2 cmH2O during deflation (Table 2). of continuous distending pressure (Paw) in one animal, with the Figure 2 shows the individual Q′S/Q′T curves with PawminOI dashed line at Q′S/Q′T =0.1. The arrows indicate the order of Paw and Paw indicated. The difference of Paw and application optimal minOI Pawoptimal was Ð1.9±4.2 cmH2O (NS) during inflation and Ð1.5±1.6 cmH2O (p<0.05) during deflation. The individu- al differences were within one Paw step (±3 cmH2O) in 15 out of 16 cases. In the remaining case, obtained during in- flation, the difference was Ð12 cmH2O. Figure 3 shows 1154

Fig. 2 Individual physiological shunt fraction (Q′S/Q′T) curves as a function of continuous dis- tending pressure (Paw) in eight animals. Highlighted data points indicate the lowest Paw where Q′S/Q′T was less than 0.1 (Pawoptimal) during the inflation (dot) and deflation phase (square). The arrows indicate the Paw where the oxygenation index (OI) is minimal on the inflation (dashed arrow) and deflation limb (solid arrow)

the hemodynamic effect of varying Paw levels in refer- nition for the optimal Paw is not easy to obtain. Clearly, ence to Pawoptimal on the deflation limb. A decrease in Paw the ultimate goal of mechanical ventilation is survival resulted in a significant increase in Q′T and DO2. without lung injury or other short- or long-term morbidity. There is increasing evidence that -induced lung injury can be minimized by an open lung approach [13, Discussion 14], a recruitment procedure followed by stabilization at lower airway pressures. The results of our study are indica- The intention of this study was to investigate if the mini- tive of the beneficial effect of a recruitment procedure, as mal OI coincided with the optimal Paw. A consistent defi- similar levels of oxygenation were obtained at a much low- 1155

also by Q′T in the presence of a ventilation-perfusion mis- match (V′/Q′<1): Decrease of pulmonary perfusion may normalize the mismatch and thereby decrease Q′S/Q′T. The OI combines the above-mentioned goals to maxi- mize oxygenation at minimal pressure cost with the mini- mization of FIO2 to prevent . The endeav- or for a minimal OI can be translated as follows: a 10% increase in Paw is only indicated if it yields a PaO2 in- crease of at least 10% while maintaining FIO2 at a con- stant level. Also, a 10% increase in Paw is only indicated if an FIO2 reduction of at least 10% can be achieved while maintaining PaO2 at a constant level. In this study we found that the OI reached a minimum at a Paw which coincided fairly well with the Pawoptimal according to the physiological shunt criterion, both on the inflation and deflation limb. Elaborating on these results, we hypothe- size that the OI can be used as a clinical guide to optimize lung volume: after initiation of HFOV, Paw is increased in steps until the OI starts to rise: the lungs are adequately recruited. Then, Paw is lowered in steps, initially result- ing in a decreasing OI, until the OI starts to rise. Then, Paw is increased one step to obtain the Pawoptimal. In previous, similar experiments we observed (unpub- lished observations), by continuous intra-arterial PaO2 measurement, a slow, ongoing recruitment at a constant Paw during the inflation phase. In our study we chose a stabilization time of 10 min, firstly to limit the total ex- perimentation time in order to assure the stability of the animal model and, secondly, to mimic the clinical prac- tice of HFOV application. A longer stabilization time may have resulted in a smaller hysteresis effect, as pro- gressive recruitment may have increased oxygenation on the inflation limb and de-recruitment may have de- creased oxygenation on the deflation limb. In previous studies [4, 5, 6, 7] the OI was used to assess

Fig. 3 Physiological shunt fraction (Q′S/Q′T), cardiac output (Q′T) severity of illness in neonatal and pediatric patients. We and oxygen delivery (DO2) as a function of continuous distending found a highly non-linear relationship between Paw and pressure (Paw) in reference to the optimal Paw (Pawoptimal) on the PaO (see example in Fig. 1). Therefore, the OI, which is deflation limb. The order of Paw application is from right to left. 2 Values were tested against those at Paw . * p<0.05 proportional to Paw/PaO2 at constant FIO2, is dependent optimal on Paw and is thus strongly related to ventilatory condi- tions. A similar reasoning accounts for the relationship be- er Paw on the deflation limb, as compared to the inflation tween FIO2 and PaO2. In previous studies the PaO2/FIO2 limb. A test of model stability (Table 1) confirmed that this relationship was found to be FIO2-dependent [17, 18]. was not due to improvement of the animal’s condition. Fol- Therefore, the OI, which is inversely related to PaO2/FIO2 lowing a recruitment procedure, there may be a range of at constant Paw, will be dependent on FIO2. Based on these Paw values that can be considered optimal, the “safe win- findings, use of the OI for assessment of severity of illness dow” according to Froese [15]. We chose to define Pawoptimal should be undertaken with some caution. as the lowest Paw where the lungs are still fully recruited In conclusion, in an animal model of experimental lung as a compromise between the minimization of atelectasis injury we found a consistent pattern in OI as a function of and the minimization of airway pressure, to prevent de- Paw, with an indicative minimum. The minimal OI is found pression of Q′T [16]. Based on our results we confirm the close to the Paw where oxygenation is optimal. Further inverse relationship between Q′T and Paw: Paw above the studies are needed to investigate the response of OI to FIO2 Pawoptimal yielded a significant decrease of Q′T and DO2 and to clarify the applicability of the index in patients. (Fig.3). We chose to define a fully recruited lung at a Acknowledgements We are grateful to Arnold Drop for technical Q′S/Q′T below 0.1. Here we must take into consideration assistance during the experiments. The assistance of Tom Leen- that shunt may not only be determined by recruitment but hoven in setting up the experiments is greatly appreciated. 1156

References

1. Gerstmann DR, Minton SD, Stoddard 6. Ortega M, Ramos A, Atkinson J, 13. McCulloch PR, Forkert PG, Froese AB RA, Meredith KS, Monaco F, Bertrand Bowman CM (1987) Oxygenation in- (1988) Lung volume maintenance pre- JM, Battisti O, Langhendries JP, dex can predict outcome in neonates vents lung injury during high frequen- Francois A, Clark RH (1996) The who are candidates for extracorporeal cy oscillatory ventilation in surfactant- Provo multicenter early high-frequency membrane oxygenation. Crit Care Med deficient rabbits. Am Rev Respir Dis oscillatory ventilation trial: improved 15:226 137:1185Ð1192 pulmonary and clinical outcome in res- 7. Arnold JH, Hanson JH, Toro-Figuero 14. Amato MB, Barbas CS, Medeiros DM, piratory distress syndrome. Pediatrics LO, Gutierrez J, Berens RJ, Anglin DL Magaldi RB, Schettino GP, Lorenzi- 98:1044Ð1057 (1994) Prospective, randomized com- Filho G et al. (1998) Effect of a protec- 2. Bryan AC, Froese AB (1991) Reflec- parison of high-frequency oscillatory tive-ventilation strategy on mortality in tions on the HIFI trial. Pediatrics ventilation and conventional ventila- the acute respiratory distress syndrome. 87:565Ð567 tion in pediatric respiratory failure. N Engl J Med 338:347Ð354 3. Gattinoni L, Pelosi P, Crotti S, Valenza Crit Care Med 22:1530Ð1539 15. Froese AB (1997) High-frequency os- F (1995) Effects of positive end-expi- 8. Lachmann B (1992) Open the lung and cillatory ventilation for adult respirato- ratory pressure on regional distribution keep the lung open. Intensive Care ry distress syndrome: Let’s get it right of tidal volume and recruitment in Med 118:319Ð321 this time! Crit Care Med 25:906Ð908 adult respiratory distress syndrome. 9. Böhm SH, Vazques de Anda GF, Lach- 16. Traverse JH, Korvenranta H, Adams Am J Respir Crit Care Med 151:1807Ð mann B (1998) The open lung concept. M, Goldthwait DA, Carlo WA (1988) 1814 In: Vincent JL (ed) Yearbook of inten- Impairment of hemodynamics with in- 4. Ortiz RM, Cilley RE, Bartlett RH sive care and emergency medicine. creasing mean airway pressure during (1987) Extracorporeal membrane oxy- Springer, Berlin, pp 430Ð440 high-frequency oscillatory ventilation. genation in pediatric respiratory fail- 10. Guyton AC (1986) Textbook of medi- Pediatr Res 23:628Ð631 ure. Pediatr Clin North Am 34:39Ð46 cal physiology, 7th edn. Saunders, 17. Herrick IA, Champion LK, Froese AB 5. Durand M, Snyder JR, Gangitano E, Philadelphia, pp 481Ð503 (1990) A clinical comparison of indi- WU PYK (1990) Oxygenation index 11. Jansen JR, Versprille A (1986) Im- ces of pulmonary gas exchange with in patients with meconium aspiration: provement of cardiac output estimation changes in the inspired oxygen concen- conventional and extracorporeal mem- by the thermodilution method during tration. Can J Anaesth 37:69Ð76 brane oxygenation therapy. Crit Care mechanical ventilation. Intensive Care 18. Gowda MS, Klocke RA (1997) Vari- Med 18:373Ð377 Med 12:71Ð79 ability of indices of hypoxemia in 12. Grotjohan HP, Heijde RMJL van der adult respiratory distress syndrome. (1992) Experimental models of the res- Crit Care Med 25:41Ð45 piratory distress syndrome: lavage and oleic acid (PhD thesis). Erasmus Uni- versity Rotterdam, The Netherlands, pp 45Ð72