14th Annual Congress ± Geneva, Switzerland ± 30 September±3 October 2001

S 135

Oral Presentations New modes of mechanical ventilation ± 1±5

3

1

Max M, Dembinski R, Kuhlen R, Bensberg R, Rossaint R. Klinik fuer Anaesthesie, Universitaetsklinikum der RWTH Aachen, Aachen, Germany

PHYSIOLOGIC EFFECTS OF PAV IN NEUROLOGICAL PATIENTS

Santos J,Cacciotti R, Bellapart J, Mancebo J. Intensive Care Medicine, Hospital Sant Pau, Barcelona, Spain INTRODUCTION. Few physiological date are available in-patients undergoing mechanical ventilation because neurological disturbances. Objective: To analyze the short term physiologic effects on breathing pattern, inspiratory muscle effort and arterial blood gases of proportional assist ventilation (PAV) in neurological patients. METHODS. We studied 10 patients who were mechanically ventilated only because neurological disturbances. Patients were comatose, free of sedation, and were capable to trigger the ventilator. We measured airflow, and airway, esophageal and gastric pressures. Patients were ventilated in CPAP (6  1 cmH2O) and two levels of PAV (high, 65 % assist, and low, 45 % assist) at same PEEP levels. Arterial blood gases were measured at the end of each ventilatory period, each lasting about 30 minutes. We calculated airway occlusion pressure (P0.1), breathing pattern and indexes of inspiratory effort were according to standard methods. Statistical analysis: ANOVA. RESULTS. Average FiO2 was 0.38. PEEPi ranged between 1 and 1.2 cmH2O (p: ns). PH ranged between 7.43 and 7.45 (p: ns). PaO2 ranged between 106 and 115 mmHg, and PaCO2 ranged between 33 and 35 mmHg (p: ns).

CPAP PAV high PAV low p

F (breaths/ min)

Vt (mL)

Ti/Ttot (%)

WOB (j/min)

PTPes (cmH2O.s /min)

245 217 236 0.6

450123 529123 47284 0.3

365 318 359 0.4

9.55.4 4.82.4 6.32.9 0.04

17250 7938 11650 0.001

PTPdi P01 (cmH2O.s (cmH2O) /min) 16366 8046 10764 0.02

2.71 1.60.6 20.8 0.02

CONCLUSION. In neurological ventilator-dependent patients, PAV downregulates inspiratory muscle effort without modifying breathing pattern or ABGs

CHANGES IN VENTILATION-PERFUSION DISTRIBUTION DURING PROPORTIONAL ASSIST VENTILATION IN ACUTE LUNG INJURY

INTRODUCTION. The aim of this study was to compare the effects of proportional assist ventilation (PAV) with controlled mechanical ventilation (CMV) both with high-PEEP levels on ventilation-perfusion distribution (VA/Q) in an animal model of acute lung injury (ALI). METHODS. Experimental ALI was established in 20 pigs by repeated lung lavage with saline. Subsequently, animals were randomized to undergo either CMV with a tidal volume of 10 ml/kg BW and a respiratory rate of 20/min (n = 10) or PAV with a flow assist and volume assist set to achieve an unloading of 80 % of the resistive and elastic respiratory work, a tidal volume > 4 ml/kg BW and a respiratory rate < 60/min (n = 10). PEEP was set at 10 mbar in both groups. VA/Q distributions were analyzed using the multiple inert gas elimination technique (MIGET) after the onset of ALI and after 2, 4, 8, and 12 hours. Data were analyzed within and between the groups using ANOVA for repeated measurements, followed by the Duncan post hoc test for multiple comparisons. RESULTS. PaO2 increased significantly in both groups at all study points when compared to ALI (p < 0,05). However, this increase was significantly more pronounced in the CMV group when compared to corresponding time points in the PAV group (p < 0,05). A decrease of QVA/QT was observed with both ventilatory modes, while an increase in Q low and Q ideal was restricted to the CMV group. Selected MIGET data of each group after ALI and after 4 and 12 h of PAV or CMV are presented in table 1 (§ p < 0,05 for comparison within each group versus ALI, * p < 0,05 for comparison between the groups at corresponding study points).

QVA/QT logSD Q VD/VT logSD V Low VA/Q high VA/Q ideal VA/Q

CMV ALI

CMV 4

CMV 12

PAV ALI

PAV 4

PAV 12

5310 1.30.9 566 0.50.1 0.030.1 1.02.0 446

129§* 1.40.8 566 0.70.1 0.20.3 0.51.1 436

1013§* 1.40.7 583 * 0.60.1 0.20.3 0.00.0 423 *

5110 1.40.6 447 0.50.1 0.030.1 0.00.0 567

3715§ 1.80.9 607 § 0.60.1 0.10.1 0.10.2 407 §

3016§ 1.80.8 656§ 0.70.3 0.10.1 0.82.1 347§

CONCLUSION. We conclude that CMV exerts a different pattern of VA/Q distribution than PAV when applied with high PEEP. It is more effective to decrease shunt and preserve ideal VA/Q areas compared to PAV, although PaO2 is improved with both ventilatory modes. REFERENCE. Supported by Deutsche Forschungsgemeinschaft, DFG (Ku 1372/1±1)

2

4

NON-INVASIVE MEASUREMENT OF RESPIRATORY SYSTEM ELASTANCE (E) AND RESISTANCE (R) DURING PROPORTIONAL ASSIST VENTILATION (PAV)

NEURALLY ADJUSTED VENTILATORY ASSIST IN ACUTE RESPIRATORY FAILURE

Grasso S1, Ranieri V M2, Brochard L3, Richard G C4, Mancebo G5, Slutsky A S6, Younes M7. 1 Intensive Care, Di Venere Hospital, Bari, Italy, 2Pisa, Italy, 3 Paris, France, 4Rouen, France, 5 Barcelona,Spain, 6 Toronto,Canada, 7Manitoba,Canada INTRODUCTION. During PAV respiratory system E and R can be non-invasively estimated from the the response of airway pressure and flow to a brief end inspiratory airway occlusion period. METHODS. PAV was applied for 3 hours in 26 ventilated patients. Flow, Volume (flow integration), airway opening (Pao) and trans-diaphragmatic (Pdi) pressures were measured. R and E obtained appliyng the equation of motion (Rmeas and Emeas) were compared with their estimation (Rcalc and Ecalc) obtained using the end inspitratory occlusion technique (300 msec occlusion every 12 breaths). Agreement between measured and calculated mechanics was evaluated by regression analisys. RESULTS. A significant correlation bertween measured and calculated values was found:

R2 P

Rmeas VS Rcalc

Emeas VS E calc

0.727 0.001

0.918 0.001

CONCLUSION. Non-invasive estimation of E and R during PAV is reliable and can be used for easy PAV implementation and continuous adaptation of ventilatory assistance to respiratory mechanics.

Gottfried SB1, Skrobik Y2, Laufer B2, Navalesi P1, Comtois N2, Beck J2, Spahija J2, Sinderby C2. 1Meakins-Christie Laboratories, McGill University Health Centre, Montreal, 2Guy Bernier Research Centre, Hopital Maisonneuve-Rosemont, Montreal, Canada INTRODUCTION. The ability to reliably record diaphragmatic electrical activity (EAdi) in real time in human subjects has enabled the development of prototype systems in which EAdi can control various aspects of ventilator operation (1). One such application is neurally adjusted ventilatory assist (NAVA), an experimental mode of ventilatory support in which positive pressure is applied in proportion to EAdi. In principle this should facilitate patient-ventilator coordination as well as allow ventilatory support to change in response to alterations in respiratory mechanics, muscle function, and drive. The aim of this study was to evaluate the feasibility of NAVA in patients with acute respiratory failure as well as to determine the effects of varying levels of support on respiratory parameters. METHODS. Flow, volume, respiratory pressures, and EAdi were measured in 9 intubated patients with acute respiratory failure of varied etiology (ARDS, COPD, pneumonia, post-operative). Following a period of spontaneous breathing (SB), NAVA was instituted and the proportionality between ventilator applied pressure (Pvent) and EAdi was incrementally increased until any of the predetermined criteria for maximal assist (NAVA,max) occurred (Pvent > 50 cmH2O, periodic breathing, 3 consecutive breaths > 15 ml/kg, desaturation, cardiovascular deterioration, intolerance). RESULTS. NAVA was well tolerated in all 9 patients. Patient-ventilator coordination was excellent during NAVA and did not deteriorate at higher levels of support. Increasing NAVA produced graded reductions in inspiratory neuromuscular activity, as reflected by integrated minute activity of EAdi and transdiaphragmatic pressure (Pdi). Breathing pattern improved, particularly at lower levels of support. Tidal volume (ml) Frequency (min-1) Minute Ventilation (L/min) Integrated Pdi (cmH20) Integrated EAdi (a. u.)

SB

NAVA 50 % max

NAVA max

36144 21.252.49 7.110.76 23032 13023

54771 * 16.632.83 8.181.02 8312 * 426 *

58976 * 17.052.93 8.781.03 499 * 273 *

* p < 0.05, SB vs NAVA CONCLUSION. NAVA appears to be a feasible mode of ventilatory support in intubated patients with acute respiratory failure. In addition to facilitating patient-ventilator coordination, NAVA can effectively reduce indices of inspiratory activity while improving breathing pattern. REFERENCE. Sinderby C, Navalesi P, Beck J, Skrobik Y, Comtois N, Friberg S, Gottfried SB,Linstrom L. Neural control of mechanical ventilation in respiratory failure. Nature Medicine 5: 1433±1436, 1999.

14th Annual Congress ± Geneva, Switzerland ± 30 September±3 October 2001

S 136

5

7

A SIMPLIFIED WEANING PROTOCOL BASED ON ADAPTIVE SUPPORT VENTILATION: EFFECT ON DURATION OF INTUBATION AND RESPIRATORY MANAGEMENT

CXC CHEMOKINE RECEPTOR EXPRESSION IN SEVERE SEPSIS IN MAN

Revelly JP1, Petter A1, Chassot PG2, Mueller XM3, ChiolØro R1. 1Soins Intensifs de Chirurgie, 2Service d'AnesthØsiologie, 3Service de Chirurgie cardio-vasculaire, Centre Hopsitalier Universitaire Vaudois, Lausanne, Switzerland INTRODUCTION. Adaptive Support Ventilation (ASV) is a microprocessor controlled ventilatory mode that maintains a preset minute ventilation (VolMin). Preliminary data suggest that ASV may accelerate postoperative respiratory weaning. The goal of this study was to assess the influence of a simplified ASV-based weaning protocol on the duration of intubation, the manipulations of the ventilator performed by the healthcare team and the occurrence of respiratory adverse events. METHODS. Patients after uncomplicated cardiac surgery were randomly assigned to an ASV (GrA) or a standard (GrS) protocol. Both protocols were divided in 3 predefined Phases (P1, P2, P3). In P1, full ventilatory support was provided by ASV 100 % Vol.Min. in GrA, and SIMV in GrS. P1 lasted until sustained spontaneous breathing occurred ( > 6 breaths/ min. for 10 min.). Respiratory weaning was performed in GrS by switching to IPS 10 cmH2O (P2), than 5 cmH2O (P3). In GrA, the settings were left unchanged until inspiratory pressure had automatically decreased to 10 cmH2O (end of P2). The patient was then extubated according to predefined criteria (identical in the 2 groups), after a short trial of IPS 5 cmH2O (P3). Weaning duration (min.), peak inspiratory pressure (Ppeak, cmH20), tidal volume (VT, ml), respiratory rate (RR, bpm), and P0.1 (-cmH20)were recorded continuously, as well as the number of changes in the settings of the ventilator (CSV), and of episodes of high pressure alarm ( > 35 cmH2O, HPA). Comparisons between groups and over time were performed by by two-ways ANOVA. *: different from the control group, p < 0.05. RESULTS. There were no differences between groups in the duration of tracheal intubation, nor in ventilatory parameters, besides a lower Ppeak in GrA at P1. Patients in GrA necessitated less manipulations of the ventilator and sustained less episodes of high pressure.

Duration P peak Vt RR P01 CSV HPA HPA

Group A Phase 1

(n = 18) Phase 2

± Phase 3

Group S Phase 1

(n = 16) Phase 2

± Phase 3

10373 17.52.5* 52296 13.12.4 1.61.8 1.20.7

6977 15.43.4 50469 16.34.8 2.62.2 0.10.3*

2320 11.81.7 46572 19.26.0 3.82.1 1.20.4

10449 22.26.2 512113 12.00.3 1.51.5 1.30.5

5641 16.92.2 549122 14.03.1 2.71.5 1.20.4

5062 12.52.0 47673 19.65.5 4.62.3 1.40.6

0.72.4* 0.72.4*

2.93.0 2.93.0

CONCLUSION. Respiratory weaning with a simplified ASV protocol was as rapid as the standard of care for rapid tracheal extubation in ªfast-trackº patients. The observation of less ventilator manipulation and high pressure alarms in ASV suggests that this mode may facilitate patient care and improve patient-machine interaction. REFERENCE. Linton DM et al. Chest 1994; 106: 1843±50

Chishti AD, Kirby J, Baudou in SV. Anaesthesia, Royal Victoria Infirmary, Newcastle Upon Tyne, United Kingdom INTRODUCTION. Understanding the molecular mechanisms responsible for neutrophil (PMN) recruitment and activation during sepsis are central in developing strategies to limit the adverse effects of inflammation. The chemokine interleukin- 8 (IL-8) is a potent and specific chemoattractant and activator of neutrophils. Two high affinity receptors, CXCR1 and CXCR2 mediate the IL-8 signal in PMN. We investigated the expression and function of CXCR1, CXCR2 and CD11 b the Beta 2 integrin essential in neutrophil migration and a marker of neutrophil activation, in patients suffering from severe sepsis. METHODS. Fifteen patients in the intensive care unit (ICU) of the Royal Victoria Infirmary were prospectively identified as having sepsis syndrome agreeddefinitions1. All subjects were enrolled within 24 hrs of the onset of developing sepsis syndrome. Eight patients admitted to the intensive care unit with diagnoses other than sepsis acted as ICU controls whilst eight volunteers served as healthy controls. Blood samples were collected from each subject every 24 h for 5 days. Receptor expression was determined by immunoflourescent flow cytometry. Plasma levels IL-8 were measured by an enzyme linked immunosorbent assay technique. Neutrophil chemotaxis against IL-8 was assessed on day 1 for six subjects in each group using 5 m pore chemotaxis chambers. RESULTS. 30 day mortality was 47 % in the sepsis group, no ICU control patients died. Mean APACHE II score for septic patients was 25 ªb 8 (mean  SD) and 16  5 (mean  SD) for ICU controls. High levels of IL-8 levels were found in both ICU groups, which subsequently decreased over 5 days in both the septic (497.4 pg/mL, day 1 to 137.1 pg/mL, day5, p = .001) and ICU control groups (21.2 to 6.2 pg/mL p < 0.0001). No detectable IL-8 was observed in healthy subjects. IL-8 levels in the sepsis group were significantly elevated compared to the ICU group (p < 0.0001), levels were between 9 to 24 times higher. Neutrophil expression of CXCR1 in ICU controls was depressed compared to CXCR1 expression from septic and healthy controls over the first 3 days, p < 0.04. PMN CXCR2 expression during sepsis was significantly reduced compared to healthy controls, p < 0.02 but compared to ICU group CXCR2 expression was significantly depressed for the first 2 days, p < 0.02. CXCR2 expression in the control groups were not significantly different, p = 0.5. In both ICU groups CD11 b expression was elevated compared to healthy controls, p < 0.0001 and p = 0.002 respectively. Septic patients had significantly higher PMN CD11 b expression than PMNs in ICU controls p = 0.0001. PMN chemotaxis was reduced in the sepsis group compared to the two control groups, p < 0.05. CONCLUSION. The main findings from our study were that the surface expression of chemokine receptors CXCR1 and CXCR2 changes on circulating neutrophils from critically ill patients compared to healthy controls. A reduced CXCR2 expression in sepsis corresponded to a decreased chemotaxis, which may be significant in terms of immunosuppression in critical illness. and a target for receptor antagonists. REFERENCE. Bone RC, Balk FB, Cerra RP et al. Chest 1992; 101: 1644±55 Funding British Journal of Anaesthesia.

Oral Presentations Sepsis on the cellular level ± 6±10

8

6

Stamer UM, Eggert C, Putensen C, Stüber F. Department of Anesthesiology and Intensive Care Medicine, University of Bonn, Bonn, Germany

CORRELATION BETWEEN IL-6 PLASMA LEVELS AND ORL-1 MRNA EXPRESSION

THE EFFECT OF CAUSATIVE ORGANISM ON MARKERS OF COAGULATION, FIBRINOLYSIS, AND INFLAMMATION IN SEVERE SEPSIS Dhainaut J1, Laterre P2, Levy H3, Opal S4. 1Serv Reanimation Medicale, Groupe Hospitalier Cochin, Paris, France, 2Emergency /Inten Care, Cliniques Universitaire Saint Luc, Brussels, Belgium, 3Crit Care, Eli Lilly, Indianapolis, IN, 4ID, Brown Univ., Providence, RI, USA INTRODUCTION. In a phase III, double-blind, placebo-controlled trial (PROWESS), recombinant human activated protein(drotrecogin-alpha (activated)) reduced 28-day all-cause mortality in patients with severe sepsis.1 As a secondary objective of the study, the effect of pathogen type on the host inflammatory and coagulopathic response to infection was assessed. METHODS. Markers of inflammation (IL-6) and coagulopathy were assessed at baseline for all placebo patients (n = 840). Markers of coagulopathy included D-dimer (D-di), platelet count (PLTS), prothrombin time (PT), Protein C activity (PC), and antithrombin activity (AT). Baseline values were grouped by survivorship (alive or dead at study day 7) and by type of pathogen: pure Gram-negative (G-); pure Gram-positive (G + ); fungal (Fgl); and suspected infection with unknown pathogen (Unk). A blinded clinical evaluation committee assessed the type of pathogen. Ranked ANOVA was used to assess differences between groups. Mixed Gram, mixed bacterial/non-bacterial, other pure non-bacterial groups, and no infection groups are not presented (N = 161). RESULTS. Summarized in the following table are the results of the analysis.

G- (S) G- (NS) G+ (S) G+ (NS) Fgl (S) Fgl (NS) Unk (S) Unk (NS)

N

D-di (mcg/mL) NR < = 0.39

PLTS (x109/L) NR 140± 400

151 61 151 66 5 8 162 75

5.9 4.5 3.6 5.1** 3.7 8.3 3.6 4.9*

177 157 191 162* 179 116 186 186

PC (%) PT (sec) NR = 81± NR < = 14.5 173 19.0 21.3** 17.7 19.0* 18.4 19.4 17.6 19.0

51 39* 53 45 63 39* 54 45*

AT (%) IL6 NR = 80± (pg/mL) 120 NR < = 10.1 60 47*** 62 52* 77 43** 67 59

503 1172** 420 2544*** 481 1180 255 479

N = number of patients; NR = normal range; (S) = 7-day survivor; (NS) = 7-day non-survivor; *p < = 0.05; ** p < 0.01; ***p < 0.001 for differences among (S) and (NS) within each causative microorganism group CONCLUSION. Median levels for markers of coagulopathy and inflammation were distinctly abnormal in patients with severe sepsis independent of the causative microorganism, including the group of patients with a suspected but unidentified infection. Baseline values for these markers were either significantly or trended towards more severe in the non-survivors than the survivors regardless of the class of pathogen. REFERENCE. Bernard GR, Vincent JL, Laterre PF et al. NEJM 2001; 344: 699±709.

INTRODUCTION. The human ORL-1 receptor (opioid receptor like) is distributed throughout the brain (1). Gene transcript were also detected in circulating lymphocytes and monocytes (2±3). A role of this opioid receptor in the function of immune system is supposed. This study investigates the mRNA expression of the ORL-1 gene in peripheral blood cells of healthy individuals and critically ill patients ex vivo and correlates these findings to the severity of inflammation (Il-6 levels). METHODS. After approval of the local ethics committee, human whole blood was drawn from healthy volunteers. Whole blood cultures with LPS 0 ng, 0.1 ng, 10 ng or 100 ng/ml were performed at 37 C and 5 % CO2 for 3, 6, 12 and 24 h. Whole blood from critically ill patients of the ICU was drawn and immediately frozen for further analysis. After RNA isolation and cDNA synthesis, reverse transcriptase polymerase chain reaction (rtPCR) using specific primers for ORL-1 was performed. PCR products were analyzed by agarose gel electrophoresis. RNA contents was estimated by semiquantitative analysis employing a housekeeping gene as internal standard. Il-6 plasma levels were measured in patients and healthy controls, as well as in supernatants of the blood culture. RESULTS. ORL-1 was expressed constitutively in human peripheral blood cells of healthy volunteers, but only in some of the critically ill patients. After incubation with LPS, semiquantitative rtPCR revealed a dose and time dependent down regulation of the ORL-1 mRNA expression in the blood of healthy individuals ex vivo. ORL-1 message was not impaired in critically ill patients without signs of systemic inflammation. In septic patients, however, ORL-1 message was suppressed. A negative correlation between ORL-1 message and Il-6 plasma levels could be detected (r = -0.7 in blood cultures). CONCLUSION. Endotoxin impaired ORL-1 expression in human peripheral blood cells ex vivo. In septic patients ORL-1 expression was down regulated at baseline and a negative correlation could be detected between ORL-1 expression and Il-6 levels. The results suggest, that ORL-1 is involved in immune response to infection and may contribute not only to neuronal functions but also to physiological functions during inflammation. REFERENCES. 1. Xie GX et al. Life Sciences 1999, 64: 2029±37. 2. Peluso et al. J Neuroimmunol 1998,81: 184±192 ; 3. Wick et al. Mol Brain Res 1995, 32: 343±347

14th Annual Congress ± Geneva, Switzerland ± 30 September±3 October 2001

9 MITOCHONDRIAL INHIBITION IS ASSOCIATED WITH THE SEVERITY OF HUMAN SEPTIC SHOCK

Brealey DA1, Brand M1, Hargreaves I2, Heales S2, Land J2, Smolenski R3, Singer M1. 1Medicine, University College London, 2Clinical Biochemistry, Institute of Neurology, London, 3 Heart Science Centre, Harefield Hospital, Harefield, UK INTRODUCTION. Nitric oxide (NO) has, in addition to its vascular effects, inhibtory actions on mitochondrial function. These include inhibition of respiratory chain components, in particular complexes I and IV. This may be an important mechanism in causing the organ dysfunction associated with septic shock (1). We have found that patients dying of septic shock have lower muscle ATP levels than survivors (2). We thus sought to investigate the link between norepinephrine (NEPI) requirement, NO production, respiratory chain activity and tissue ATP. METHODS. A vastus lateralis muscle biopsy was taken from patients with septic shock within 24 hours of ICU admission and frozen immediately at ±200 C. The samples were later assayed for 1. complex I and IV activity by spectrophotometric methods. Results are expressed as a ratio to citrate synthase activity (to control for mitochondrial enrichment). 2. Total tissue nitrite (NOx) levels by a modified Griess reaction,expressed as micromol/mg protein. 3. ATP levels by HPLC, expressed as nmol/mg protein. RESULTS. 23 patients were recruited to the study. The relationship between NEPI, tissue NOx, complex I activity and tissue ATP levels are shown in Table. Due to variation in the amount of tissue obtained, we were unable to perform a full set of analyses on all patients (indicated in Table). No relationship was found between complex IV and any other variable.

Complex I activity ATP nmol/mg protein NOx micromol/mg protein NEPI mcg/min

Complex I activity

ATP Nmol/mg protein

NOx micromol/ mg protein

±

r = 0.6, p = 0.02, n = 15

r = -0.8, p = 0.01, n = 8 r = -0.7, p = 0.05, n = 8

r = 0.6, p = 0.02, n = 15 r = -0.8, p = 0.01, n = 8 r = -0.7, p < 0.01, n = 23

± r = -0.7, p = 0.05, n = 8 r = -0.5, p = 0.04, n = 16

± r = 0.9, p = 0.01, n=9

NEPI mcg/min r = -0.7, p < 0.01, n = 23 r = -0.5, p = 0.04, n = 16 r = 0.9, p = 0.01, n=9 ±

CONCLUSION. This is the first demonstration in patients that the increasing severity of septic shock (assessed by norepinephrine requirements) is related to raised tissue NOx levels, decreased complex I activity and ATP levels. These results have important implications for the pathogenesis of septic shock and organ dysfunction. REFERENCES. 1. Singer M, Brealey D. Biochem Soc Symp. 1999; 66: 149±66. 2. Brealey D, Smolenski R, Singer M. Muscle ATP levels in early septic shock predict outcome. Intensive Care Med. 2000; 26: S321.

S 137

Oral Presentations Novel haemodynamic interventions (I) ± 11±15 11 HAEMODYNAMIC SPLANCHNIC EFFECTS OF ATP-MGCL2 DURING HYPERDYNAMIC PORCINE ENDOTOXAEMIA Asfar P1, Nalos M1, Pittner A1, Theisen M1, Ichai C1, Ploner F1, Georgieff M1, Ince C2, Brückner UB1, Leverve XM3, Radermacher P1, Fröba G1. 1Abt. Anästhesie and Sekt. Chir. Forschung, Universitätklinik für Anästhesiologie, Ulm, Germany, 3Lab. BioØnergetique Fond. et Appl., UJF, Grenoble, France, 2 Exp. Anesthesia, AMC, Amsterdam, The Netherlands INTRODUCTION. ATP-MgCl2 has been reported to exert beneficial effects in short-term shock models characterized by decreased cardiac output. Given the particular role of the gut in multiple organ dysfunction, we tested ATP-MgCl2 effects on intestinal haemodynamics, O2 exchange and metabolism in hyperdynamic porcine endotoxaemia . METHODS. 12 h after starting continuous i. v. endotoxin (LPS), pigs received either placebo (CON = 8) or 0.3 Dmol/kg/min of ATP-MgCl2 (ATP = 9). Before as well as at 12 and 24 h LPS we assessed systemic and portal venous (Qpv) blood flow, mucosal microcirculation (Cytoscan), gut O2-extraction (EO2) and ileal-arterial PCO2 gap. RESULTS. Data are median (25 %;75 %), *p < 0.05 vs pre-shock (Friedman) and § p < 0.05 vs CON (Mann-Withney). Changes of PCO2 gap were inversely correlated with Qpv variations (r ±0.68; p < 0.05).

CON MAP(mm Hg) ATP MAP(mm Hg) CON CI(ml/kg/min) ATP CI(ml/kg/min) CON Q pv(ml/kg/min) ATP Q pv(ml/kg/min) CON Gut EO2 % ATP Gut EO2 % CON PCO2 gap(mm Hg) ATP PCO2 gap(mm Hg) CON pv pH ATP pv pH CON perfused villi (no) ATP perfused villi (no) CON non-perfused villi (no) ATP non-perfused villi (no)

Before LPS

12 h LPS

24 h LPS

92 (91;97) 94 (90;96) 116 (106;123) 112 (90;140) 22 (18;23) 19 (18;24) 32 (31;35) 35 (30;45) 15 (10:22) 16 (14:24) 7.46 (7.45;7.46) 7.42 (7.39;7.43) 14 (12;16) 13 (10;16) 0 (0;0) 0 (0;0)

101 (97;106) 99 (91;115) 148 (117;165)* 134 (120;148)* 25 (22;28)* 22 (20;29) 26 (23;28)* 34 (27;36) 30 (21:43)* 41 (32:60)* 7.35 (7.31;7.36)* 7.34 (7.34;7.38)* 5 (0;9)* 7 (0;12)* 6 (0;11)* 4 (0;10)*

97 (87;107) 78 (74;87)*§ 148 (140;171)* 175 (156;204)* 28 (26;31)* 42 (36;45)*§ 22 (19;23)* 20 (17;24)* 50 (32:64) 18 (12:22)§ 7.34 (7.31;7.4)* 7.33 (7.30;7.38)* 8 (0;13)* 7 (0;11)* 3 (0;8)* 3 (0;10)*

CONCLUSION. Given the unchanged ileal mucosal microcirculation the ATP-MgCl2- induced fall in PCO2 gap suggests a wash out of the metabolically derived CO2 rather than the existence of stagnant hypoxia.

10

12

SELECTIVE SEROTONIN-RECEPTOR ANTAGONISM AND LEUKOCYTEINDEPENDENT PLASMA EXTRAVASATION DURING ENDOTOXAEMIA

EFFECTS OF A NOVEL POLY-(ADP-RIBOSE) SYNTHETASE INHIBITOR IN SERUM AND LUMINAL GUT LACTATE RELEASE IN A RABBIT MODEL OF ENDOTOXIC SHOCK

Walther A1, Peter C1, Schmidt W1, Secchi A1, Gebhard MM2, Martin E1, Schmidt H1. 1Department of Anesthesiology, 2Department of Experimental Surgery, University of Heidelberg, Heidelberg, Germany INTRODUCTION. Previous work from our laboratory showed a leukocyte-independent plasma extravasation during early endotoxaemia (1), that can be inhibited by the serotonin (5-HT)receptor-antagonist methysergide (mixed 5-HT1/2- receptor-antagonist) (2). It is unknown whether a more selective serotonin-receptor antagonism is also effective in reducing macromolecular efflux. Therefore, it was the aim of the study to compare the effects of the 5-HT-receptor antagonists methysergide, cinanserin (5-HT2-receptor-antagonist), and ketanserin (5-HT2Areceptor-antagonist) on leukocyte-independent endothelial damage. METHODS. In male Wistar rats, microvascular permeability (MP) and leukocyte-endothelialinteraction (leukocyte rolling, LR) were determined in mesenteric postcapillary venules using intravital microscopy at baseline, and at 60, and 120 min after start of a continuous infusion of endotoxin (ETX; 2 mg/kg/hr, E.coli O26:B6) (group A, n = 12). Animals underwent laparotomy and the mesentery was exposed beneath an in-vivo videomicroscope. MP was measured using fluorescein isothiocyanate (FITC) labelled albumin. Leukocyte-endothelial interaction was blocked by fucoidin (25 mg/kg bw) 10 min before laparotomy. Animals in group B (n = 12) received methysergide (1 mg/kg bw), animals in group C (n = 12) received cinanserin (5 mg/kg bw), and animals in group D (n = 12) received ketanserin (1 mg/kg bw) prior to baseline measurement additionally to the procedure described above. Animals in group E (saline, n = 12) only received equivalent volumes of NaCl 0.9 %. Statistical analysis was performed using twoway repeated measures ANOVA followed by the ScheffØ test. A p-value < 0.05 was considered significant. RESULTS. In groups A-D, fucoidin prevented LR during the entire experiment. In group E, LR increased significantly, starting at 60 min. Differences between the fucoidin-treated groups and the saline-group were significant at 60 min and at 120 min. However, in all groups MP increased significantly, starting at 60 min. Animals in group A were characterized by a stronger increase in MP and showed significantly higher values in MP in comparison to groups B-E at 120 min. All 5-HT-receptor-antagonists used in this experiment reduced MP to values comparable to saline group. CONCLUSION. Selective 5-HT2A-receptor antagonism is as well effective in reducing leukocyte-independent plasma extravasation during endotoxaemia as selective 5-HT2-receptor antagonism, and as well effective as mixed 5-HT1/2-receptor antagonism. Thus, serotonin effects in that patho-physiology seemed to be mainly mediated via the 5-HT2A receptor subtype. REFERENCES. 1. Walther A, Weihrauch M, et al. Crit Care Med 28: 2943±2948, 2000. 2. Walther A, Yilmaz N, et. al. Int Care Med 25 (Suppl 1): A64, 1999.

Lobo SM1, Dimopoulos G1, DeBacker D1, Sun Q1, Tu Z1, Preiser JC1, Szabo C2, Vincent JL1. 1 Dept of Intensive Care, Erasme Hospital, Free University of Brussels, Brussels, Belgium 2Inotek Corporation, Beverly, Massachusetts, USA. INTRODUCTION. During septic shock, peroxynitrite-mediated DNA strand-breaks activate the nuclear enzyme poly-(ADP)-ribose synthetase (PARS) resulting in cytotoxic depletion of cellular energy and cell dysfunction. Strategies directed to restore normal cellular energetics and function may be a key approach in the treatment of sepsis. We aimed to investigate the effects of a PARS inhibitor in systemic and regional flow and lactate production during lipopolysaccharide (LPS)-induced shock in rabbits. METHODS. After baseline measurements 12 anesthetized, paralyzed, and ventilated NewZealand rabbits were randomized into 2 groups: a) Group I (n = 6) received LPS (E. coli055:B5) 1 mg/kg and b) Group II, (n = 6), received 10 mg/kg bolus and 3 mg/kg/h continuous infusion of PARS-inhibitor PJ34 (1) 15 minutes after the same doses of LPS. Superior mesenteric artery (QSMA) and aortic (Qaorta) flows were measured using ultrasonic flow probes. A segment from the ileum was isolated using two multilumen tubes with inflated balloons allowing the perfusion and recovery of a solution to measure lactate released from the mucosa. RESULTS. The PARS inhibitor significantly reduced serum lactate (4 h, 6.7  2.6 vs 4.5  1.9; 6 h, 5.9  1.9 vs 4.0  1.6; groups I and II, respectively, p < 0.05) as well luminal gut release of lactate (6 h, 2.7  1.4 vs 0.5  0.3; groups I and II, respectively, p < 0.05) occurring in the late phase of shock and improved Qaorta (2 h, 70  16 ml/min vs 109  40 ml/min; 4 h, 70  9 ml/ min vs 94  36 ml/min; 6 h, 52  14 ml/min vs 113  8 ml/min; groups I and II, respectively, p < 0.05). There were no significant differences in QSMA between groups (2 h, 106  27 ml/ min vs 138  41 ml/min; 4 h, 116  85 ml/min vs 155  51 ml/min and 6 h, 114  95 ml/min vs 124  34 ml/min; groups I and II, respectively, p < 0.05). CONCLUSION. PARS inhibition decreased serum and luminal gut lactate concentrations probably by reducing inflammatory damage. Our data further support the view that PARS inhibition may be a suitable approach for the treatment of septic shock. Disturbed cellular oxygen utilization due to ªcytopathic hypoxiaº may occur specially in gut. REFERENCE. Soriano et al., Nat Med. 2001 Jan;7(1): 108±13)

S 138

14th Annual Congress ± Geneva, Switzerland ± 30 September±3 October 2001

13

15

EFFECTS OF DOPEXAMINE AND PEEP ON SPLANCHNIC TISSUE OXYGEN PERFUSION: THE PERFUSION PRESSURE PERSPECTIVE

SYSTEMIC AND SPLANCHNIC HAEMODYNAMIC EFFECTS OF VASOPRESSIN ADMINISTRATION IN VASODILATORY SHOCK

Lehtipalo SG, Biber B, Fröjse R, Arnerlöv C, Johansson G, Winsö O. Department of Surgical and Perioperative Sciences, Anesthesiology and Intensive Care, Umeå, Sweden

Bracco D, ChiolØro RL, Revelly JP. Surgical Intensive Care Unit, CHUV University Hospital, Lausanne, Switzerland

INTRODUCTION. Dopexamine has been reported to prevent PEEP-induced decreases in mesenteric blood flow. Whether these findings were dependent on alterations in intestinal perfusion pressure is not clear. Concurrent intestinal metabolic effects are also not well described. We designed this study to evaluate net effects of the concomitant use of PEEP and dopexamine on intestinal tissue perfusion and oxygen kinetics during pre-defined artificial reductions of intestinal perfusion pressure.

INTRODUCTION. Vasopressin is increasingly used as vasoconstrictor for refractory hypotension. During septic shock and post cardiopulmonary bypass (CPB), The effet of vasopressin therapy on regional circulation is not established. Since this agent is used for its potent splanchnic vasoconstrictive property during variceal bleeding, we hypothetized that the same phenomenon may occur in shock. The aim of this study was to assess the systemic and splanchnic haemodynamic effect of vasopressin during vasodilatory shock. METHODS. The study was conducted in accordance with the recommendation of Lausanne Ethics Committee. Patients with septic shock or post CPB vasoplegia were enrolled when unresponsive to large doses of vasopressors (norepinephrine > 1 mg/kg/min), despite adequate cardiac preload and inotropic support. Haemodynamic parameters including MAP, CI, SVRI, gastric mucosal pCO2 gradient (Delta-PCO2) were measured before and after vasopressin administration. Patients received a 50 mU/kg bolus vasopressin followed by 50 mU/kg/hour. If the response was insufficient, patients received a second bolus and continuous infusion was increased to a maximal dose of 100 mU/kg/h. Vasopressin infusion was tapered down by 10 mU/kg/h steps each 3 hours, when norepinephrine infusion decreased below 0.5 mg/kg/min. RESULTS. Seven patients aged 68  9 yrs completed the study. Three of them survived. MAP and SVRI increased in all patients, but delayed splanchnic hypercapnia developed leading us to reduce or withdraw the vasopressin infusion.

METHODS. In 7 barbiturate-anesthetized pigs we measured portal blood flow (QPORT; transit- time ultrasonic flowmetry) jejunal mucosal perfusion (laserDoppler flowmetry), tissue oxygen tension (PO2 TISSUE; microoximetry) and metabolic parameters. These measurements were performed at 3 predefined intestinal perfusion pressure (IPP) levels. At each IPP level, measurements were performed prior to and during PEEP (10 cmH2O) both with and without simultaneous dopexamine infusion (0.5 and 1.0 mg/kg/min). RESULTS. Within the perfusion pressure range 33±77 mmHg, intestinal perfusion and oxygenation were maintained irrespective of whether or not PEEP or dopexamine were applied. At severe hypotension (17±22 mmHg), QPORT and intestinal oxygenation deteriorated, resulting in regional net lactate production. At this IPP range, tissue oxygen perfusion was entirely pressure dependent and even small reductions in IPP led to prominent increases in intestinal net lactate production. CONCLUSION. Below the perfusion pressure range for effective autoregulation, intestinal tissue oxygen perfusion deteriorated and regional ischaemia occurred. In this situation, dopexamine was unable to counteract perfusion pressure dependent decreases in intestinal tissue oxygen perfusion produced by PEEP. A regional ischaemic threshold was described, with IPP lower than 35 mmHg associated with an intestinal tissue oxygen tension less than 45 mmHg. REFERENCES. Steinberg S, Azar G, Love R, Lee R, Choe E, Flint L. Dopexamine prevents depression of mesenteric blood flow caused by positive end-expiratory pressure in rats. Surgery (1996) 120, 597±601.

Norepinephrine [mg/kg/min] MAP [mmHg] SVRI [dyne s cm-5] Delta-PCO2 [mmHg]

Before

15 min

6 hours

12 hours

1.500.16

1.500.16

0.560.58*

0.740.58*

638 1396362 86

9410* 2399530* 1310

7210* 1752219* 1717*

7410* 2088527* 4856*

Meam  SEM, * < 0.05 CONCLUSION. Vasopressin increased peripheral vascular resistances. All patients developed an increase in Delta-PCO2, suggestive of decreased gastric mucosal perfusion Further studies assessing the effect of vasopressin on regional circulation are necessary before widespread use can be recommended. Meanwhile, close monitoring of gut perfusion by tonometry may be warranted. REFERENCES. 1. Tsuneyoshi I, Yamada H, Kakihana Y, et al. Hemodynamic and metabolic effects of low-dose vasopressin infusion in vasodilatory septic shock. Crit Care Med 2001, 29: 487. 2. Morales DL, Gregg D, Helman DN, et al. Arginine vasopressin in the treatment of 50 patients with postcardiotomy vasodilatory shock. Ann Thorac Surg 2000; 69: 102±6.

THE EFFECT OF NOREPINEPHRINE ON JEJUNAL MUCOSAL AND GLOBAL SPLANCHNIC PERFUSION

Oral Presentations Antibiotic therapy: New insights ± 16±20

Nygren A, ThorØn A, Ricksten SE. Department of Anaesthesiology and Intensive Care, Institute of Surgical Sciences, Göteborg, Sweden

16

14

INTRODUCTION. Norepinephrine is a commonly used vasopressor drug in the treatment of volume-resuscitated septic shock (1). It has both alpha and beta-adrenergic effects on the splanchnic vascular bed (2). The effect of norepinephrine on intestinal mucosal perfusion in man has so far not been studied. We have therefore evaluated the effect of norepinephrine on jejunal mucosal perfusion (JMP) by the use of laser Doppler flowmetry (3) in uncomplicated postcardiac surgical patients. METHODS. Nine male, mechanically ventilated postcardiac surgical patients aged 57±77 (average 67) years, with an ejection fraction > 45 % were included after informed consent. Postoperatively, using the nasogastric route and fluoroscopic guidance, we placed a two-probe Laser Doppler catheter (Perimed, Sweden) endoluminally in the proximal jejunum. We also catheterised the hepatic vein, the pulmonary artery and the femoral artery for blood sampling and haemodynamic profiles. A gastric tonometry catheter was placed for measurements of gastric mucosal pCO2. (Tonocap, Datex, Finland). Propofol was used for sedation (0,1  0,01 mg/kg/ min) to maintain a mean arterial pressure (MAP) of 70 mmHg. Norepinephrine was titrated (46  6 ng/kg/min) to a target MAP of 90 mmHg. Data on JMP were sampled during two consecutive 20±30 min periods of intestinal quiescence without (control) and with norepinephrine. Arterial and hepatic vein blood samples for lactate and oxygen saturation were taken at the end of each sampling period. A paired Student's t-test was used for statistical analysis. RESULTS. Norepinephrine increased MAP and systemic vascular resistance by 25±30 % (p < 0.0001), while cardiac index, central venous pressure, pulmonary capillary wedge pressure and mixed venous saturation remained unchanged. Splanchnic lactate and oxygen extraction as well as JMP and the gastric-arterial pCO2 gradient was not affected by norepinephrine. CONCLUSION. In postcardiac surgical patients, norepinephrine does not seem to jeopardise global splanchnic or intestinal mucosal perfusion. This might be due to a beta2-adrenergic stimulation antagonising alpha-adrenergic intestinal mucosal and global splanchnic vasoconstriction. REFERENCES. 1. LeDoux et al: Effects of perfusion pressure on tissue perfusion in septic shock. Crit Care Med Vol. 28,No.8: 2729±2732. 2. Reinelt H et al. Impact of exogenous betaadrenergic receptor stimulation on hepatosplanchnic oxygen kinetics and metabolic activity in septic shock. Crit Care Med. 1999 Feb; 27(2): 325±31. 3. ThorØn A, Elam M, Ricksten SE: Differential effects of dopamine, dopexamine, and dobutamine on jejunal mucosal perfusion early after cardiac surgery. Crit Care Med. 2000 Jul; 28(7): 2338±43.

THE APPLICATION OF BIOIMPEDANCE MEASUREMENT FOR THE ESTIMATE OF PHARMACOKINETICS OF ANTIBIOTICS IN SEVERE CAPILLARY LEAK SYNDROME

Balik M1, Kolµù M1, ƒediv™ J1, Hendl J1, Pachl J1. 1Anaesth. and Intensive Care, University Hospital Kralovske Vinohrady, Prague 10, Czechia INTRODUCTION. The authors were looking for corellation between distribution volume of the antibiotics assessed from the pharmacokinetic model and bioimpedance measured extracellular water (ECW) in septic mechanically ventilated patients with capillary leak syndrome. METHODS. 10 patients were treated with vancomycine (van) and 7 with netilmycine (net) more than 48 hours. The initial doses of the antibiotics were corrected according to the renal functions. Patients with presence of ascites, pleural effusion or treated with renal replacement therapy were excluded. Serum concentrations were measured before, 1 hour and 4 hours after the antibiotic infusion. Fast distribution compartment (V1), slow distribution compartment (Vd-area) and total body clearence (Cl) were measured. At the moment of the drawing of the third blood sample the total body multi-frequency bioimpedance analysis at 1, 5, 50 and 100 kHz was performed, the volume of ECW, TBW and ECW/TBW ratio were calculated. The protocol was repeated after 24 hours. RESULTS. Vd-area of both van and net correlated well with ECW (r = 0.53, p < 0.02, resp. r = 0.64, p < 0.03) as well as Cl of van and net with ECW (r = 0.68, p < 0.001,resp. r = 0.80, p < 0.002). Significant relationship was revealed for Vd-area of van and ECW/TBW (r = 0.76, p < 0.001) and for Cl of van and ECW/TBW (r = 0.69, p < 0.001). V1 of net correlated with ECW (r = 0.70, p < 0.02). The estimate of Vd-area of van using bioimpedance measured ECW is: Vd-area = 2.852*ECW-0.026. Vd-area of net can be estimated similarly: Vd-area = 3.626*ECW-0.248. In 9 patients (90 %) on van and in 4 patients on net (57 %) the toxic levels were reached. CONCLUSION. Bioimpedance measurement can help to estimate the increased V1, Vd-area as well as Cl of both drugs in severe sepsis with capillary leak syndrome. The correlation between the ratio ECW/TBW and Vd-area and Cl of van and between ECW and V1 of net represents probably better penetration of van into the tissues in comparison to net. The antibiotics are distributed poorly into the extravascular space and routine treatment may lead to significant overdose. The next steps should be directed towards monitoring of the tissue level of the antibiotics. REFERENCE. Patel RV, Peterson EL, Silverman N, Zarowitz BJ: Estimation of total body and extracellular water in post-coronary artery bypass graft surgical patients using single and multiple frequency bioimpedance, Crit Care Med,1996, 24: 1824±8

14th Annual Congress ± Geneva, Switzerland ± 30 September±3 October 2001

S 139

17

19

VARIOUS SYSTEMIC ANTIBIOTICS AND THEIR ANTINOCICEPTIVE EFFECT IN RATS

SHORT COURSE MONOTHERAPY FOR ICU-ACQUIRED GRAM-NEGATIVE BACTERAEMIA

Luger TJ, Geisler H, Farkas W, Lorenz IH. Dept of Anesthesiology and Critical Care Medicine, University of Innsbruck, Innsbruck, Austria

Corona A1, Wilson P2, Singer M1. 1Bloomsbury Institute of Intensive Care Medicine, 2Department of Clinical Microbiology, University College, London, United Kingdom

INTRODUCTION. Antibiotics are widely used in ICU patients for prophylaxis as well as treatment of infections. Systemic gentamicin produces mild antinociception in rodents (1,2). Less is known about antinociceptive effects of other antibiotics. The goal of this study was to examine the antinociceptive action of various randomly chosen antibiotics. METHODS. Sprague Dawley rats (male, 230±280 g; n = 6±8) received ciprofloxacin lactate (cipro), gentamicin sulfate (genta), amoxicillin (amox), vancomicin hydrochlorate (vanco) i. p. or saline intraperioneally (i. p.) in a study approved by the Animal Care Committee. Measurements were performed at baseline, 5, 15, 30, 60, 90 and 120 minutes after administration by using the tail-flick test (TF), hot-plate test (HP), catalepsy test (CT) and the sedation score (SD). Statistics: Wilcoxon Test and ANOVA for repeated measures; P < 0.05 = sign. RESULTS. The effect of submaximal doses of various antibiotics and saline is demonstrated in nociceptive tests (Table 1). Various antibiotics show a mild antinociceptive effect of short duration (5±15 minutes) in the TF and of longer duration (5±90 minutes) in the HP. In the dose-response curve, genta 0.1±8 mg/kg (max%MPE: HP = 23.8  4.4; TF = 33.7  5.0), amox 11±44 mg/kg (max%MPE: HP = 22.7  3.2; TF = 25.3  3.2) and cipro 1±64 mg/kg (max%MPE: HP = 18.7  2.6; TF = 38.4  7.0) showed a mild dose-dependent antinociceptive effect in both nociceptive tests. vanco 8±64 mg/kg (max%MPE: HP = 23.8  6.4; TF = 24.8  9.9) only in the HP. Rats did not show motor dysfunction or hyperreactivity; higher doses than used were excluded due to catalepsy. Baseline measurements did not differ between the groups. Saline and the solvent of cipro were ineffective.

INTRODUCTION. We undertook a prospective audit to assess the efficacy of our routine practice of short course monotherapy for treating ICU-acquired Gram negative bacteraemia [GNB] unless specifically indicated, e. g. endocarditis, osteomyelitis (1,2). This was judged by clinical response, relapse rate and patient outcome. METHODS. For a 6 month period from February 2000, all patients with clinically significant bacteraemia were identified by daily prospective surveillance of all positive blood cultures. For those developing GNB the following were recorded (i) micro-organism identification and antibiotic sensitivity pattern; (ii) type and duration of therapy (iii) patient clinical response, relapse rate and outcome. Microorganism isolation and identification (API 20E system Biomerieux,) was made by standard techniques and anti-microbial susceptibility tested by the Stokes disc diffusion method. RESULTS. 713 patients (455 male; 47 % surgical; median age 62 years Inter-Quartile Range (IQR) 45±72 y were admitted with a median ICU stay of 3 days (IQR 2±5). 143 (20 %) died in the ICU, 51 within 3 days. Twenty-six (3.6 %) patients developed 27 GNB episodes while in the ICU, with a median onset of 9 days (IQR 7±14). The major organisms isolated were Klebsiella spp. (14), Enterobacter spp. (7), Escherichia coli (3), and Pseudomonas spp. (3). The most frequent source was line-related (n = 8). Septic shock and severe sepsis occurred in 15 (56 %) and 7 (26 %) episodes, respectively. Only two (6.5 %) isolates were multi-drug resistant (MDR) and no extended-spectrum beta-lactamase producing microorganisms were found. Two patients were not treated as a therapy withdrawal decision was taken. A median 5-day course (IQR 4±7) of monotherapy (usually tazocin) was used for 20 (78 %) of the episodes. Clinical response occurred in 67 % of cases (of the total 20 treated with monotherapy) and relapse in only one (4.7 %). Four patients were given combination therapy for a median course of 8 days (IQR 5±12) because of severe deep-seated abdominal infections; all relapsed. For those treated with monotherapy, crude and bacteraemia-related mortality were 43 % and 19 %, respectively. Of the 4 patients treated with combination therapy, 3 died; all were directly attributable to the bacteraemia. CONCLUSION. Short course monotherapy produces a satisfactory clinical response and a low relapse rate in the majority of patients with ICU-acquired GNB. The low incidence of antibiotic resistance may also be related to this practice. Prospective studies are required to confirm this observation. REFERENCES. 1. Rello J., Ricart M, Mirelis B, et al. Nosocomial bacteremia in a medicalsurgical intensive care unit: epidemiologic characteristics and factors influencing mortality in 111 episodes. Intensive Care Med. 1994; 20: 94±98. 2. Ibrahim EH, Sherman G, Ward S, Fraser VJ, Kollef MH. The influence of inadequate antimicrobial treatment of bloodstream infections on patients outcomes in the ICU setting. Chest. 2000; 118: 146±55.

vanco64 amox44 cipro32 genta4 saline vanco64 amox44 cipro32 genta4 saline

Baseline TF

5 mins

15 mins

30 mins

60 mins

90 mins

2.360.2 2.140.18 2.290.13 2.30.23 2.70.08

2.50.39 2.830.18* 2.590.24 3.110.46* 2.790.15

2.610.17 2.690.11* 3.010.33* 3.480.49* 2.780.07

2.450.25 2.220.20 2.640.17 2.990.57 2.710.1

2.410.34 2.340.22 2.450.18 3.180.47* 2.750.12

2.360.19 2.340.2 2.870.23 3.030.45* 2.720.06

60 mins

90 mins

Baseline HP

5 mins

15 mins

30 mins

7.420.71 8.880.8 8.940.97 8.50.54 8.51.19

11.830.95* 10.831.14 9.890.79 15.131.01* 8.290.99

12.01.59 15.171.45* 12.330.88* 14.03.08* 8.861.68

13.171.74* 16.51.57* 12.561.69 14.03.81* 8.571.29

15.172.07* 13.51.34* 17.02.89* 15.832.1* 13.562.0* 13.562.79* 12.03.12 12.131.64 8.711.15 8.711.27

Table1: Effect of submaximal doses (in mg/kg i. p.) of various antibiotics in nociceptive tests CONCLUSION. We were able to demonstrate that cipro, genta, amox and vanco resulted in a mild dose-dependent antinociceptive effect. The clinical importance will be discussed. REFERENCE. Ocana&Baeyens,Neurosci Lett,1991.(2) Prado,Pain,1990.

18

20

CEFEPIME/CEFPIROME PHARMACODYNAMICS IN ICU PATIENTS: AUIC FOR RECOMMENDED DOSAGES

ANTIBIOTIC POLICY BASED IN A ROTATION PATIENT-TO-PATIENT

Laterre PF, Di Fazio V, Wittebole X, France MN, Reynaert MS, Wallemacq P. Critical Care Medicine, St Luc University Hospital, Brussels, Belgium INTRODUCTION. Beta-lactams serum levels well above minimal inhibitory concentration (MIC) are recommended to achieve rapid bacterial eradication, clinical cure and avoid resistance emergence. ICU patients (patients) with severe sepsis (SS) have often an increased distribution volume (DV) and are infected with less susceptible pathogens. Recommended 4 th generation Cephalosporins dosage might be inadequate in critically ill patients. METHODS. ICU patients (patients) with SS. were randomly assigned to receive Cefepime (CEP) or Cefpirome (CEF) ( + Amikacin) 2 gr bid or tid. Pts were divided in 4 groups (G) based on antibiotic dosage and measured creatinin clearance (CCl); G I (2 gr bid, CCl > 50), G II (2 gr bid, CCl < 50), G III (2 gr tid, CCl > 50), G IV (2 gr tid, CCl < 50). CEP and CEF serum levels were determined by HPLC (UV) at 0, 0.5, 1, 3, 6, 8 (tid) or 12 (bid) hours after antibiotic injection. MIC was determined for positive cultures. DV, antibiotic clearance (ACl), Area under the curve (AUC) and AUC/MIC (AUIC) were determined. AUIC for a MIC of 8 mg/l was simulated for all patients (AUIC8). Pts with AUIC < 125 are expressed in %. CEF and CEP data were pooled because of similar half life. RESULTS. 31 Pts were included in this study. Data are shown in the Table and expressed in mean  SD. DV was increased in most patients. ACl was well correlated with CCl. When CCl was > 50 ml/min, AUIC < 125 was observed in 16 and 22 % of the patients after 2 gr bid and tid respectively and partially explained by some resistant pathogens (MIC > 8). When AUIC was simulated for a MIC of 8 mg/l, 66 and 60 % of the patients were < 125 after 2 gr bid and tid respectively. For a CCl < 50, AUIC8 was > 125 for both dosages.

GI (n = 6) GII (n = 9) GIII (n = 10) GIV (n = 6)

Lorente L, Lecuona M, Mµlaga J, Delgado T, Mora ML. Intensive Care Unit, Hospital Universitario de Canarias, La Laguna, Spain INTRODUCTION. To evaluate ICU-acquired infections and sensibility to antibiotics by an antibiotic policy bassed in a rotation patient-to-patient and not for periods of time. We have compared our dates with EPIC study(1). METHODS. It is a prospective observational study in a 20-bed medical-surgical Intensive Care Unit. Were included all patients admitted in ICU during 3 periods of time (first period from 1±1-1999 to 30±7-1999, second period from 16±6-2000 to 16±10±2000 and third period from 1±1-2001 to 31±3-2001), with a lengt of stay in ICU longer than 24 hours. The infections were diagnosed according to the criteria of the CDC. RESULTS. Were studied 705 patients (371, 163 and 171 patients in respectives periods of time), 454 males and 251 females. Mean age was 57.61  16.85 years. APACHE-II was 13.4  6.2. Mortality was 12 ¢ 18 %. Patients distribution was: 45 % cardiac surgery, 14 % cardiologic, 11 % neurologic, 8 % traumathology, 8 % pulmonary, 5 % digestive, 9 % others. We diagnosed 315 UCIacquired infections in 182 patients. Distribution infections was: pneumonia 29 %, tracheobronchitis 11 %, urinary tract infection 30 %, central venous catheter 12 %, bloodstream 11 %, wound surgical 6 %, nervous central system 1. We isolated 322 germs: 46 % Gram-negative, 44 % Gram-positive, 10 % fungi. We found the following diferences between EPIC-HUC. Rate infections for pseudomonas aeruginosa 28 %±7 %, for acinetobacter 9 %±0 ¢ 3 %, for fungi 17 %±10 %. Rate resistance of pseudomonas to gentamicin 46 %±21 %, to ureidopenicillin 37 %±13 %, to ceftazidime 27 %±17 %, to ciprofloxacin 26 %±4 %, to imipenem 21 %±13 %. Rate resistance of Staphylococci aureus to methicillin 60 %±16 %. Rate resistance of coagulase-negative Staphylococci to vancomycin 3 %±0 %.

DV (l/kg)

ACl (ml/h/kg)

AUC (24 h)

AUIC ( < 125)

AUIC8 ( < 125)

CONCLUSION. An antibiotic policy bassed in a rotation patient-to-patient is able to control of infectious mape in our ICU.

0.460.3 0.570.25 0.440.24 0.81.06

12899 31.515 117109 40.425

757634 2280859 957471 33742790

16.6 % 37.5 % 22.2 % 20 %

66.6 % 0% 60 % 0%

REFERENCE. Vincent JL et al. Results of the European Prevalence of infection in Intensive Care (EPIC) Study. JAMA 1995; 274: 639±644.

CONCLUSION. In critically ill patients with sepsis and CCl > 50, CEF or CEF 2 gr bid or tid do not guarantee adequate AUIC in all patients. Fro a targeted AUIC > 125 in MIC of 8 mg/l, proposed dosages are inadequate in 2/3 of the patients. Higher CEF and CEP dosages or continuous infusion should be considered in ICU patients with severe sepsis.

14th Annual Congress ± Geneva, Switzerland ± 30 September±3 October 2001

S 140

23 Oral Presentations OF THREE DIFFERENT METHODS TO CONFIRMATION OF Initial management of trauma patients ± 21±25 RELIABILITY TUBE PLACEMENT IN EMERGENCY INTUBATION 21

Grmec S1, Piberl S1. 1Emergency Medical Service, Teaching Hospital Maribor, Maribor, Slovenia

THE SUCCESS RATE OF CARDIOPULMONARY RESUSCITATION DEPENDS ON THE QUALITY OF TEAM BUILDING Marsch SC, Marquardt K, Conrad G, Spychiger M, Eriksson U, Hunziker PR. Medical Intensive Care Unit, University of Basel, Basel, Switzerland INTRODUCTION. Extensive guidelines have been developed for a variety of medical aspects during cardiopulmonary resuscitation. Cardiopulmonary resuscitation is a team endeavour and several health care workers have to co-ordinate their actions in order to ensure adherence to the guidelines. The aim of the present study was to determine whether the success rate of cardiopulmonary resuscitation depends on behavioural aspects of team activity in a simulated cardiac arrest. METHODS. A patient simulator, consisting of a commercially available resuscitation mannequin was demonstrated at a workshop during an international congress of Intensive Care. Participants included registered nurses with a diploma in intensive care and physicians working in intensive care. Simulations were performed with teams consisting of 3 participants. The scenario employed was a witnessed cardiac arrest in intensive care. One participant played the role of the nurse in charge of the patient while the other two participants were summoned to assist. Cardiopulmonary resuscitation was rated to be successful if, within 2 min after the onset of the cardiac arrest, either defibrillation was performed or cardiac massage and mask ventilation were commenced. Simulation sessions were recorded on videotapes. After the simulation, the participants' theoretical knowledge of the algorithms of cardiopulmonary resuscitation was assessed. The quality of the medical performance and issues related to individual or team behaviour were analysed using the videotapes according to preset criteria. RESULTS. 16 teams with 3 participants were studied. According to the preset criteria, only 6 teams were successful while the remaining 10 teams failed. Within each team the participants shared among them sufficient theoretical knowledge on cardiopulmonary resuscitation to successfully handle the scenario presented.

Leadership Task distribution Information transfer Conflicts

Success (n = 6)

Failure (n = 10)

4/6 6/6 4/6 1/6

2/10* 4/10* 2/10* 6/10*

INTRODUCTION. The aim of our study was to evaluate three different methods for immediate confirmation of tube placement: auscultation, capnometry and capnography. METHODS. Tube position was initially evaluated by auscultation ( infraclavicular fossa,fifth intercostal space in the midaxillary fossa and over the epigastrium. Capnometry was determine with infrared capnometry BCI Capnocheck Model 20600A1, BCI International,waukesha,Wisconsin,USA. Capnography was determine with Propaq Encore, Model 200EL,Protocol Systems INC.,Beaverton,Oregon,USA. Determination of final tube placement was by direct visualization with laryngoscope.Analysis of the data included standard calculations of sensitivity and specificity and McNemar`s test for paired samples Chi- square and p value. RESULTS. Data were collected from February 1998 to Febryary 2001. We was tested 345 emergency intubations. Indications for intubations included cardiac arrest(n = 246;71 %) and nonarrest conditions(n = 99).11(3,1 %)of the 345 intubations in the study was an oesophageal tube placement.Results are showed in Table 1 and Table 2.Capnometry vrs capnography ± p < 0,01;capnometry vrs.auscultation ± p < 0,01 in cardiac arrest patients. Capnometry vrs.auscultation ± p < 0,05;capnography vrs.auscultation ± p < 0,05 in nonarrest patients. Method of verification

Sensitivity (%)

Specificity (%)

Negative predictive Value(%)

Positive predictive Value(%)

Auscultation Capnometry Capnography

98 80 100

100 100 100

99 38 100

100 100 100

Calculation of specificity,positive and negative predictives value for cardiac arrest patients

Method of verification Auscultation Capnometry Capnography

* = p < 0.05

Sensitivity (%)

Specificity (%)

Negative predictive Value(%)

Positive predictive Value(%)

94 100 100

100 100 100

93 100 100

100 100 100

CONCLUSION. Though sufficient theoretical knowledge of the algorithms of cardiopulmonary resuscitation was present in all teams, the majority of the teams failed to translate their knowledge into effective team activity. Thus, theoretical knowledge alone is insufficient to handle complex problems in highrisk domains like intensive care units. Failure to perform effective cardiopulmonary resuscitation was correlated with absence of leadership, absence of task distribution, lack of information transfer, and conflicts.

Calculation of sensitivity,specificity,positive and negative predictives value for nonarrest patients

22

24

PRE-HOSPITAL TRAUMA MANAGEMENT IMPLEMENTING THE PHTLSAND ATLS- GUIDELINES AFFECTS OUTCOME FOLLOWING ICU

THE SONOGRAPHIC ªDEEP SULCUS SIGNº: VALUE IN CRITICALLY ILL PATIENTS WITH PNEUMOTHORAX

Filos K S1, Fliggou F1, Sklavou C1, Zbouki A1, Vagianos C E2. 1anesthesiology And Intensive Care, 2surgery, University Of Patras, School Of Medicine, Rion ± Patras, Greece

Soldati G1, Iacconi P2. 1Emergency Medicine, General Hospital, Lucca, Italy, 2Department of Surgery, University of Pisa, Pisa, Italy

INTRODUCTION. ATLS is world-wide used, but the improvement of patient outcome following the ATLS- and PHTLS- guideline implementation has been documented only in developing countries (1). No studies have evaluated the impact of these guidelines on outcome following ICU admission after severe trauma in Europe.

INTRODUCTION. Diagnosis of a pneumothorax (PNX) on standard chest radiography requires identification of a radiolucent air space between the visceral pleural line and the lateral parietal pleura. However the reliability of the supine anterioposterior chest radiograph is not absolute and misdiagnosis may occur in up 40 % of all PNXs. The radiologic deep sulcus sign, present in a supine patient with PNX, represents air collecting anteriorly within the lower hemithorax, causing in chest X-ray a relatively radiolucent focus along the iuxtacardiac region and eventually extending into the lateral costophrenic recess. This is often a subtle and overlooked sign, expecially with small anterior retroparietal airpockets (1). The sonographic appearence of an isolate anterior air collection (occult PNX) has been described in on our previous work (2). Briefly, it consist in the disappearence of the gliding sign and in the lack of all vertical pleural artifacts (ring down) on a background of acoustic shadowing, along the pleural collapse. METHODS. We have studied in Emergency Room (ER) 65 critically ill patients with symptoms of dyspnea and pleuritic pain and clinical diagnosis of possible PNX. 47 were patiens with blunt thoracic trauma. Average age was 42 years (range 18±87) and 52 were male. In ER all patients underwent portable plain chest X-ray (Anterioposterior supine projection) and ultrasound chest examination. Longitudinal scans of the anterior, lateral and posterolateral (or posterior in the sitting patient) chest wall were taken using a Toshiba SSA250A portable unit equipped with a 3,75 MHz convex transducer. Subsequently, in Radiology Department, the primitive ER diagnosis (radiologic and sonographic) was validated in blind with the use of spiral Computerized Tomography. RESULTS. The diagnosis of PNX was confirmed in 22 patients (34 %). One subject (blunt trauma) showed bilateral PNXs. Six pneumothoraces were spontaneous. Urgent chest X ray showed 13 pneumothoraces (sensibility 56,5 %) while sonography was 100 % sensitive e specific for lung collapse. All the occult PNXs showed an exclusively anterior and inferior airpocket, not reaching the coronal mediothoracic line. This air collection, independently from its thickness, was readly visualized with the use of ultrasound, according to the morphological findings previously described. CONCLUSION. In a supine subject, a small PNX frequently has an isolate anterior retroparietal location. This pleural air collections (ªdeep sulcus signº in Radiology) are frequently overlooked in a supine ER plain chest radiography (43,5 %). In such cases, as well as with loculate pneumothoraces, CT scan can lead to their immediate identification. In our experience, however, chest sonography appear as a useful, economic and potent diagnostic tool for critically ill patients with a potentially occult pneumothorax. REFERENCES. 1. Knisely BL. Diseases of the pleura. In: Juhl JH, Crummy AB, Kuhlman JE Eds. Paul and Juhl's Essential of Radiologic Imaging. Lippincott Raven, Philadelphia, 1998, 1141±1169. 2. Soldati G, Rossi M. Pneumotorace traumatico: diagnosi ecografica in urgenza. G Ital Ecografia 2000; 3: 269±274

METHODS. All adult trauma patients (n = 1011, 68 % male, mean Injury Severity Score [ISS] 12.8  22.8) transported by ambulance to a third level university hospital between May 1997 and April 1998, were prospectively studied. Out of these, 140 patients (APACHE II = 18.8  7.1, ISS = 23.7  9.4) were transferred to the ICU. Univariate statistical analysis included ANOVA and Chi2. Logistic regression was used to calculate the effects of two predictive models: (a) the implementation of PHTLS and ATLS guidelines and (b) 45 ICU- treatment specific or patient related variables on outcome (30-day mortality). RESULTS. ICU mortality was 44.3 %. Polytraumatized patients and those with thoracic and spine trauma were not different distributed between the groups. However, the non-survivor group included more head-trauma patients, either as polytraumatized patients (75.8 % vs. 51.3 %, P < 0.01) or as severe head trauma (GCS < 8) alone (58.1 % vs. 38.5 %, P < 0.05) compared to the survivor-group, respectively. Table 1 summarizes the univariate analysis of prehospital and baseline characteristics of survivors vs. non-survivors. Logistic regression model (a) revealed only one factor: the delayed prehospital intubation (P < 0.05, ROC = 0.61). The second logistic regression model revealed the following variables: ISS (p < 0.001), age (p < 0.01), severity of head trauma (p < 0.01) and the duration of mechanical ventilation (p < 0.05), with an area under ROC = 0.87.

APACHE II ISS Transport Time (min) Pre-hospital O2: Necessary, not performed Pre-hospital ITN: Necessary, not performed Prehosp. cervical collar

Survivors (n = 78)

Non-survivors (n = 62)

P ± Value

16.76.6 21.08.97 83.551.9

22.36.2 27.57.21 81.744.1

< 0.001 < 0.001 ns

35.1 %

53.1 %

< 0.05

28.1 %

49.0 %

< 0.05

52.6 %

57.1 %

ns

CONCLUSION. The level of pre-hospital care and specifically the timely performed intubation have significant impact on delayed patient outcome following ICU transfer, especially, when trauma is complicated by head trauma, a high ISS- score and increased age. Education provided by ATLS might further improve ICU outcome after trauma. REFERENCE. 1. Ali J. et al.: J Trauma 1997; 42: 1018±21.

CONCLUSION. Our dat suggest that capnography is the most reliable method for detecting tube position. REFERENCE. Knapp S,Kofler J, Stoiser B et al: The Assessment of Four Different Methods to Verify Tracheal Tube Placement in the Critical Care Setting.Anesth Analg 1999;88: 766±770.

14th Annual Congress ± Geneva, Switzerland ± 30 September±3 October 2001

S 141

25

27

VALUE OF CT-SCAN IN PREDICTION OF PNEUMOTHORAX RECURRENCE

CUMULATIVE LOS-ICU UNDERESTIMATES MEDICAL WORKLOAD SPENT ON NON-SURVIVORS IN THE ICU AS INDICATED BY CUMULATIVE DAILY SOFA (CDSOFA) SCORES

Ouanes-Besbes L, Golli M2, Fkih-Hassen M1, Dachraoui F1, Boussarsar M1, Nouira S1, Gannouni A2, Abroug F1. 1ICU, 2Radiology, CHU F Bourguiba, Monastir, Tunisia INTRODUCTION. The usefulness of CT Scan findings in predicting pneumothorax recurrence is not well established. The aim of this study is to assess in patients with primary spontaneous pneumothorax (PSP) the incidence of CT-Scan revealed bullae and to determine if such lesions are predictive of recurrence. METHODS. All patients admitted from February 1995 to March 2001 for primary spontaneous idiopathic pneumothoraces were included. All patients were treated with pleural drainage by a thoracic tube or a small-bore catheter (Pleurocath*). Once discharged with no evidence of residual pneumothorax, thoracic CT-Scan was performed. The presence of lung parenchymal dystrophies (bullae and blebs) and their number were analysed and compared between patients in whom pneumothorax secondarily recurred and patients who remained recurrence free. The follow up was regular for all patients. RESULTS. 80 patients (78 men and 2 women, mean age: 27  8 years, SAPSII:7  3) were included. CT-Scan showed pathological lung changes in 58/80 patients (blebs in 10 patients, bullae in 27, bullae and blebs in 2). Within a mean follow-up of 33  20 months, 15 patients (19 %) experienced pneumothorax recurrence (ipsilateral in 14 and controlateral in 1). Neither clinical data nor CT-Scan findings were predictive of such recurrence (table).

Age (years) % initial collapsus CT scan findings Normal Bullae Blebs Bullae + blebs Drainage tubes Moderate-size chest tube Small-bore catheter

Recurrent pneumothorax (n = 15)

No recurrence (n = 65)

p

25 60

27 41

6 4 2 3

16 23 8 18

0.37 0.12 0.66

8

43

7

22

INTRODUCTION. The aim of this study was to analyse treatment efforts in ICU patients (LOS-ICU) and medical workload (CDSOFA) between hospital survivors and non-survivors. METHODS. All patients admitted to a 18 bed medical surgical ICU and treated > 48 hours in the unit were included in the study. Cumulative daily SOFA and LOS ICU were recorded in the period from September 1998 until March 2001. Hospital mortality was used as indicator of survival. RESULTS. 6622 scores were obtained during the studyperiod in 1187 patients. APACHE-II scores were lower in survivors than non-survivors (P < 0.001). Cumulative LOS was higher in survivors (P < 0.001), whereas median LOS was lower in this group (P < 0.001). Median CDSOFA was distinctly lower in survivors than in non-survivors (P < 0.001).

survivors non-survivors All patients

N=

mean score APACHE II

cum. LOS (days)

median LOS (days)

CDSOFA

median CDSOFA

855 332 1187

19 26 21

5749 3419 9168

3,6 5,0 3,9

30639 27031 57670

15 38 20

CONCLUSION. Non survivors consumed 47 % of CDSOFA points and 37 % of treatment days. If CDSOFA is used to proxy medical workload, this study indicates that cumulative LOS-ICU underestimates actual medical workload spent on non-survivors in the ICU. 0.38

CONCLUSION. CT-Scan revealed lung parenchymal dystrophies in more than 70 % of patients with primary spontaneous pneumothorax. We did not find any relation between recurrences and lung parenchymal dystrophies anatomical status (number, size and distribution of blebs/bullae) as assessed by CT-Scan.

Oral Presentations Emerging issues in modelling and evaluation of the critical patient ± 26±30 26 CALIBRATION AND DISCRIMINATION OF DAILY LOD SCORE IN PREDICTING HOSPITAL MORTALITY OF ICU PATIENTS: COMPARISON WITH SOFA SCORE (THE OUTCOME-REA DATABASE) Timsit JF1, Fosse JP2, Troche G3, Delassence A4, Alberti C5, Garrouste-Orgeas M6, Bornstain C7, Cheval C6. 1Hôpital Bichat, RØanimation mØdicale et infectieuse, 6Hopital St Joseph, 7Hôpital St Louis, 5 DØpartement de biostatistique, Paris, 3Hôpital BØcl›re, Clamart, 2Hôpital Avicenne, Bobigny, 4Hopital Louis Mourier, Colombes, France INTRODUCTION. LOD score should be used to compare ICU patients at ICU admission but has not been tested during ICU stay. However, to evaluate the over-risk of death induced by any nosocomial events, accurate organ dysfunction scores measurable during ICU stay are needed. Purpose: To evaluate the accuracy of daily LOD and daily SOFA scores in predicting hospital death in ICU patients. METHODS. Design: Prospective multicenter study. Setting: 6 intensive care units (ICUs) in France. Patients: 1685 ICU patients (511 deaths) were hospitalized for more than 48 hours. Median Age and SAPS II scores were 66 and 38 respectively. 881 were transferred from hospital ward, 1168 were medical. For each patients, a senior trained physician recorded daily variables needed to calculate organ dysfunction scores. RESULTS. SOFA and LOD scores were computed between day 0 (admission) and Day 7 (7 th calendarday). As the calibration of LOD and SOFA scores, at any time point, was not good (HL stat, p p < 0.001), crude scores were recalibrated. Prognostic value of daily customized LOD and SOFA scores were evaluated using patients hospitalized in ICU at least one more calendar-day. Models performances were assessed using the Hosmer-Lemeshow C statistic (Lower C values and higher p value is associated with better fit) and the receiver operating characteristics (ROC) area under the curve (AUC) (A ROCAUC of 1 is a perfect discrimination and the ROC-AUC of 0.5 is random chance).

Day1 Day2 Day3 Day4 Day5 Day6 Day7

Spronk PE, Bosman RJ, Oudemans-van Straaten HM, Van der Spoel JI, Zandstra DF. Intensive Care Medicine, Onze Lieve Vrouwe Gasthuis, Amsterdam, The Netherlands

LOD AUC ROC

LOD C-stat (p value)

SOFA AUC ROC

SOFA C stat (p value)

Alive/deaths n

0.726 0.739 0.753 0.752 0.737 0.754 0.728

9.33 (0.22) 12.7 (0.05) 13.7 (0.03) 5.6 (0.6) 10.4 (0.17) 6.8 (0.45) 8.2 (0.31)

0.727 0.743 0.756 0.754 0.738 0.754 0.730

9.4 (0.22) 12.4 (0.2) 13.2 (0.15) 6.3 (0.7) 13.7 (0.13) 9 (0.3) 5 (0.7)

1174/511 1097/473 908/428 738/387 605/344 492/319 406/294

CONCLUSION. When recalibrated in a particular population LOD and SOFA scores provided reasonable performances between day one and day 7. These scores should be used to compare severity of patients during the first week of ICU stay. (http://www.outcomerea.org) REFERENCE. Le Gall JR et al(1996) JAMA 276(10): 802±10

REFERENCES. 1. Knaus WA, et al. APACHE II: a severity of disease classification system. Crit Care Med 1985;818±829. 2. Vincent JL, et al. The SOFA (Sepsis-related Organ Failure Assessment) score to describe organ dysfunction/failure. On behalf of the Working Group on Sepsis-Related Problems of the European Society of Intensive Care Medicine. Int Care Med 1996; 22: 707±10

28 OUTCOME AND SOFA (SEQUENTIAL ORGAN FAILURE ASSESSMENT) SCORE AFTER CARDIOPULMONARY RESUSCITATION Janssens U, Dujardin R, Karassimos E, Graf J, Lepper W, Koch K, Ortlepp J, Hanrath P. Medical Clinic I, University Hospital of RWTH Aachen, Aachen, Germany INTRODUCTION. The SOFA score describes quantitatively the degree of organ dysfunction. The aggregate score (total maximum SOFA score [TMS]) is composed of scores from six organ systems respiratory (R), cardiovascular (C), hepatic (H), coagulation (Co), renal (Re), and neurological (N) graded from 0 to 4 according to the degree of dysfunction. We applied the SOFA score and TMS in patients (patients) with cardiopulmonary resuscitation (CPR) prior to admission and compared them to patients with acute cardiovascular disorders admitted to the same intensive care unit (ICU). METHODS. 813 patients (64  14 years, 69.7 %male, SAPS II 29  14) were included between 4/99 and 4/00. SOFA score was determined daily and TMS was calculated. All patients with CPR prior to admission were identified, neurologic impairment (NI) and hospital mortality (HM) were assessed. Simple / multiple logistic regression (LR) was performed to assess the relationship between SOFA, TMS / organ systems and HM and NI in patients with CPR. RESULTS. 62 (7.6 %) patients (62  13 years, 62.9 % male, SAPS II 44  18; initial rhythm: 44 [71 %] patients ventricular fibrillation, asystole 14 [22.6 %] patients, ventricular tachycardia 3 patients and torsades de pointes 1 pt) were resuscitated. Time to initiation of CPR was 3.5 (2.5±4.4 95 % confidence interval [CI]) minutes (min) and duration of CPR 26 (14±39 95 % CI) min. 28/62 (45.2 %) patients vs 109/ 751 (14.5 %) died (p < .001). Pts with CPR stayed longer in the ICU (11.4 vs 3.2 days, p < .001). Time to initiation of CPR and duration of CPR longer than 30 min was associated with NI (Risk Ratio (RR) 1.6[1.2±2.0 95 % CI] and 4.8[1.2±20.2]). NI (30/62 patients) was strongly associated with HM (risk ratio 28 (7±111). SOFA Score day 1 to 5 was significantly higher in patients with CPR as well as TMS (10  5.3 vs. 3.7  4.4, p < .001). SOFA score day 1 (RR 1.3 [1.1±1.4]) and TMS (RR 1.3[1.1±1.4]) were significantly linked to NI as well as HM (RR 1.4[1.1±1.6] and 1.2[1.1±1.4] respectively). Simple LR for SOFA score of each organ system on day 1 with HM and NI as the response variable: R (3.8[1.9±7.8], 2.2[1.3±3.7]), C (2.0[1.3±3.0], 1.9[1.3±2.9]), H (not significant [ns], ns), Co (ns, ns), Re (2.4[1.2±4.7], 1.9[1.0±3.7]), N (1.6[1.1±2.1], 1.6[1.2±2.2]). In multiple LR including all organ systems as the explanatory variables only R was independently associated with HM (3.3[1.4±7.7]) and C with NI (1.9[1.3±2.9]). TMSR (3.4[1.8±6.6]), TMSC (2.1[1.4±3.3]) and TMSN (1.5[1.1±2.0]) were associated with HM. Moreover TMSR (3.6[1.8±7.0]), TMSC (1.9[1.3±2.9]) and TMSN (1.8[1.2±2.6]) were linked to NI. In a multiple LR only TMSR was an independent predictor of HM (3.4[1.8±6.6]) and NI (3.6[1.8±7.0]). CONCLUSION. In a subgroup of cardiovascular patients with CPR prior to ICU admission coupled with a high mortality and poor neurologic outcome SOFA score and TMS were significantly higher compared to patients without CPR. Both SOFA score and TMS were associated with HM and NI. Liver and coagulation systems (day 1) did not predict HM and NI whereas respiratory, cardiovascular and neurological systems on day 1 as well as the maximum score were associated with outcome. Hence SOFA score and its derived variable TMS may serve well to describe the degree of organ dysfunction in critically ill post resuscitation patients. REFERENCE. Vincent JL et al. Intensive Care Med 1996;22: 707±710

14th Annual Congress ± Geneva, Switzerland ± 30 September±3 October 2001

S 142

29

30

MODELLING TOTAL COSTS IN ADULT INTENSIVE CARE UNITS: NEW MODELS FOR OLD QUESTIONS

A COST-EFFECTIVENESS STUDY OF ENTERAL IMMUNE MODULATING NUTRITION IN INTENSIVE CARE PATIENTS

Moran JL 1, Peisach AR 1, Solomon P 2. 1Intensive Care Unit, The Queen Elizabeth Hospital, Woodville, Australia, 2Department of Applied Mathematics, University of Adelaide, Adelaide, Australia

Coates E1, Hibbert C1, Edbrooke DL1. 1Medical Economics and Research Centre, Sheffield (MERCS), Royal Hallamshire Hospital (Intensive Care Unit), Sheffield, United Kingdom

INTRODUCTION. Advances in generalised linear modelling (GLM) have added to the repertoire of predictive cost models available beyond simple linear regression [1]. METHODS. Total patient ICU episode costs in three adult ICUs were calculated over a three month period and expressed as dollars (Australian). Costing used a ground-up approach with dedicated data-collectors recording daily activity / utilisation. Predictor models used admission day variables only, including APACHE III score, as judged by penalised likelihood. Models compared were: ordinary least squares [OLS], OLS with logarithmic cost transformation and Duan's smearing back transformation [OLSLS] and GLM with (i) gamma family, identity [GLMGI] and log link function [GLMGL] (ii) inverse gaussian family with identity [GLMII], log [GLMIL] and power [GLMIP] link. Analysis excluded re-admissions, deaths and cumulative activity indices (TISS). Model predictive ability was assessed by root mean square error (RMSE) and mean absolute error (MAE) and overall model fit by residual analysis [1]. RESULTS. The patient cohort (n = 1050) was of mean(SD) age 51(22) years and 59 % male. APACHE III score was 51(22) and 51 % were initially ventilated. Total costs were $6039(8804) per patient episode, range 242±70406, and exhibited significant (p = 0.000) kurtosis and (right) skewness. Significant cost predictor variables (p < 0.05) were APACHE III score, age and age-squared, ventilation status and source-hospital as a three level factor. A power relation, SD = mean^1.028 was demonstrated for costs and this relation was used in the GLMIP link function. Both OLS and OLSLS models demonstrated significant heteroskedasticity. Residual probability plots and plots of residuals against covariates and fitted values suggested that the GLM inverse gaussian identity [GLMII] and power links [GLMIP] demonstrated the best overall fit.

RMSE MAE

OLS

OLSLS

GLMGI

GLMGL

GLMII

8305 4676

8283 4676

8328 4669

8266 4672

8345 4641

GLMIL GLMIP 8330 4796

8369 4530

Predictive model comparisons CONCLUSION. Traditional cost prediction models using OLS and dependent variable transformations may be supplemented by appropriately specified generalised linear models which more closely model the error structure. REFERENCE. Diehr P et al. Ann Rev Public Health 1999;20: 125±44

INTRODUCTION. Patients admitted to an ICU are in, or at imminent risk of single or multiple organ system failure. Malnutrition is a major problem and patients' immune systems are generally compromised and resulting infections contribute to increased lengths of stay, mortality, morbidity and costs. There is growing clinical evidence emerging on the advantages of immunonutrition (IMN) in ICU patients. A recent meta-analysis of twelve RCTs [1] found that patients receiving IMN during their ICU stay experienced significant reductions in their number of ventilator days, infection rates and hospital length of stay, when compared to standard enteral feeds. The aim of this study was to estimate the cost-effectiveness of IMN in reducing infection in critically ill adult patients, from the perspective of the ICU. METHODS. Evidence of effectiveness was derived from a meta-analysis of twelve RCTs (1482 patients). The CONSORT statement [2] was used to assess the quality of the studies. Daily and total patient-related costs of care were calculated using an activity-based costing method, for a sample of 213 patients admitted to an adult general ICU over a ten-month period. The cost components included medical and nursing time, drugs, fluids, disposable equipment and medical imaging. The 213 patients were classified as to whether they were septic at any point during their ICU stay or non-septic. Length of stay and cost profiles were determined upon which the cost-effectiveness of IMN in reducing the duration of infection could be evaluated. RESULTS. The results of the meta-analysis found IMN to reduce infection by an average of 40 %, (CI: 14 %, 59 %). Thirty-six of the 213 patients studied (16.9 %) were diagnosed as septic and the added cost of treating this infection was estimated at £111.81 per day (£6151.21 per patient). If the number of days of infection were reduced by 40 %, the average cost per sepsis patient would reduce by £629.50 to £6928.37. When the cost of the immune-enhancing feed is included at £35.00 per day, the cost reduction was still £136.70. This calculation used an average length of stay of 14.08 days for the sepsis patients that changed to 8.45 days with infection and 5.63 days without infection from using IMN. CONCLUSION. Several factors can affect the relative cost-effectiveness of IMN such as the costs per day of infection, the number of days of infection and the cost of the nutrition itself. The results of our study suggest that the clinical benefits associated with the use of IMN have the potential to produce real cost savings in the ICU. If we accept the evidence on the effectiveness of IMN, further work should be directed towards rigorous evaluation of its cost-effectiveness. It is suggested therefore that future clinical trials on IMN incorporate an economic evaluation. REFERENCES. 1. Beale RJ, Bryg DJ, Bihari D. Immunonutrition in the critically ill: A systematic review on clinical outcome. Critical Care Medicine 1999;27: 2799±2805. 2. The Standards of Reporting Trials Group. A proposal for structured reporting of randomised controlled trials. JAMA 1994; 272:1926±1931.