INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

SEPTIC SHOCK IS ASSOCIATED with significant alterations in cellular oxygen utilization (1, 18, 26), even though oxygen delivery (O₂) to the tissues is typically maintained or even increased. Microregional zones of hypoxia (8, 23), secondary to microcirculatory disturbances (12), have been incriminated in these alterations, in addition to altered cellular metabolism. A complex interaction between an increased release of many mediators, leukocyte activation, endothelial injury, and interstitial edema has been largely implicated in this process (30) and may lead to multiple organ failure (22).

Red blood cell (RBC) entrapment may also participate in these microcirculatory abnormalities. Arterial hypotension can decrease microcirculatory RBC flow, resulting in a decrease in the perfused capillary density (12) and altered oxygen availability to the cells. Several studies have also demonstrated increased RBC stiffness in sepsis (6, 9, 16, 19), rendering the RBC less deformable for penetrating the microcirculation (13), a feature that could be explained by several mechanisms, including oxidation by oxygen free radicals (14), cellular energy depletion, and increase in intracellular calcium content (24).

Normovolemic hemodilution may result in beneficial rheological properties, leading to a better penetration of RBC into the microcirculation (10, 15, 17, 29). In a model of hemorrhagic shock, our laboratory previously reported (28) that tissue oxygen extraction capabilities during hemorrhage were greater when the hematocrit (Hct) was initially reduced. Whether similar results may be observed in sepsis, when microcirculatory alterations are present, has not yet been investigated. Therefore, we tested the hypothesis that normovolemic hemodilution could improve oxygen extraction capabilities in a canine model of endotoxic shock.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Surgical preparation. The study project was approved by the animal research committee of the School of Medicine of the Free University of Brussels. Eighteen mongrel dogs (22.7 ± 4.3 kg) were anesthetized with pentobarbital sodium at an initial intravenous dose of 30 mg/kg, followed by a continuous infusion of 4 mg · kg¹ · h¹ (pump Infusomat II, Melsungen, Germany) through the left forepaw vein. After endotracheal intubation with a cuffed endotracheal tube, each dog was mechanically ventilated with room air using a Servo ventilator 900B (Siemens-Elema, Solna, Sweden). Controlled ventilation was facilitated with pancuronium bromide given as an initial bolus of 0.15 mg/kg followed by an infusion of 0.075 mg · kg¹ · h¹. Respiratory rate was set at 12 breaths/min, and tidal volume was adjusted to obtain an end-tidal PCO₂ between 28 and 34 Torr. These ventilatory conditions were not changed thereafter. A right femoral arterial catheter was inserted and connected to a pressure transducer for arterial pressure monitoring. The right forepaw vein was cannulated for infusion of normal saline. A balloon-tipped pulmonary arterial catheter (model 93A-131-7-Fr, Swan-Ganz catheter, Baxter, Irvine, CA) was inserted through the right external jugular vein under guidance of pressure waves, as determined from a four-channel monitor (Sirecust 302 A, Siemens, Erlangen, Germany). A left thoracotomy between the fourth and the fifth intercostal space was performed, with bleeding controlled by electrocautery. Via a 2- to 3-mm incision in the anterior pericardium, a 16-gauge polyethylene catheter (Intracath, Deseret Medical, Sandy, UT) with multiple side holes around the tip was positioned in the pericardial space with its tip adjacent to the diaphragmatic surface of the left ventricle. The catheter was secured with purse-string sutures; 30 ml of warm sterile (37°C) saline were injected into the pericardial cavity to ensure that there was no leakage and was then drained before sealing. The thoracic cavity was then carefully closed in three layers, and a chest tube (Argyle, Trocar catheter A75, 28Ch-40 cm; Sherwood Medical, Tullamore, Ireland) was placed through the seventh intercostal space to allow gentle drainage of the chest. Through a midline laparotomy, a splenectomy was performed after maximal splenic contraction to 1-mg epinephrine (spread on the surface of the spleen) to prevent autotransfusion during hypotension. Ultrasonic flow probes were placed around the common hepatic artery (3-4 mm), the portal vein (10-12 mm), and the left renal artery (3-4 mm) for simultaneous measurement of blood flow in these vessels. In each dog, a multipurpose catheter (5F, Cook, Bjaerskov, Denmark) was inserted via the right jugular vein into the superior hepatic vein. Good position was confirmed by direct hand-feeling on the hepatic vein. A 16-gauge, 20-cm intravenous catheter (Argyle, Intramedicut, Sherwood Medical) was inserted via the splenic vein into the portal vein. A tonometer (TRIP, NGS catheter, Tonometrics, Helsinki, Finland) was placed in the lumen of the distal ileum via a small antimesenteric enterostomy and secured with a purse-string suture.

Experimental protocol. After surgical preparation, the dog was placed in the supine position and allowed to stabilize for 30 min. The dogs were randomly divided into three groups: Hct between 39.6 and 43.5% (Hct 40%, n = 6), between 27.6 and 32.1% (Hct 30%, n = 6), and between 21.0 and 24.5% (Hct 20%, n = 6). To achieve the three Hct levels, blood was slowly withdrawn from a femoral artery and simultaneously replaced by the same volume of a 6% hydroxyethyl starch (HES) solution (200/0.5; HAES-steril). The pericardial cavity was emptied using a 5-ml syringe to ensure a slightly negative intrapericardial pressure before the control measurements (B1) were obtained. The animals then received a slow intravenous bolus of 2 mg/kg Escherichia coli endotoxin (055:B5, control no. 3120-10-7; Difco, Detroit, MI), and a second set of measurements (B2) was obtained 30 min later. A normal saline infusion was then started and titrated to restore pulmonary occlusion pressure to baseline. A third set of measurements was obtained after 30 min (B3). The saline infusion was then kept constant at a rate of 20 ml · kg¹ · h¹ throughout the study. Cardiac tamponade was then induced by repeated injections of normal saline, heated to 37°C, into the pericardial sac. Measurements were repeated every 15 min thereafter in all animals, except the tonometry data, which needed 30 min of equilibration time. When the mean arterial pressure had declined to 20% of the baseline level, the dog was considered to be in a decompensatory state, the data collection was ended, and the dog was killed.

Measurements and calculations. All pressures were determined from a strip-chart recorder (2600S recorder; Gould, Cleveland, OH) at end expiration. Cardiac index (ml · min¹ · kg¹) was measured by the thermodilution technique (cardiac output computer, COM-2; Baxter) using three to five 5-ml injections of cold (<5°C) 5% dextrose in water. Each injection was started at end inspiration. A temperature probe was used on-line to control for variations in injectate temperature. Core temperature was continuously given by the pulmonary arterial catheter thermistor. During the study, core temperature was kept constant at its initial level with warming lamps. Regional blood flow was estimated simultaneously in the common hepatic artery, portal vein, and left renal artery by a previously calibrated blood flowmeter (model T208; Transsonic Systems, Ithaca, NY).

Statistics. In each animal, the determination of the whole body and liver critical O₂ (O_2 crit) was obtained from a plot of O₂ vs. O₂ using the method described by Samsel and Schumaker (20). O_2 crit was defined as the point of intersection of two best-fit regression lines as determined by a least sum of squares technique. Paired sets of linear regressions were calculated for all possible combinations of points separated into low (supply dependent) and high (supply independent) O₂ groups. Points were constrained to fall on either regression line but not on both. The pair of regressions with the lowest sum of standard errors of estimate was taken as the set that best fit the data. The values of O₂ and O₂ at the intersection point were then calculated using the two regression equations and were called O_2 crit and O_2 crit, respectively. As O₂ and O₂ were derived from independent techniques of measurement, oxygen extraction ratio at critical point (ERO_2 crit) was calculated by dividing O_2 crit by O_2 crit. An example is shown in Fig. 1. Statistical analysis included an ANOVA for repeated measurements followed by Dunnett's test. The difference in the slopes of O₂/O₂ was tested by an analysis of covariance. A P value < 0.05 was considered statistically significant. All values are expressed as means ± SD.

View larger version (23K): [in this window] [in a new window]   Fig. 1.   Relation between whole body O2 uptake (O2) and whole body O2 delivery (O2) in a hematocrit (Hct) of 40% in dog 6. Lines are dual-regression lines, and the point of intersection of these lines defines the whole body critical O2 (O2 crit) and the corresponding whole body critical O2 (O2 crit). ERO2 crit, critical O2 extraction ratio. See METHODS for explanation of Hct groups.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Hemodilution procedure. The amount of blood withdrawn was 320 ± 120 ml in the Hct 40% group, 470 ± 180 ml in the Hct 30% group, and 1,120 ± 310 ml in the Hct 20% group. As shown in Table 1, Hct levels at baseline were 41.4 ± 1.9, 30.8 ± 2.2, and 22.2 ± 1.8%, respectively. At baseline, cardiac index and regional blood flows were significantly greater in the Hct 20% group than in the Hct 40% group (Figs. 2 and 3). Nevertheless, systemic O₂ was significantly lower in the Hct 20% group than in the two other groups (Fig. 2).

View larger version (20K): [in this window] [in a new window]   Fig. 2.   Changes in mean arterial pressure (MAP), cardiac index (CI), and whole body O2 in relation to incremental changes in intrapericardial pressure (IPP) in the 3 groups of animals. B1, baseline; B2, 30 min after endotoxin; B3, 30 min after fluid resuscitation. Values are means ± SD. * P < 0.05 vs. Hct 40% group.

View larger version (20K): [in this window] [in a new window]   Fig. 3.   Changes in regional blood flow () during incremental changes in IPP in the 3 groups of animals. port, Portal vein; hep art, hepatic artery; ren, renal artery. Values are means ± SD. * P < 0.05 vs. Hct 40% group.

Effects of endotoxin. Endotoxin administration resulted in sharp decreases in arterial pressure, cardiac index, systemic O₂, and regional blood flows, whereas blood lactate levels increased (Figs. 2, 3, and 4, respectively). After initial fluid resuscitation, arterial pressure remained low, but cardiac index, systemic O₂, and regional blood flows increased and systemic vascular resistance decreased, reflecting a hyperdynamic state (Figs. 2 and 3; Table 2).

View larger version (22K): [in this window] [in a new window]   Fig. 4.   Changes in arterial lactate concentration and mucosal-arterial PCO2 gradient (PCO2 gap) in relation to incremental changes in IPP in the 3 groups of animals. Values are means ± SD. * P < 0.05 vs. Hct 40% group.

Effects of cardiac tamponade. Mean arterial pressure was similar in the three groups throughout the study (Fig. 2). Cardiac tamponade resulted in a progressive decrease in cardiac index and regional blood flows (Figs. 2 and 3). Cardiac index and portal, hepatic, and renal arterial blood flows remained higher in the Hct 20% group than in the other groups in the early phase but converged to similar values in the late phase of cardiac tamponade (Figs. 2 and 3). O₂ remained lower in the Hct 20% and Hct 30% groups than in the Hct 40% group (Fig. 2).

1

1

1

1

1

1

View larger version (16K): [in this window] [in a new window]   Fig. 5.   Individual (points) and mean (horizontal line) whole body O2 crit, O2 crit, and ERO2 crit in the 3 groups of dogs.

View larger version (11K): [in this window] [in a new window]   Fig. 6.   Individual (points) and mean (horizontal line) liver (Hep) ERO2 crit in the 3 groups of dogs.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The main finding of our study is that normovolemic hemodilution can increase oxygen extraction capabilities in endotoxic shock. Normovolemic hemodilution to a Hct of 20% was associated with an improvement in tissue oxygen extraction capabilities, both systemically and regionally in the liver, as indicated by a significantly lower O_2 crit and a higher ERO_2 crit in the lower Hct groups. The higher slopes of systemic O₂/O₂ relationships during the dependency phase in the lower Hct groups than in the Hct 40% group reflect this improvement in oxygen extraction capabilities. Van der Linden et al. (28) previously demonstrated in our laboratory that normovolemic hemodilution improves tissue oxygen extraction capabilities during acute hemorrhage in anesthetized dogs. The present study extended these observations to endotoxic shock, a situation in which microcirculatory alterations are more complex (12). The model used was slightly different because we used cardiac tamponade in lieu of progressive controlled hemorrhage to reduce cardiac output, but De Backer et al. (2) previously reported that the relation between O₂ and O₂ is similar in these two dog models.

Although we did not study the precise mechanisms involved, normovolemic hemodilution must have resulted in beneficial rheological properties, leading to a better penetration of RBC in the microcirculation. The principal effect of hemodilution on the microcirculation is an increase in blood velocity, which helps to maintain the RBC flux in the capillaries down to a Hct of ~25% (15, 17). Mirhashemi et al. (17) reported that RBC flux was not significantly altered in the subcutaneous tissue of the Syrian hamster skinfold when hemodilution reached 50% and systemic Hct was ~25%. Capillary RBC velocity increased by 60%, and capillary Hct decreased by 30%, with both of these changes being statistically significant. Using intravital microscopy, Lindbom et al. (15) studied the effects of acute normovolemic hemodilution with dextran on microvascular RBC flow in the tenuissimus muscle of the rabbit. Whereas the systemic Hct decreased from 50%, capillary Hct, measured by video densitometric methods, decreased by only 20%. A 45% increase in RBC velocity compensated for the decrease in capillary Hct so that the RBC flux, calculated from RBC velocity and capillary Hct, remained unchanged. It is thus well established that Hct can fall to values as low as 20% without a significant decrease in the RBC capillary flux.

A more homogeneous distribution of microcirculatory perfusion after normovolemic hemodilution also seems to be a key mechanism for the preservation of tissue oxygenation. Vicaut et al. (29) studied the effects of changes in Hct on the microcirculation. Even though the number of capillaries containing RBCs was not significantly influenced, the number of capillaries with no flow or low flow was reduced after hemodilution, suggesting a more homogeneous capillary perfusion. Arteriolar diameter was not significantly altered so that changes in vascular tone were unlikely to contribute. Hutter et al. (10) used radioactive microspheres to study blood flow distribution of skeletal muscle during normovolemic hemodilution in dogs. Although cardiac index increased during the procedure, O₂ to skeletal muscle was reduced to 74% of baseline. Nevertheless, tissue PO₂ was preserved. Heterogeneity of muscle perfusion (relative dispersion of perfusion) was reduced, again suggesting an homogeneous distribution of muscle perfusion after hemodilution.

Another factor operating to preserve tissue oxygenation during hemodilution may be a reduced oxygen loss by precapillary diffusion. Duling and Berne (3) showed that there can be a loss of oxygen from arterioles by precapillary diffusion either to surrounding tissue or to parallel veins in close proximity. This oxygen "leak" or diffusive shunt is determined by the PO₂ gradient between the vessel and the surrounding tissue or parallel veins. By shortening the transit time, an increase in RBC velocity may reduce the loss of oxygen before the capillaries and thereby improve oxygen transfer to the tissues.

The type of solution used in the normovolemic hemodilution procedure, and especially its molecular weight (MW), may influence erythrocyte rheology. The effects of HES on erythrocyte aggregation depend on their in vivo MW, as large molecules, such as fibrinogen, can form bridgelike structures between the erythrocyte membranes, whereas smaller molecules can displace the larger molecules and, therefore, decrease aggregation (25). Treib et al. (25) reported an increased erythrocyte aggregation for HES with large in vivo MW but a decrease in erythrocyte aggregation for easily degradable HES 200/0.5 or low-MW HES 70/0.5. Medium- or low-MW HES with a low degree of substitution resulting in low in vivo MW have better rheological properties (25). The 6% HES solution used in this study is a medium-MW HES solution, which is quickly split in vivo into smaller molecule sizes, resulting in a decrease in blood viscosity (5). This solution is commonly used clinically to induce normovolemic hemodilution for faster renal elimination and fewer adverse effects on coagulation than larger molecules (25).

Sepsis is typically characterized by alterations in microvascular perfusion due to capillary endothelial cell injury, leukocyte plugging of capillaries, and the development of interstitial or intracellular edema leading to a reduced perfused capillary density and an increase in the diffusion distance for oxygen (12). The reported decreased RBC deformability in sepsis (6, 9, 16, 19) can further impair cellular oxygen supply. In a model of experimental peritonitis in rats, Lam et al. (12) demonstrated a reduction in perfused capillary density, assessed by intravital microscopy. In these conditions, the rheological effect of hemodilution may be particularly beneficial, as shown in the present study. In a model of hyperdynamic endotoxic shock, Vallet et al. (27) also observed better tissue oxygenation after hemodilution. Endotoxin administration resulted in a shift to lower values of the frequency distribution of tissue PO₂ at the surface of skeletal muscle. After 30 min of resuscitation with dextran infusion, mean tissue PO₂ was restored close to the preendotoxin value, but, with continued resuscitation, reaching a Hct of ~17%, the mean tissue PO₂ even increased above the baseline value.

Normovolemic hemodilution may influence the distribution of blood flow to the tissues. Using the microsphere technique, Fan et al. (4) reported, during normovolemic hemodilution in dogs, a selective increase in blood flow to the myocardium and the brain but a stable blood flow to the liver, the intestine, and the kidney. In our study, there was no evidence of interorgan blood flow redistribution after normovolemic hemodilution because hepatic arterial, portal venous, and renal blood flows increased in proportion to the increase in cardiac index. Nevertheless, at the end of the cardiac tamponade, despite a very low-flow state leading to lower global O₂ in the Hct 20% and 30% groups, the PCO_2 gap was lower in these groups than in the Hct 40% group. A lower PCO_2 gap suggests a better balance between ileum mucosal oxygen supply and metabolism, which can here only be explained by a selective improvement in gut mucosal perfusion, and a subsequently better CO₂ removal. Similarly, Kleen et al. (11), during normovolemic hemodilution to a Hct level of ~20% in dogs, observed, using radioactive microspheres, a selective increase in gut mucosal perfusion, indicating a favorable redistribution of blood flow within the intestinal layers. We also observed, in the late stages of cardiac tamponade, lower arterial blood lactate concentrations in the Hct 20% and 30% groups than in the Hct 40% group. Even though hypoxia may not be the only cause of hyperlactatemia in endotoxic shock (7), in our model, the lower lactate concentrations in the lower Hct groups were likely explained by a better tissue oxygenation. Accordingly, a lower PCO_2 gap as well as lower lactate concentrations were likely to reflect improved tissue oxygenation in the lower Hct groups during the reduction in blood flow.

Our study demonstrates that normovolemic hemodilution down to Hct levels as low as 20% improves oxygen extraction capabilities during endotoxic shock and may even benefit tissue oxygenation. As long as cardiac output can be increased or sustained, Hct levels between 20 and 30% may be acceptable (or even desirable) in sepsis.