The conventional view is that fluid requirements are greatly increased in septic patients because of increased vascular permeability and more rapid loss of fluid from the intravascular space (1). No studies have quantified changes in the retention of crystalloid fluids within the vascular space in clinically relevant experimental models of sepsis.
In our laboratories we have developed a hyperdynamic model of sepsis in conscious, chronically instrumented sheep that replicates many of the hemodynamic features of clinical sepsis (2–4). We used the criteria for sepsis as defined by the Consensus Conference Committees of the Society of Critical Medicine, the European Society of Intensive Care Medicine, the American College of Chest Physicians, the American Thoracic Society, and the Surgical Infection Society (5,6). In this experimental model, sheep receive an infusion of live Pseudomonas aeruginosa (P. aeruginosa), which is associated with increased body temperature, increased heart rate (HR), and increased neutrophils in the setting of an infusion of live bacteria (7). These features meet the criteria for sepsis. The model does not produce septic shock, which would include evidence of tissue hypoperfusion such as lactic acidosis, although sepsis in this model has been associated with organ dysfunction, including renal failure (8), increased pulmonary transvascular fluid flux (9), and loss of hypoxic pulmonary vasoconstriction (10).
Clinical sepsis is typically hyperdynamic if adequate IV fluids have been provided for resuscitation (11). Hyperdynamic sepsis is characterized by increased core temperature, increased energy expenditure, peripheral vasodilation, increased cardiac output (CO) and decreased systemic vascular resistance (SVR) (11,12). In the experimental model of Pseudomonas sepsis used by our laboratory, CO increased and SVR decreased (13).
To quantify the influence of sepsis on the kinetics of infused crystalloid, we administered 0.9% saline by IV infusion to conscious sheep in which hyperdynamic sepsis had been induced by continuous infusion of live P. aeruginosa, as described previously (2,4). Serial measurements of blood hemoglobin concentration ([Hb]) were used to calculate plasma dilution, and from that we calculated the distribution of the infused fluid by two different approaches—mass balance analysis (14) and volume kinetic analysis (15–17). We hypothesized that infusion of 0.9% saline in conscious, chronically instrumented sheep with hyperdynamic, bacteremic sepsis would be associated with less plasma volume expansion (PVE) and greater interstitial fluid volume expansion than in conscious, nonseptic sheep.
The protocol for this study was approved by the Institutional Animal Care and Use Committee of the University of Texas Medical Branch at Galveston and adhered to the National Institutes of Health Guide for the Care and Use of Laboratory Animals.
Six adult female merino sheep were anesthetized with halothane in oxygen. A pulmonary arterial catheter (Swan-Ganz, Baxter, Irvine, CA) and bilateral femoral arterial and venous catheters (Intracath, Becton Dickinson, Sandy, UT) were inserted under sterile conditions. To eliminate the influence of a contractile spleen on [Hb], all animals underwent splenectomy through a left subcostal incision and the abdomen was closed in three layers. The catheters were connected to continuously flushed transducers. After emergence from anesthesia, analgesia consisted of buprenorphine 0.3 mg bid IM. The sheep were maintained in metabolic cages with free access to food and water for a minimum of 5 days of postoperative recovery. Twenty-four hours before the three experimental protocols, a urinary bladder catheter (Sherwood Medical, St. Louis, MO) was inserted and food and water were withheld.
All animals were heparinized with 3000 IU of heparin IV 5 min before each protocol. Conscious, chronically instrumented sheep weighing 42 ± 5 kg received 25 mL/kg of 0.9% saline over 20 min and blood was sampled every 5 min for a total of 180 min (“control”) (see Fig. 1 for a flowchart of the protocol). The 0.9% saline infusion (Baxter, Irvine, CA) was kept in a temperature range of 39°C to 40°C (via a warming coil and thermistor-regulated, temperature-controlled bath) and given through a femoral venous catheter over 20 min using a high-flow roller pump (Travenol Laboratories, Morton Grove, IL). After an interval of at least 24 h, an infusion of live P. aeruginosa (6 · 106 colony-forming units · kg−1 · h−1) was initiated and continued for the duration of the experiment. Four hours (“early sepsis”) and 24 h (“late sepsis”) after the Pseudomonas infusion was begun, sheep received another infusion of 25 mL/kg of 0.9% saline over 20 min and [Hb] was measured every 5 min for 180 min. Before each fluid infusion started, an Evans blue calculation of plasma volume (PV) was performed. Lactated Ringer’s solution was infused during the bacterial infusion at a rate of 2 mL · kg−1 · h−1, a rate of infusion that we have used empirically in this model to provide maintenance fluids after the onset of sepsis.
Hemodynamic variables, including HR and mean arterial blood pressure (MAP), were monitored continuously via a 4-channel hemodynamic monitor (Model 78304; Hewlett Packard, Santa Clara, CA). CO, pulmonary arterial pressure (PAP), central venous pressure (CVP), and pulmonary arterial occlusion pressure (PAOP) were measured using a thermistor-tipped pulmonary arterial catheter and a computer (9530; Baxter Edwards Critical Care, Irvine, CA). CO was measured hourly. The zero reference level for all hemodynamic data was set at 12 cm above the standing animal’s sternal plane. SVR was calculated hourly using the standard formula. Urinary output was measured every 5 min using a 250-mL graduated cylinder.
Hematocrit ([Hct]) and [Hb] were measured and recorded 3 times during baseline measurements and every 5 min during the 180-min duration of each experiment via analysis of 1 mL arterial blood samples (HemaVet; CDC Technologies, Oxford, CT). At the beginning of each protocol and at 60, 120, and 180 min after beginning the protocol, total plasma protein ([Prot]) was analyzed via a refractometer (Shuco; Tokyo, Japan) and plasma colloid osmotic pressure (COP) was analyzed using an oncometer (4100 Colloid Osmometer; Wescor, Logan, UT). Arterial blood gas and pH samples were withdrawn, analyzed (System 1302; Instrumentation Laboratory, Lexington, MA), and recorded three times during baseline measurements and hourly during the experiment.
PV was measured for each sheep before each protocol using Evans blue dye (14,18). An initial 45-mL arterial blood sample was taken before the first measurement as a standard. After rapid injection of 4 mL of Evans blue, 5-mL arterial blood samples were collected every 2 min for a total of 4 samples. Blood samples were centrifuged at 4500 rpm for 10 min and the plasma dye concentration was measured in a spectrophotometer at a wavelength of 805 nm (Spectronic, Milton Ray Company, Rochester, NY). The obtained values were fitted using least-squares regression to a logarithmic time decay curve of plasma dye concentration. Calibration standards were constructed for each animal from the plasma collected before dye infusion.
The dilution of arterial plasma was used to quantify PVE. For this purpose, repeated measurements of [Hb] in arterial blood were performed, as described above, at 5-min intervals for a duration of 180 min in each experiment, on a bedside [Hb] analyzer (HemaVet). Calculated expansion was corrected to account for the loss of 2 mL of blood for each sample withdrawn throughout the experiments, based on the assumption that baseline blood volume in sheep was 6% of body weight. The calculated PV at time (t) was as follows:
where [Hb]0 and [Hb]t represent [Hb] at the beginning of the infusion (0) and at each 5-min time interval, respectively. From the calculation of PVE at each timepoint, individual PVE curves were constructed for the entire 180 min of each experiment.
The distribution of the fluid given by IV infusion was analyzed using a two-volume kinetic model. In this model (15), fluid infused at a rate ki is distributed in an expandable space having a volume (v1) and communicating with a peripheral fluid space (v2). The net rate of fluid exchange between v1 and v2 is proportional to the relative difference in deviation from their baseline target volumes, V1 and V2, by a constant, kt. Elimination occurs at a baseline rate kb, which represents basal losses of fluid, and at a rate proportional by a constant (kr) to the deviation from the target volume, V1. kb was initially assumed to be 0.3 mL/min (19,20), but was further corrected, using a value of (–) 0.1 mL/min to account for flushing the catheters. The kinetic curves were modeled using Matlab version 4.2 (Math Works Inc., Natick, MA), whereby a nonlinear, least-squares regression routine is repeated until no parameter changes by more than 0.001 (0.1%) in subsequent iterations. The output of the kinetic analysis consists of the lowest possible squared average difference between the data points as predicted by the model and the data points actually measured, as well as the corresponding best estimate and the standard error (se) for each parameter in the model. The following differential equations describe the dilution changes in v1 and v2, respectively (Equations 2 and 3):
Because PV is assumed to constitute part of V1, PV and v are closely related and the dilution of arterial plasma can be used to indicate (v1– V1)/V1:
Kinetic analysis was performed on all experiments within each protocol as a group (i.e., data from all six animals were analyzed together). The best estimates of the model parameters V1, V2, and kt and their associated standard errors were obtained by fitting the mathematical solutions to Equations 2 and 3. To reduce the number of unknown parameters, the mean renal clearance (urinary excretion divided by the area under the dilution-time curve) for infused fluid was taken to represent kr used in each of the three group analyses. Hence, the analysis was made more robust by assuming that elimination occurred solely by renal excretion.
The three conditions (control, early sepsis, and late sepsis) were assessed for HR, MAP, CO, SVR, plasma dilution, PVE, cumulative urine, and calculated elimination rate constant (kr) at 180 min. Because of the uncertainty of homogeneity of variance and normality of random error terms, those measurements were analyzed using the Friedman test. The condition was assessed at the 0.05 level of significance. Data analysis was conducted using SAS®, Release 8.2 (SAS, Cary, NC).
The kinetic analysis is presented as a result of a nonlinear regression fitting of data in one single analysis and the parameters (V1, V2, and kt) are therefore given as best estimates with the corresponding standard errors of the estimates.
The animals tolerated all experimental procedures well, with the exception of one animal that died 20 h after induction of sepsis.
HR at the end of analysis intervals (at 180 min after beginning the infusions) in early and late sepsis was significantly more rapid than at 180 min in the control protocol (Table 1 displays all hemodynamic data). HR at the end of early sepsis was significantly more rapid than at the end of late sepsis. The three conditions were not significantly different at 180 min for MAP. CO at 180 min tended to be higher in the sepsis protocols, although the changes were not statistically significant. SVR at 180 min during early and late sepsis was significantly lower than at the end of the control interval. Temperatures at 180 min were significantly higher in early as well as late sepsis compared with controls. Although PAP and CVP tended to be higher and PAOP lower in the sepsis protocols, there were no significant differences at 180 min.
Arterial blood gases did not show any differences among the three protocols (Table 1). In both the early and late sepsis protocols at 180 min, [Prot] decreased significantly in comparison to the control [Prot]. Protein concentration was also significantly decreased at 180 min during late versus early sepsis. Plasma COP at the beginning and the end of the early septic interval was significantly less than at the beginning and the end of the control interval, respectively. Similarly, plasma COP at the beginning and the end of the late septic interval was significantly less than at the beginning and the end of the early septic interval, respectively (Fig. 2).
PV estimated by the Evans blue method were 1750 ± 114, 1678 ± 100 and 2128 ± 156 mL, respectively, at the start of the control, early sepsis, and late sepsis protocols. PV changes are shown both as dilution in percentage (Table 1) and PVE (by mass balance analysis) in mL (Fig. 3). At the end of the infusions (20 min into the protocols), PV dilution was 19% ± 3% in the control protocol, 21% ± 2% in the early sepsis protocol and 19% ± 3% in the late sepsis protocol. The corresponding PVE was 312 ± 50 mL, 386 ± 34 mL, and 400 ± 51 mL in the control, early sepsis, and late sepsis protocols, respectively (Fig. 3). Three hours after initiating the infusions, PV dilution was 6% ± 1%, 9% ± 2%, and 6% ± 2%. PVE at the end of the experiments was 97 ± 21, 151 ± 40 and 102 ± 44 mL, respectively (Fig. 3).
The volume kinetic analysis showed similar results at all three intervals (Table 2, Fig. 4 a–c). In each individual experiment, the two-volume model was statistically preferable to the one-volume model. The elimination rate constant kr did not show statistical differences, with a mean of 83 ± 16 mL/min in the control protocol, 109 ± 55 mL/min in the early sepsis protocol and 41 ± 13 mL/min in the late sepsis protocol. If one high outlier in the early sepsis group had been excluded, the mean would have been 55 ± 19 mL/min in that protocol.
Data are shown as best estimates and standard errors of the estimates as the result of a single nonlinear regression analysis of all experiments. Data marked by * were not results of the curve fitting but rather were calculated as mean renal clearances (urinary excretion divided by the area under the dilution-time curve) for infused fluid. One animal had an extremely high kr value at 180 min in the early sepsis group. Excluding this observation would have resulted in a value of 55 ± 19 in that group. The kr values of this animal in the control and the late sepsis states were normal. V1 and V2, central and peripheral body fluid space; kt, distribution rate constant; kr, elimination rate constant.
Neither mass balance nor volume kinetic analysis showed any differences in crystalloid distribution or elimination between conscious, nonseptic sheep and conscious, normovolemic, hyperdynamic, septic sheep. Theoretically, changes in capillary permeability during sepsis should reduce intravascular retention of a colloid bolus, whereas the influence of a bolus of crystalloid fluid is more difficult to predict. Again theoretically, decreases in COP could result in less retention of crystalloid fluid in the PV and greater loss into the interstitium. Crystalloid fluid is transported readily between the PV and interstitial fluid space in the normal, nonseptic state. Even if plasma proteins are extravasated more rapidly during sepsis, as the reductions in COP and [Prot] (Table 1 and Fig. 2) suggest happened in this experiment, the COP gradient between the intravascular and interstitial space apparently did not change sufficiently to alter the distribution of the crystalloid bolus.
The relevance of these observations to human sepsis requires discussion of the characteristics of this experimental model of sepsis. Models of experimental sepsis have become progressively more sophisticated and specific over the past two decades. Early experimental models of sepsis typically consisted of administration, usually to anesthetized animals, of large doses of endotoxin that produced circulatory shock. Models of progressive sepsis progressing to septic shock are also produced by cecal ligation and puncture, usually in rodents (21,22). Recently, our laboratory has preferred the experimental model of hyperdynamic sepsis produced by a continuous infusion of live Pseudomonas aeruginosa in conscious sheep. Both continuous endotoxin infusion and continuous Pseudomonas aeruginosa infusion in conscious sheep produce highly similar hemodynamic changes that reflect those in septic patients (but not patients in septic shock); in fact, responses to hemodynamic interventions in endotoxin-infused and Pseudomonas-infused conscious sheep have closely resembled responses in septic humans (23). The hemodynamic responses in the present study showed that these animals were hyperdynamic because HR and temperature were significantly higher and SVR was significantly lower in the early and late sepsis protocols compared with controls (Table 1).
Our failure to confirm our hypothesis with regard to fluids is surprising because the data seem contrary to clinical experience with fluid management in sepsis. Moreover, the data do not agree with the only previous study of volume kinetics in experimental sepsis (24). In the previous study, in which hypodynamic sepsis was induced in rabbits by a bolus of endotoxin, Svensen et al. (24) found that the central fluid space V1 was significantly smaller and that the apparent central retention of infused crystalloid solution was increased during sepsis. V1, which should be considered a functional rather than a physiologic or anatomic volume, is probably influenced both by the magnitude of PV and by the fraction of well-perfused tissues. In the previous study by Svensen et al. (24), the authors speculated that the forces acting to preserve central volume outweighed those that promoted diffusion of fluid to more peripheral fluid spaces. In this respect, the volume kinetics of crystalloid infusion in rabbits with endotoxic shock resembled the kinetics of crystalloid infusion in mildly and moderately hemorrhaged volunteers (16), in whom antecedent hemorrhage enhanced plasma dilution produced by a given volume of fluid. We speculate that the primary difference between the previous study in rabbits and the present study in sheep is that the model in rabbits was hypodynamic and produced relative hypovolemia.
These studies in septic sheep again illustrate the principle that IV-infused crystalloid fluids do not produce static responses. Rather, they produce an acute increase in intravascular volume that rapidly diminishes as fluid is redistributed and excreted. In addition, acute changes in intravascular volume alter the physiologic state and may secondarily generate compensatory responses that may be superimposed on underlying acute or chronic physiologic disturbances. Consequently, kinetic analysis of changes in PVE in response to fluid infusion under a variety of physiologic circumstances should enhance the clinical process of estimating fluid requirements.
The mass balance technique in this study was used to complement volume kinetic analysis by quantifying changes in PV. For this purpose, Evans blue dye was used to measure initial PV. The distribution of a dye or a tracer may be influenced by changes in capillary permeability. Presumably, errors induced by increased capillary permeability in estimation of PV using Evans blue dye, which is heavily protein bound, would result in measurements of PV that are artifactually low. Preservation in this model of PV during early and late sepsis argues against errors induced by increased capillary permeability to protein.
In contrast to mass balance analysis, volume kinetic analysis does not require determination of PV. Volume kinetics assesses the distribution of fluid based on the dilution of plasma. In this model, the body’s tendency to excrete infused fluid can be described as the renal clearance, kr. Because of inter-animal variations in kr calculated from the individual plasma dilution curves, we calculated a fixed kr for each protocol, based on actual urinary output (25). The fixed kr was reduced by late sepsis, although the change was not statistically significant; however, the magnitude of change was considerably less than the previously reported influence of isoflurane anesthesia on kr (26).
One important limitation of this study is the small number of animals, one of which died 4 hours before the late sepsis infusion. However, because the differences in kinetic and mass balance variables between protocols were small, even markedly larger groups would be unlikely to alter the conclusions from this study. Thus, in sheep, the distribution and elimination of an infused crystalloid is apparently not changed by an IV infusion of bacteria that induces systemic changes that fulfill the criteria for sepsis. Further investigation is required to determine whether this finding is clinically relevant in humans. Future experimental studies should extend these observations to severe sepsis and septic shock and should include analyses of the kinetics of colloid solutions.
In summary, it is apparent from this study that hyperdynamic, septic sheep were readily able to eliminate infused fluid while maintaining similar levels of PVE in response to fluid infusion. The elimination pattern of crystalloids was similar to that of nonseptic animals.
The authors wish to thank Jordan Kicklighter, BA, in the editorial office of the Department of Anesthesiology, The University of Texas Medical Branch, for manuscript preparation and review.
1. Marx G, Cobas MM, Schuerholz T, et al. Hydroxyethyl starch and modified fluid gelatin maintain plasma volume in a porcine model of septic shock with capillary leakage. Intensive Care Med 2002;28:629–35.
2. Fischer SR, Deyo DJ, Bone HG, et al. Nitric oxide synthase inhibition restores hypoxic pulmonary vasoconstriction in sepsis. Am J Respir Crit Care Med 1997;156:833–9.
3. Booke M, Hinder F, McGuire R, et al. Nitric oxide synthase inhibition versus norepinephrine for the treatment of hyperdynamic sepsis in sheep. Crit Care Med 1996;24:835–44.
4. Booke M, Westphal M, Hinder F, et al. Cerebral blood flow is not altered in sheep with Pseudomonas aeruginosa sepsis treated with norepinephrine or nitric oxide synthase inhibition. Anesth Analg 2003;96:1122–8.
5. Bone RC, Balk RA, Cerra FB, et al. Definitions for sepsis and organ failure and guidelines for the use of innovative therapies in sepsis. Chest 1992;101:1644–55.
6. Levy MM, Fink MP, Marshall JC, et al. 2001 SCCM/ESICM/ACCP/ATS/SIS International Sepsis Definitions Conference. Crit Care Med 2003;31:1250–6.
7. Lingnau W, McGuire R, Dehring DJ, et al. Changes in regional hemodynamics after nitric oxide inhibition during ovine bacteremia. Am J Physiol 1996;270:R207–16.
8. Hinder F, Booke M, Traber LD, et al. Nitric oxide synthase inhibition during experimental sepsis improves renal excretory function in the presence of chronically increased atrial natriuretic peptide. Crit Care Med 1996;24:131–6.
9. Hinder F, Booke M, Traber LD, Traber DL. Nitric oxide and endothelial permeability. J Appl Physiol 1997;83:1941–6.
10. Theissen JL, Loick HM, Curry BB, et al. Time course of hypoxic pulmonary vasoconstriction after endotoxin infusion in unanesthetized sheep. J Appl Physiol 1991;70:2120–5.
11. Kumar A, Haery C, Parrillo JE. Myocardial dysfunction in septic shock. Crit Care Clin 2000;16:251–87.
12. Kim PK, Deutschman CS. Inflammatory responses and mediators. Surg Clin North Am 2000;80:885–94.
13. Bone HG, Fischer SR, Schenarts PJ, et al. Continuous infusion of pyridoxalated hemoglobin polyoxyethylene conjugate in hyperdynamic septic sheep. Shock 1998;10:69–76.
14. Brauer KI, Svensen C, Hahn RG, et al. Volume kinetic analysis of the distribution of 0.9% saline in conscious versus isoflurane-anesthetized sheep. Anesthesiology 2002;96:442–9.
15. Svensen C, Hahn RG. Volume kinetics of Ringer solution, dextran 70, and hypertonic saline in male volunteers. Anesthesiology 1997;87:204–12.
16. Drobin D, Hahn RG. Volume kinetics of Ringer’s solution in hypovolemic volunteers. Anesthesiology 1999;90:81–91.
17. Hahn RG, Resby M. Volume kinetics of Ringer’s solution and dextran 3% during induction of spinal anaesthesia for caesarean section. Can J Anaesth 1998;45:443–51.
18. el Sayed H, Goodall SR, Hainsworth R. Re-evaluation of Evans blue dye dilution method of plasma volume measurement. Clin Lab Haematol 1995;17:189–94.
19. Riesenfeld T, Hammarlund K, Norsted T, Sedin G. The temperature of inspired air influences respiratory water loss in young lambs. Biol Neonate 1994;65:326–30.
20. Cox P. Insensible water loss and its assessment in adult patients: a review. Acta Anaesthesiol Scand 1987;31:771–6.
21. Williams DL, Ha T, Li C, et al. Modulation of tissue Toll-like receptor 2 and 4 during the early phases of polymicrobial sepsis correlates with mortality. Crit Care Med 2003;31:1808–18.
22. Qiu G, Gribbin E, Harrison K, et al. Inhibition of gamma interferon decreases bacterial load in peritonitis by accelerating peritoneal fibrin deposition and tissue repair. Infect Immun 2003;71:2766–74.
23. Booke M, Lingnau W, Hinder F, et al. Endotoxin versus bacteremia: a comparison focusing on clinical relevance. Prog Clin Biol Res 1995;392:393–403.
24. Svensen CH, Hjelmqvist H, Hahn RG. Volume kinetics of Ringer solution in endotoxinaemia in conscious rabbits. J Endotox Res 1997;4:425–30.
25. Hahn RG, Drobin D. Urinary excretion as an input variable in volume kinetic analysis of Ringer’s solution. Br J Anaesth 1998;80:183–8.
© 2005 International Anesthesia Research Society
26. Connolly CM, Kramer GC, Hahn RG, et al. Isoflurane but not mechanical ventilation promotes extravascular fluid accumulation during crystalloid volume loading. Anesthesiology 2003;98:670–81.