Isoflurane but Not Mechanical Ventilation Promotes Extravascular Fluid Accumulation during Crystalloid Volume Loading

The experimental protocol was reviewed and approved by the Institutional Animal Care and Use Committee of the University of Texas Medical Branch at Galveston, Texas, with adherence to the Guide for Care and Use of Laboratory Animals . 16Seven adult female merino sheep weighing 28 ± 4 kg (range, 22–35 kg) were studied.

All animals were surgically prepared in sterile operative conditions more than 5 days before the experimental protocols. Each sheep had a splenectomy performed under halothane anesthesia to remove the effects of splenic contraction and allow hemoglobin dilution to be used as a direct measure of PV expansion or contraction. 17The following indwelling catheters were inserted: bilateral femoral arterial and venous catheters (Intracath; Becton Dickinson, Sandy, UT) and a pulmonary arterial thermodilution cardiac output catheter (Edwards Lifesciences, Irvine, CA) through the internal jugular vein. After surgery, lactated Ringer's solution was continuously infused at 2.0 ml · kg⁻¹· h⁻¹, and the catheters were connected to continuously flushed transducers. After emergence from anesthesia, analgesia consisted of 0.3 mg buprenorphine via  intramuscular injection twice a day or as needed (Reckitt and Colman Pharmaceuticals, Richmond, VA). The sheep were maintained in large cages with free access to food and water during 5–7 days of postoperative recovery.

In four separate experimental protocols, the disposition of a 25-ml/kg, 20-min bolus of 0.9% saline was investigated under four protocols: conscious and spontaneously ventilating (CSV), conscious and mechanically ventilated (CMV), isoflurane-anesthetized and spontaneously ventilating (ISOSV), and isoflurane-anesthetized and mechanically ventilated (ISOMV). During the protocols, which were separated by more than 24 h, all animals were subjected to the same schedule of fluid administration, hemodynamic measurements, and blood chemistry measurements. The protocols were performed in random order except for the CMV protocol, which required a tracheostomy and was performed last to minimize animal discomfort and to reduce the risk of respiratory infections after the required tracheostomy.

Between protocols, the animals were returned to the central holding area. Temperature and fluid input and output were monitored hourly, and food and water were available ad libitum . When all the protocols were completed, the animals were euthanized by intravenous anesthesia with 1 ml xylazine at 20 mg/ml (Vetus Animal Health, Burns Veterinary Supply Inc., NY) and 10 ml ketamine (Fort Dodge Animal Health, IA) followed by 60 ml saturated KCl solution.

On the day before the first protocol, all vascular catheters were attached to pressure transducers (Baxter Pressure Monitoring kit; Baxter Healthcare, Irvine, CA). The sheep were instrumented with urinary bladder catheters (C.R. Bard Inc., Covington, GA), and food and water were discontinued at midnight.

On the morning of the study, the sheep were removed from the central holding area, and systemic hemodynamic variables, temperature, and urinary output were measured during a 2-h period. Heparin, 3,000 IU (Pharmacia and Upjohn, Kalamazoo, MI), was administered intravenously at the end of this baseline period. Blood chemistry and hemodynamic measurements were taken three times before giving the intravenous fluid bolus, which consisted of a 20-min infusion of 0.9% NaCl in a volume of 25 ml/kg through the femoral vein via  a calibrated roller pump (Cole-Farmer Instrument Co., Chicago, IL). During the protocol, blood samples and hemodynamic measurements were taken at 5-min intervals for the first 30 min (0–T30) and at 10-min intervals thereafter (T40–180). Heparin, 3,000 IU, was given intravenously again halfway through the experiment (T90) to maintain the patency of the catheters.

Preinfusion PV was measured for each sheep using indocyanine green (ICG) dye (Akorn Inc., Buffalo Grove, IL). 12,18,19After rapid injection of 12.5 mg ICG 5 ml, 5-ml arterial blood samples were collected every 2 min for a total of six samples. Blood samples were centrifuged at 4,500 rpm for 10 min (Dupont, Newtown, CT), and the ICG concentration was measured in the plasma from 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 ICG concentration. The calculated concentration of ICG at time zero was extrapolated from the decay curve and was representative of the plasma ICG concentration at the time of the dye injection with vascular mixing complete. ICG calibration standards were constructed for each animal from the plasma collected before dye infusion. Repeated measurements in individual sheep exhibited a small difference of baseline PV on different days.

For each blood sample, 10 ml blood was initially withdrawn into a syringe from the arterial catheter to avoid sample dilution. Then, a 1.0-ml arterial blood sample was taken for analysis of hemoglobin, hematocrit, and plasma protein concentration. One milliliter of the 10 ml was discarded to avoid infusion of clot or air, and 8.0 ml was reinfused into the arterial catheter, which was then flushed with 10 ml heparinized saline at a concentration of 3 × 10⁻³IU/ml heparin. Mean hemoglobin was measured twice (CO-Oximeter 482; Instrumentation Laboratories, Lexington, MA), and the average value was recorded. The remaining sample was spun for hematocrit analysis (Micro Capillary Centrifuge; Damon International Equipment Company, Needham Heights, MA). Hematocrit was recorded as the average of duplicate samples measured by a single operator on a microcapillary reader (Damon International Equipment Company). Plasma protein was measured using a clinical protein refractometer (Schuco, Japan) that had been calibrated for sheep plasma.

Heart rate (HR) and mean arterial pressure (MAP) were measured via  a pressure transducer connected to the femoral arterial catheter. Right atrial pressure (RAP), continuous cardiac output (CCO), and core body temperature (T) were measured from the pulmonary arterial thermodilution catheter. The zero reference levels for all hemodynamic data were set at 12 cm above the sternal plane. HR, MAP, and RAP were monitored on a hemodynamic monitor (Model 78534C; Hewlett Packard), and CCO and T were monitored on a continuous thermal dilution cardiac output computer (Baxter Healthcare Corporation, Edwards critical care division). Urinary output was accurately measured and recorded at 5-min intervals using an ultrasonic urine collection system (C.R. Bard Inc.).

In this protocol, baseline blood chemistry, hemodynamic measurements, and PV measurements were made, and the fluid bolus and measurement schedules were performed with the animal conscious and spontaneously breathing.

For this protocol, a tracheostomy was performed 24–48 h before the protocol. Under ketamine sedation (Fort Dodge Animal Health), a low-pressure, cuffed tracheostomy tube (Mallinckrodt Inc., St. Louis, MO) was inserted with aseptic technique. On the morning of the study, baseline measurements were taken, and mechanical ventilation was administered with a mechanical ventilator (Bear Medical Systems Inc., Riverside, CA) using the pressure-support mode with synchronized intermittent mandatory ventilation. Pressure support was maintained between 13 and 15 mm H²O, and synchronized intermittent mandatory ventilation was set at a frequency of 4–10 breaths/min and a tidal volume of 10–15 ml/kg. Fraction of inspired oxygen (Fio²) was maintained at 0.5, and peak flow was maintained in a range of 30–60 l/min. The assist sensitivity function on the ventilator was set to minimize the effort required to initiate pressure support. In one sheep that tolerated synchronized intermittent mandatory ventilation poorly, i.e. , coughed and strained excessively in response to mandatory artificial breaths, the pressure-support mode alone was used. Baseline measurements were taken when the sheep were comfortable and stable on mechanical ventilation. Subsequently, the fluid bolus was administered, and measurements of PV, hemodynamics, and blood chemistries were made.

In the third protocol, the sheep were anesthetized and spontaneously ventilating. Anesthesia was induced with 2.5% intravenous thiopental at a dose of 0.66 ml/kg (Abbott Laboratories, N. Chicago, IL). The animals were intubated with a cuffed endotracheal tube (Mallinckrodt Medical Inc.), and placement was assessed by auscultation and capnography (Datex–Engstrom Div., Helsinki, Finland). Anesthesia was maintained at 1.5% isoflurane (Abbott Laboratories) or a minimum alveolar concentration (MAC) of 1.0 for sheep. 20,21When a delay in the onset of spontaneous ventilation occurred, manual ventilation was performed to deliver isoflurane via  the ventilator's reservoir. When stable spontaneous ventilation under anesthesia was achieved, a second set of preinfusion measurements were made, the fluid bolus was administered, and measurements of PV, hemodynamics, and blood chemistries were made.

This fourth protocol was performed when the animals were anesthetized and mechanically ventilated. After induction and intubation, isoflurane was delivered as in the ISOSV protocol. Ventilation was mechanically controlled using an anesthesia ventilator (American Drager, Telford, PA). The tidal volume was set at 15 ml/kg, and the respiratory frequency was set initially at 10 breaths/min, then adjusted to maintain an end-tidal carbon dioxide (ETco²) of 35 mmHg, with an Fio²set to 50%. When a MAC of 1.0 had been achieved, baseline variables were measured, the fluid bolus was administered, and measurements of PV, hemodynamics, and blood chemistries were made.

Initial (preinfusion) plasma volume (PVⁱ) was measured using ICG dilution. 19,22,23In the following equation, initial blood volume (BVⁱ) was derived from PVⁱ, and hematocrit. The calculation of BVⁱrequired correction for known errors in large vessel hematocrit (F cell ratio, 0.92) and the correction for plasma trapped in the erythrocyte column of the hematocrit tube (0.95).

Initial erythrocyte volume (RBCVⁱ) was calculated as the difference between PVⁱand BVⁱ:

Serial erythrocyte volumes (RBCV^t) were calculated in the equation below, accounting for blood volume withdrawn during sampling (S) and changing hematocrit values. The same correction factors for hematocrit were used as in the initial calculation of blood volume.

Serial PVs were calculated from previous blood volumes (BV^t-1) subtracting previous erythrocyte volume.

This equation is based on the calculation of blood volume expansion, which is proportional to the changes in arterial hemoglobin concentration.

The change in PV or PV expansion is calculated as:

The application of mass balance provides further information based on the assumption that at any time, the infused volume is equal to the sum of PV expansion, urinary excretion, and extravascular expansion or contraction. We assume that the extravascular expansion due to an isotonic crystalloid is entirely interstitial for an isotonic fluid that contains a sodium concentration similar to the baseline plasma sodium concentration, as we previously reported using an extracellular tracer. 24Serial net change in interstitial fluid volume (ΔISFV) was calculated as the difference between fluid infused (F^t) and the sum of cumulative urinary output (ΣUO) and net change in blood volume (ΔBV).

Volume expansion efficiency (VEE) was calculated as net change in PV divided by cumulative fluid volume infused (ΣF^t).

The distribution of the fluid administered by intravenous infusion was analyzed by using a one- or two-volume fluid-space model as previously described. 12,14,25In this kinetic model, the fluid is given at a rate kⁱinto a central fluid space having a baseline volume V (one-compartment model) or V¹(two-compartment model). The expanded space after infusion is termed v or v¹. For the two-compartment model, v¹communicates with a peripheral fluid space with a baseline volume of V²and an expanded volume of v²(equations 8 and 9and fig. 1). The net rate of fluid exchange between v¹and v²is proportional to the relative difference in deviation from their baseline values V¹and V²by a constant (k^t). Elimination from the kinetic system occurs at a baseline rate k^b, which represents basal losses of fluid and was assumed to be 0.3 ml/min (430 ml/24 h) 26,27(fig. 1). The renal fluid losses occur from v¹and are estimated as the product of a “renal” constant k^rand the deviation from the target volume (v − V)/(V). The following differential equations describe the dilution changes in v¹and v², respectively.

The model was further corrected to account for flushing the catheters and the loss of fluid and hemoglobin due to blood sampling not shown in the equations. As the plasma is a part of v¹, the dilution of the arterial plasma was used to indicate (v¹− V¹)/V¹. Hence:

The best estimates of the model parameters V¹, V², k^r, and k^tand their associated SDs were obtained by fitting the mathematical solutions to equations 8 and 9, which have previously been described, 12,14,25to each data set separately by using a nonlinear least-squares regression routine programmed in Matlab version 4.2 (Math Works Inc., Notich, MA). In addition to the k^restimated by this curve-fitting procedure, a theoretical k^rwas calculated based on the urinary excretion divided by the area under the curve for the plasma dilution curve. 28 

In the volume kinetic analysis of each experiment, a statistical F test was applied to examine whether the use of a biexponential function (two-volume model) was statistically preferable to a monoexponential (one-volume model). 28If not, the results were presented according to a one-volume model, where the resulting parameters were estimates of V and k^r. When a series of experiments is presented graphically, the curve is the weighted average of the two curves obtained for the experiments in which the one-volume and two-volume models, respectively, were statistically justified.

In the indicator dilution and mass balance analysis, the measured variables in the four protocols were compared at baseline (time zero) and at 20, 60, 120, and 180 min and were described as mean ± SEM. Statistical comparisons were made using a two-way analysis of variance with a post hoc  Tukey test for pair-wise multiple comparisons. In volume kinetic analysis, mechanical ventilation and isoflurane were tested as predictors of outcome and data were presented as median and 25th and 75th percentiles. Variables in the fasting study were compared by the Student paired t  test. Significance was accepted at P < 0.05.

Seven experiments were completed in the CSV, ISOSV, and ISOMV protocols, and six were completed in the CMV protocol; one animal did not complete the CMV protocol because of the development of a respiratory tract infection. Mean infused volume was 25 ml/kg or 728 ± 15 ml.

Repeated measurements in individual sheep showed little variance of baseline PV between protocols. Mean baseline PVs were 51.5 ± 3.6, 53.8 ± 4.8, 48.5 ± 4.2, and 49.1 ± 4.4 ml in the CSV, CMV, ISOSV, and ISOMV protocols, respectively (P = not significant). Mean blood volumes were also similar between groups and averaged 66.2 ± 5.6 ml/kg.

Vascular expansion is directly observed in the serial samples of blood, the dilution of hemoglobin and plasma protein, and the calculated changes in PV (ΔPV) at selected time points from equations 4 and 5. Figure 2shows the detailed time course of ΔPV both during and after infusion in all four groups. Only small, statistically insignificant differences in PV expansion were present between the protocols (fig. 2). PV increased significantly during the 20-min infusion and remained greater than baseline for the first 2 h of the experiment (P < 0.001). Immediately after infusion (T20), the VEE (milliliter expansion/milliliter infused) of the LR bolus in the CSV protocol was 0.40, i.e. , 40% of the 25 ml/kg infused load remained in the intravascular space at the immediate end of the infusion. The rapid increase in PV during the 20-min infusion was followed by a rapid decline in PV until T50 or 30 min after infusion (VEE = 0.14) and then was followed by a slower phase of declining PV until the end of the experiment, T180, at which time VEE was 0.04.

The mean results (± SEM) of indicator dilution and mass balance analysis and statistical comparisons of protocol differences for PV (ΔPV), urinary output (ΣUO), and interstitial fluid volume (ΔISFV) following a 25-ml/kg fluid bolus for selected time points are shown in table 1. That is, the time course of the relative distribution of the infused volume over time with respect to changes in the vascular volume, the interstitial volume, and the urine volume is shown. These variables are plotted for all time points as area graphs (fig. 3), providing a visual comparison of the three fluid spaces versus  time for all four protocols.

For all time points analyzed, ΣUO was significantly reduced by ISO regardless of ventilatory pattern (P < 0.001), but ΣUO was not reduced by mechanical ventilation. Under isoflurane anesthesia, only 12 ± 4% of the infused load had been excreted at T180 as compared to 63 ± 10% in the conscious protocols.

A fluid bolus increased extravascular fluid accumulation, assumed to be ΔISFV, significantly more under isoflurane anesthesia than in the conscious state (P < 0.001), whereas mechanical ventilation did not affect the ΔISFV after a fluid bolus. ΔISFV was comparable in all protocols at T20, the end of the infusion (range, 11–13 ml/kg), but at 180 min, 89 ± 6% of the infused fluid remained in the interstitial fluid space in the anesthetized protocols compared with 33 ± 6% with the conscious protocols.

Three adaptations of volume kinetic analysis were applied to these data sets. The first adaptation was a “split k^t”method in which a two-volume model was applied to quantify the accumulation of edema in V²during the isoflurane protocols. k^rwas determined by the measured urinary excretion, divided by the area under the plasma dilution curve. Peripheral accumulation of fluid could then be indicated as a difference between k^tfor the translocation of fluid from V¹to V²and another k^tgoverning the flow occurring in the opposite direction. This model was abandoned, however, since it consistently yielded poor curve fits.

In the second curve-fitting procedure, k^rwas determined as the total measured urinary excretion divided by the area under the plasma dilution curve. Losses of fluid to an extravascular space in the body were modeled by letting loss from the kinetic system equal a model-determined constant, k^b, as determined by a best fit. This was instead of having k^bfixed at 0.3 ml/min to account for evaporative losses. This model yielded good curve fits with small, evenly distributed residual errors (figs. 4–6).

The parameter k^bindicated that very limited transfer of fluid out of (or into) the kinetic system occurred in the absence of isoflurane (table 2).However, the model simulation was associated with a significant transport of fluid out of the kinetic system for the isoflurane protocols with k^bequal to 3.2 and 4.2 ml/min during spontaneous ventilation (ISOSV) and during mechanical ventilation (ISOMV), respectively. Calculated fluid losses over 180 min and normalized by body weight were 20.2 ± 0.5 and 26.5 ± 0.3 ml/kg in the ISOSV and ISOMV protocols, respectively, at the end of the 180-min study. The two-volume-of-fluid-space model resulted in a statistically lower squared difference between the theoretical and experimental data points in 22 of 27 experiments. In the other 5 of 27 experiments, fluid was rapidly eliminated, and the one-volume-of-fluid-space model fit the data better (table 2).

In the third curve-fitting procedure, the extravascular accumulation of fluid was quantified by comparing the model-predicted rate of elimination with the measured urinary excretion. This was done by using a fixed value of 0.3 ml/min for k^bwhile calculating k^rusing the kinetic model. The amount of eliminated fluid was then calculated as the product of k^rand the area under the plasma dilution curve. In the absence of isoflurane, the comparison showed good agreement between the model-predicted elimination and the measured urinary excretion; however, during isoflurane anesthesia, model-predicted elimination and measured urinary excretion did not agree. Isoflurane was associated with fluid losses out of V¹of between 560 and 600 ml, which were not recovered as urine (table 3).

As shown in table 4, MAP was not significantly altered after infusion but was greater in both anesthetized protocols (ISOSV and ISOMV) than in the control group (CSV) (P < 0.001 and P = 0.002, respectively). MAP was also lower in the CSV protocol than in the CMV protocol; however, this difference was not statistically significant. Cardiac output increased during the infusion in all protocols and was 150% baseline at T20 and then returned toward baseline. HR was reduced during the infusion in all protocols. HR was significantly higher during the anesthetized protocols (123 ± 8 and 116 ± 6 beats/min in ISOSV and ISOMV, respectively) than during the conscious protocols (95 ± 6 and 102 ± 15 beats/min in CSV and CMV, respectively;P < 0.001). RAP increased during the infusion but was significantly lower during the ISOSV protocols than in either the ISOMV protocol or either of the conscious protocols (P < 0.001, P < 0.001, and P = 0.006, respectively). RAP in the CMV protocol was significantly greater than in the ISOSV and ISOMV protocols (P < 0.001 and P = 0.03, respectively). Therefore, isoflurane tended to decrease RAP in this animal model, while mechanical ventilation tended to increase it. Total peripheral resistance decreased following the infusion, but this change was not different among the protocols. Baseline core body temperature was 39.4 ± 0.1 in the two conscious protocols and was 0.6–1.0° lower in the anesthetized protocols than in the conscious protocols, but there was a significant difference only between the CMV and ISOMV protocols (P < 0.04).

In an attempt to minimize the vasodilatory effect of hypercapnia, 29a target ETco²level of 35 mmHg was achieved in both the CMV (35.9 ± 2.8 mmHg) and ISOMV (35.5 ± 4.7 mmHg) protocols. The mean ETco²in the ISOSV protocol was 54.3 ± 4.2 mmHg (range, 40–70 mmHg). Spontaneous ventilation following induction of anesthesia in the ISOSV protocol was achieved within 7–10 min after induction.

These data demonstrate that isoflurane anesthesia markedly influences the distribution and elimination of infused fluids and suggests that some of the tendency of fluid to accumulate perioperatively can be attributed to the influence of anesthesia. This study amplifies the recent finding of a marked reduction of the urinary excretion of a fluid load of 0.9% saline when infused during isoflurane anesthesia and mechanical ventilation in sheep. 12We performed a similar fluid bolus study in four experimental protocols to investigate the hypothesis that the different fate of a fluid bolus during general anesthesia was attributable to mechanical ventilation rather than to isoflurane. Our hypothesis was based on reports that positive pressure associated with mechanical ventilation can alter circulating blood volume and peripheral fluid accumulation by impeding venous return and by changing (increasing or decreasing) cardiac output. 10,11Our data disprove our hypothesis.

The technique combining indicator dilution and mass balance and the technique of volume kinetic analysis were both applied to analyze alterations in body fluid handling caused by mechanical ventilation or isoflurane anesthesia. Both techniques were based on frequent serial measurements of hemoglobin dilution and urinary excretion during and after the infusion. Indicator dilution and mass balance demonstrated that isoflurane but not mechanical ventilation was associated with a significant decrease in urinary excretion and a corresponding increase in interstitial fluid volume. No statistically significant differences in the time course of PV expansion or VEE between the protocols could be discerned by the indicator dilution technique.

Volume kinetic analysis was performed to provide more insight into the fate of the extravascular fluid losses. Volume kinetic analysis in conscious humans has shown that the elimination constant k^rcorresponds closely to urinary excretion. In healthy, conscious volunteers after blood loss, both k^rand urinary excretion were reduced. 30Patients who had sustained hip fractures had volume elimination coefficients that were 50% of that of controls, which is in keeping with posttraumatic fluid retention. 31,32In the sheep study 12that prompted the current experiment, we found that isoflurane anesthesia and mechanical ventilation were associated with reduced urinary output in sheep and that the kinetic variable k^rdid not correspond to reduced urinary output, but rather to the sum  of urinary output and extravascular fluid accumulation. Therefore, in the kinetic analysis of the current study, we used a fixed k^rvalue in two of the kinetic models.

Splitting k^tinto bidirectional components between the central (V¹) and peripheral (V²) fluid spaces was suggested in the antecedent study to describe bidirectional fluid exchange between the spaces. 12During that study, it was shown that the combination of mechanical ventilation and isoflurane anesthesia causes an apparent reduction in fluid transfer from V²and V¹following dilution of V²in isoflurane-anesthetized and mechanically ventilated sheep. Therefore, the first kinetic approach applied in this study tested for directional transfer of fluid from V¹to V²and from V²to V¹. The results were disappointing as this analysis yielded poor curve fits and no support was obtained for the view that fluid accumulation of V²developed as a consequence of isoflurane use. The failure of the split k^tmodel to fit the data suggests that all the peripheral fluid is not distributed to V², but rather some is transferred to a “third space,” which acts functionally as if it were isolated from V¹and V². Although the split k^tused in this model was not successful, the split k^tmodification may prove to be a useful tool for the estimation of edema in future volume kinetic analysis models.

The steeper elimination curves for the isoflurane experiments shown in figure 5suggest that elimination of the infused fluid was more rapid in the presence of isoflurane, while in fact, the measured urinary excretion showed the opposite. This discrepancy leads to the hypothesis that fluid accumulated in a “third space” in the body that does not equilibration with V². This could be an anatomically separate space as in the gut or abdominal cavity or just a functionally sequestered part of the interstitial space from which volume equilibration with the plasma does not readily occur.

The theory of perioperative third-space loss, i.e. , that fluid leaves the vascular volume to accumulate extravascularly in a space that does not equilibrate with the vascular space, is consistent with kinetic analyses. The amount of fluid that escaped volume equilibration was estimated using volume kinetics. Here, k^band k^rwere used to describe such accumulation depending on whether it occurred according to a zero-order or first-order process, respectively. Good curve fits for these analyses were obtained for both approaches. The simulations indicate that third-space losses were negligible in conscious sheep, but with isoflurane, the lost fluid was equal to most of the infused volume, approximately 21 ml/kg after 180 min. This was calculated from the product of the average isoflurane k^b(3.2 ml/min) from table 2divided by mean body weight and multiplied by 180 min. From a kinetic perspective, most of the 25 ml/kg infused fluid thus remained in a body fluid space different from V¹and V². Computer simulations were performed to examine the impact of these third-space losses. By considering that the measured urinary excretion represented all the eliminated fluid, the elimination curves for the isoflurane sheep would be notoriously flat, with the final dilution at 180 min being approximately 0.15. In reality, however, the dilution had returned to zero at that time. The increased k^bor third-space losses measured in the isoflurane protocols corresponded well with the ΔISFV of 22–23 ml/kg calculated from indicator dilution and mass balance.

Previous work has shown that the kinetics of a crystalloid fluid load are often sufficiently explained by a one-volume model when urinary excretion is prompt, while a two-volume model is likely to be statistically justified when the fluid is excreted more slowly. 33Thirteen out of the 14 isoflurane experiments in the current study were most consistent with the two-volume model, which could be expected since the urine output was small. Some of the nonisoflurane experiments could be sufficiently explained by the one-volume model, while the two-volume model was justified for others (table 2). The parameter estimates for the latter show that the size of V¹was 35 ml/kg (mean V¹of 1,069 divided by mean weight of 28.5 kg), which was smaller than the measured PV (67.2 ml/kg). The sum of V¹and V²usually amounted to approximately 15% of the body weight, which is an expected ratio when the majority of the experiments are reported according to the two-volume model. 30 

In this study, we used two techniques, both based on repeated measurements of the vascular tracer hemoglobin, to assess changes in volume after infusion of intravenous crystalloid. The first technique, which combines a conventional indicator dilution measurement of PV with mass balance quantification of fluid distribution, provides repeated static estimates of PV and interstitial fluid volume. The mass balance technique requires the assumptions that hemoglobin is distributed only in the vascular space, that hemoglobin is not added or removed from the vascular space during the course of the measurements, and that dilution of hemoglobin occurs uniformly in the blood volume. In contrast, volume kinetics estimate clearance rates and transfer rates between modeled fluid spaces that do not correspond to conventional concepts, such as PV or interstitial volume. In volume kinetics, the calculated volumes are not assumed simply to represent changes in PV or interstitial fluid volume but may be influenced by differences in the rate of equilibration of dilution in vascular beds of differing rates of perfusion and different baseline blood volumes. The resulting model fluid spaces, transfer coefficients, and clearance coefficients describe physiologic fluid handling without precisely measuring physiologic spaces. Both approaches can be used to design predictive models for use in clinical fluid therapy and can estimate the effects of factors, such as hemorrhage, cardiac function, vascular permeability, colloid osmotic pressure, renal function, and neuroendocrine changes, that alter fluid homeostasis.

A practical advantage of a technique based on indicator dilution and mass balance analysis is that it describes changes in traditional physiologic compartments that are familiar to both scientists and clinicians, whereas volume kinetic models calculate parameters that are less intuitive. However, the ability of volume kinetics to quantify clearance and transfer of fluids provides useful insights into fluid handling.

In this study, the specific effects of isoflurane and mechanical ventilation were explored while avoiding the intraoperative factors, such as surgical trauma and tissue edema formation, that alter circulating blood volume during surgery. However, the mechanisms of altered fluid handling require speculation. Anesthetics reduce urinary output directly by the action of nephrotoxic metabolites or indirectly due to systemic hypotension, renal vasoconstriction, sympathetic stimulation, altered levels of antidiuretic hormone, and catecholamine release. 34The sheep response to isoflurane is increased arterial blood pressure, which is different from the human response of hypotension. Isoflurane decreases sympathetic activity in man and reduces catecholamine release in a dose-dependent manner. 4,6However, cardiovascular function in sheep is more dependent on vagal tone. 20,21The direct tubular effects of isoflurane are presumably minimal in that only 0.17% of isoflurane is excreted as urinary metabolites. Isoflurane does, however, decrease glomerular filtration by 30–50%, renal blood flow by 40–60%, and urinary flow rate to 34% of unanesthetized controls. 3 

The low urinary volume after fluid infusion in the anesthetized protocols did not result in increased intravascular retention of fluid, but rather in increased interstitial fluid volume. Alterations in the Starling equilibrium do not appear to explain these findings. Plasma protein concentrations, serum osmolarity, and serum sodium concentrations are reportedly unchanged by isoflurane anesthesia. 3Venous pressures are generally reduced by anesthetics, suggesting reduced capillary pressure, but if precapillary resistance is reduced more than postcapillary resistance, an increase in capillary hydrostatic pressure is possible. Isoflurane does, however, increase secretion of antidiuretic hormone, 7which causes antidiuresis by acting directly on transmembrane aquaporin proteins. 35The aquaporin proteins which have been isolated to date are found not only in the kidney but in several locations in the body. 36,37The transmembrane movement of fluid caused by antidiuretic hormone in the kidney may occur in more diverse sites than previously appreciated, thus contributing to interstitial fluid accumulation.

Although hormonal levels were not measured in this study, HR, which is a factor known to influence release of atrial natriuretic peptide, 38increased significantly in the anesthetized protocols. Atrial natriuretic peptide, in addition to facilitating renal excretion of sodium, has been shown to increase transcapillary movement of water, electrolytes, and albumin 39and could play a role in the net increase of interstitial fluid volume under anesthesia. In vivo  cardiovascular function is largely preserved during isoflurane anesthesia, 3as in these experiments; therefore, a deterioration in cardiac function would be an unlikely cause of increased interstitial fluid accumulation during isoflurane anesthesia.

In conclusion, isoflurane and not mechanical ventilation inhibits the diuretic response to a volume load with 0.9% saline in sheep and is associated instead with a corresponding increase in extravascular fluid. The physiologic mechanism by which isoflurane causes these changes has not been determined. Kinetic analysis of the hemodilution suggests that the “lost” fluid accumulates in a body fluid space that does not readily equilibrate with arterial blood. Further studies are required to determine whether other inhalational and intravenous anesthetics cause similar patterns of fluid retention and to separate the influence of anesthesia from that of surgical trauma. Confirmation of the clinical relevance of these findings requires an evaluation in humans.

The authors thank Mary Townsend (Administrative Secretary, Department of Anesthesiology, University of Texas Medical Branch, Galveston, Texas) for her careful preparation and editing of the manuscript.