Capillary Filtration of Plasma Is Accelerated During General Anesthesia: A Secondary Population Volume Kinetic Analysis

Authors: Robert G. Hahn

Abstract

How infusion fluids are distributed and eliminated is of importance to how much and how fast they should be administered. This manuscript applies population pharmacokinetic modeling to intravenous infusions of crystalloid fluid, which is a common therapy in hospital care and mandatory during surgery. The analysis was based on the hemodilution and urine output measured during and after 262 infusions of 1647 ± 461 mL (mean ± SD) of fluid over 30 min in adults. The result shows that distribution of fluid from the plasma to the interstitial fluid space occurred twice as fast during general anesthesia as compared to the conscious state. The increased rate ensures adequate nutritional flow to the cells despite decreased flow in the macrocirculation, which is a characteristic of general anesthesia. This increased capillary leakage of fluid was coupled with an even greater reduction of the urinary output and accumulation of fluid in both the fast‐exchange interstitial fluid space and a remote “third fluid space,” the latter of which apparently serves as an overflow reservoir. During the first hour of the experiments, 88% more fluid resided extravascularly in the presence of general anesthesia than in the awake state. General anesthesia increased the half‐life from 1.8 to 16.6 h, showing marked impairment in the handling of infused crystalloid fluid.

Introduction

“Volume kinetics” is pharmacokinetics of infusion fluids and deals with plasma dilution instead of drug concentrations for the calculations. Infusion fluids behave like drugs in the body and are associated with adverse effects if given in too small or too large amounts in those who needed volume therapy. ^1^ , ^2^ , ^3^ , ^4^ , ^5^ Their medical effects are exerted solely by volume expansion of the physiological body fluid spaces.

Volume kinetics has been used to characterize the turnover of all marketed infusion fluids ^6^ and highlighted changes that occur in response to dehydration and bleeding ^7^ but is still not well known among clinical pharmacologists. The approach operates by identifying fictitious “walls” that delay the distribution of administered fluid. These walls represent functional barriers but may be interpreted in quasi‐physiological terms as crystalloid fluid is not metabolized or bound to any tissue.

This study presents volume kinetics by exploring whether fluid‐induced capillary filtration occurs at different rates during general anesthesia compared to the awake state. The question is not trivial because the macrocirculation is depressed by general anesthesia, as evidenced by low arterial pressure and cardiac output. ^8^ The extracellular fluid needs to be in constant motion by circulating between a high‐speed and a low‐speed system, that is, the plasma and the interstitial space/lymphatics. Disturbances of the equilibrium between these two systems may occur during ongoing surgery due to the anesthesia‐induced vasodilatation and trauma‐induced activation of inflammatory and hormonal systems. A combination of depressed microcirculation and depressed transcapillary fluid exchange might cause stagnation of the nutritional flow to the cells and impair the removal of metabolic waste products. ^1^

The specific aim of the present study was to compare the forces that govern the distribution and elimination of infused Ringer´s solution in the awake and anesthetized state with special emphasis of the capillary filtration. For this purpose, 262 infusion experiments were chosen from a data bank based on similar infusion times and infusion volumes. The hypothesis was that the fluid‐induced acceleration of the transcapillary leakage of fluid is reduced during anesthesia and surgery in accordance with the depression of the macrocirculation.

Methods

Data Acquisition

Material was derived from a database with intravenous infusion experiments in humans performed during the past 25 years. For the current presentation, 262 experiments were included where approximately 25 mL/kg of Ringer´s acetate/lactate was infused over approximately 30 min in euhydrated subjects (<18 years) who were in the awake or anesthetized state. Fifteen of the subjects received isotonic (0.9%) saline. The data were retrieved from 12 studies ^9^ , ^10^ , ^11^ , ^12^ , ^13^ , ^14^ , ^15^ , ^16^ , ^17^ , ^18^ , ^19^ , ^20^ that are detailed in Table 1. The author planned and supervised all studies, which were performed using similar protocols.

No subject had restricted kidney function or severe cardiovascular disease. No subject had undergone deliberate dehydration, withdrawal of blood (in addition to sampling volume), administration of adrenergic drugs, or, if the experiment was performed under general anesthesia, no fluid had been administered during onset of the anesthesia. Surgical hemorrhage, if any, was minimal (50‐100 mL).

Procedure

The subjects were allowed to ingest one glass of liquid and one sandwich on the morning of the experiment, which usually started between 8 and 9 a.m., to prevent blunt dehydration. Those who underwent general anesthesia were in the fasting state and either received no premedication or diazepam by mouth.

The excreted urine volume was measured two to three times per experiment, and 3 mL samples of whole blood for measurement of the hemoglobin (Hb) concentration were obtained every 5 min during the infusion, for 30 min thereafter, and at 15‐30 min intervals for a total length of up to 4 h. Hb was analyzed at same hospital´s certified clinical chemistry laboratory with a coefficient of variation of 0.5%‐1%. Data obtained after awakening from anesthesia were excluded.

Kinetic Analysis

Base Model

A three‐volume kinetic model with five rate constants (k12, k21, k10, k23, and k32; unit is 1/min for all) and one scaling factor between dilution and volume (Vc, central volume, mL) were fitted to the dependent variables, which were the frequently measured plasma dilution and the urinary excretion. ^4^ Vc can be regarded as a quasi‐measure of the plasma volume. ^6^

The kinetic model is illustrated in Figure 1a and is intended to mimic human physiology. Briefly, fluid is infused into the plasma (Vc,), from which distribution occurs to an extravascular space, Vt1 (mL), and further to a more remote fluid space, Vt2 (“third fluid space,” mL). Redistribution of the fluid occurs in the reverse order. Elimination is governed by the rate constant k10, which is obtained as the measured urine output divided by the modeled volume expansion of Vc during the same time as urine was collected.

The differential equations describing the kinetic model are given in Supplementary File 1.

The Hb‐derived fractional plasma dilution used to indicate the volume expansion of Vc resulting from the Plasmadilution=Hbbaseline/Hblater−1]/1−Hematocritbaseline

Each dilution underwent a minimal correction to account for surgical hemorrhage, if any, and blood sampling. ^17^ , ^21^ How this was done is explained in Supplementary File 1.

Covariates

Kinetic differences between the awake and anesthetized state were examined and quantified by covariate analysis. Identification of appropriate covariates was guided by plots of random effects (“eta:s”) using both forward addition and backward removal of the most promising candidates. ^22^ The exponential covariate model and the power model were used. The criterion for accepting a covariate was that its inclusion should reduce the −2 LL (log likelihood) for the model by >3.84 points (P < .05) or >6.6 points (P < .01). In addition, the 95% confidence interval (CI) for the estimate of the covariate was not allowed to include 0.

Analysis

The kinetic model was simultaneously fit to all measurements of plasma dilution and urinary excretion (dependent variables) using the Phoenix software version 8.3.4 for nonlinear mixed effects (Pharsight, St. Louis, MO) with the first‐order conditional estimation extended least squares (FOCE ELS) as search routine. The sandwich method was employed as the variance estimator, as it is robust for covariance misspecification. The goodness‐of‐fit and performance of the final model were evaluated by predictive checks, residual plots, and plots of the conditional weighted residuals (CWRES). ^23^

Half‐Lives

The half‐life of the infused fluid in the body was derived as the time required for a decrease of 50% to occur of the sum of the simulated volumes of residual fluid present in the three fluid compartments (Vc + Vt1 + Vt2) over time.

Flow Rates

The flow of fluid between compartments was calculated as the product of a rate constant and the volume expansion of the compartment from which the flow is initiated. Hence, the flow from the plasma (Vc) to the fast‐exchange interstitial space (Vt) at any time is the product of k12 and the induced volume expansion of Vc.

Parameter Variations Over Time

Variations in the fixed parameters over time could not be studied in the main analysis due to the maximum limit of 15 parameters in a single run. Instead, exploratory analyses were performed using only the three strongest covariates from the main analysis and examining variations in the fixed parameters during the first 50 min of the study by using a single exponential covariate where each 10‐min period was given a value from 1 to 5. This potential covariate parameter was, in turn, tested for covariance with each of the fixed parameters.

Statistics

Data showing a normal distribution are reported as the mean ± standard deviation (SD). Kinetic parameters are reported as the best estimate and 95% CI according to the output from the Phoenix program. The significance levels for inclusion of the covariates were taken from the Phoenix program.

Results

Basic Data

Data for the study consisted of 262 infusion experiments performed on 134 females and 128 males. Ringer´s solution had been administered to 172 fully awake subjects and to 90 patients while undergoing surgery under general anesthesia. One third of the surgeries were performed via the laparoscopic route (33%).

The subjects were aged 35 ± 14 years and weighed 73 ± 13 kg. They received 1647 ± 461 mL of Ringer's solution at a constant rate over 30 ± 11 min.

Table 1 shows the basic data for the included studies and Table 2 compares these data depending on whether subjects were studied in the awake or anesthetized state.

Kinetic Analysis

The kinetic analysis comprised 6030 measurements of plasma dilution (mean, 23 per experiment) and 604 urine collections (2.3 per experiment).

The base model was successfully fit to all infusion experiments in one single run. The Akaike criterion showed that the three‐volume model was statistically superior to two‐volume and one‐volume models (−14,598 vs −14,270 and −9756) and was then selected for the presentation.

The final estimates of the kinetic parameters are shown in Table 3. Goodness‐of‐fit and model performance measures for all subjects are illustrated by CWRES and residual plots in Figure S1 and separately for awake and anesthetized subjects in Figure S2. A predictive check including all subjects is shown in Figure 1b and separately for awake and anesthetized subjects in Figure S3.

The predicted plasma volume expansion (Vc over time) when 800 and 1600 mL is infused at various rates are shown graphically in Figure 1c,d.

General anesthesia served as a statistically significant covariate to the elimination (k10 −78%), capillary filtration rate (k12 +309%), and the size of the central fluid space (Vc −65%). Females had slightly faster elimination (+22%). The volume of infused crystalloid fluid slightly decreased the rate of elimination (k10) and the return flow of distributed fluid to the plasma (k21). Replacing Ringer´s by isotonic (0.9%) saline decreased the elimination rate by 30%. Covariate effects are illustrated in Figure 2.

Simulating Fluid Distribution

A simulation of the fluid distribution following an infusion of 1600 mL of Ringer's solution over 30 min, based on the best parameter estimates in Table 3, is shown in Figure 3. General anesthesia was associated with greater volume expansion of the two interstitial spaces while the urine output was very low. Figure 4 illustrates that greater filling of Vt1 was needed before Vt2 opened for fluid accumulation during anesthesia and surgery.

The half‐life of the infused fluid in the body averaged 1.8 h in the awake state and 16.6 h in the anesthetized state.

Flow Rates and Parameter Variations Over Time

During the first hour, the flow from Vc to Vt1, that is, the fluid‐induced increase of the capillary filtration, averaged 53 mL/min in the awake state and 98 mL/min in the anesthetized state. The return flow (from Vt1 to Vc) averaged 42 and 78 mL/min, respectively, while the flow from Vt1 to Vt2 averaged 4.0 and 7.3 mL/min. The inflow and outflow were better matched during the subsequent 2 h, but the differences between the two settings increased to a factor of 3.

The examination of variations in the fixed parameters over time showed that k12 was higher while Vc was lower than the overall average during the first 20 min of the infusion (Figure 5). Making the same explorations with only the anesthetized patients yielded the same pattern but the variations had lower amplitudes.

Discussion

Key Result

The analysis shows that the flow rates of fluid to and from the plasma and the extravascular space were twice as high during general anesthesia than in the conscious state. This result is not intuitive, as general anesthesia is associated with slowing of the macrocirculation. ^1^ , ^8^ The acceleration would help to maintain nutritional flow and removal of metabolic waste products despite moderately severe hemodynamic depression. However, the downside is that general anesthetic drugs inhibit lymphatic pumping, ^24^ , ^25^ whereby an excessive amount of filtered fluid accumulates outside the bloodstream. ^26^

Capillary Leakage

The covariate analysis confirmed that the rate constant for capillary leakage of fluid to the extravascular space (k12) increased greatly in the presence of general anesthesia. This can be explained by decreased activity of the adrenergic system, which favors transcapillary filtration by increasing the capillary hydrostatic pressure. The involved mechanisms are relaxation of the arterioles, increased precapillary resistance, and reduced postcapillary venodilatation. ^27^ Another reason for the accelerated flow could be the marked inhibition of urinary excretion by general anesthesia (−78%), which would increase the intravascular hydrostatic pressure by retaining a larger fraction of the infused fluid in the plasma as well as in the whole body. The capillary hydrostatic pressure was not measured, but the higher k12 during infusion than post‐infusion suggests that the fluid pressure affects estimates of this parameter (Figure 5). Moreover, anesthesia‐induced vasodilatation may have increased the availability of capillaries for fluid exchange. By contrast, the transcapillary flow seems to be strongly retarded when general anesthesia has just been induced, but this is a transient phase only. ^28^

Other Covariances

The size of Vc increased with the body weight, which is natural as the plasma volume increases with the body weight. A reduction of k21 occurred for larger infused volumes and might be due to gradual loss of elasticity of the interstitial matrix when the extravascular space is expanded. ^29^ , ^30^ The slightly faster elimination in females ^31^ and the slower elimination when 0.9% saline instead of Ringer's is infused ^32^ is known previously. The mild reduction of k10 with the largest fluid volumes might implicate some exhaustion of the kidneys but can also be due to inclusion of perivascular areas in the estimate of Vc when very much fluid is infused.

The “Third Fluid Space”

The remote interstitial fluid space (Vt2) has only recently been demonstrated since its detection requires many experiments and a population analytical approach to be disclosed. The term “third space” stems from a concept developed by Shires et al in the 1960s, ^33^ but has often been considered to be a methodological artifact. ^34^ , ^35^ However, a third extracellular fluid space clearly exists from a kinetic point of view. For the space to open, >1 L of crystalloid fluid must be administered relatively fast (approximately >50 mL/min). ^36^ This view is confirmed by the exploratory analysis is Figure 5, which suggests that fluid easily flows through Vt2 up to when 2/3 of the fluid volume of 1.6 L has been administered, whereafter k23 increases and k32 decreases to create a “lock‐out” effect.

It has been unclear if the two interstitial compartments communicate with the plasma via a parallel or serial connection, but recent evidence suggests that the connection is serial; hence, the fluid must first enter Vt1 to reach Vt2. ^37^ The “third fluid space” might correspond to the colloid‐rich gel phase of the interstitium, which is difficult to expand but binds water tightly. ^38^ , ^39^ The fast‐exchange interstitial compartment (Vt1) probably corresponds to the free fluid phase of the interstitium and the lymphatic vessels combined.

The “third space” seems to operate as a fluid reservoir that prevents overloading of the blood with fluid. Its main influence on the kinetic analysis consists in a marked prolongation of the half‐life. ^36^ Hence, accumulation of fluid in this colloid‐rich phase might explain why weight gain might last several days after surgery. ^4^ , ^40^

Sizes of the Fluid Spaces

The sizes of the fluid spaces are of theoretical interest. Volume kinetic analysis of isotonic crystalloid fluid detects only expandable fluid spaces, whose sum is typically smaller than the size of the physiological extravascular space; for example, the interstitium of bone tissue and the skull cannot be expanded by volume loading alone. In the present study, the size of Vc of 3.5 L for a conscious human was given directly by the analysis. The fluid distributed to Vt1 occupies a space of similar size (as given by Vc k12/k21) while Vt2 represents only 40% of that volume; hence, the infused Ringer solution is distributed in a total space of 8.4 L, which should be compared to the expected total extravascular fluid space of 15 L. However, the size of Vt2 becomes much larger in inflammatory states and may even occupy a supraphysiological volume, ^26^ which suggests absence of free flow.

Anesthesia Induction

The period of anesthesia induction was not included in this study, but it involves events of importance. If crystalloid fluid is infused during this period, the anesthesia‐induced decrease of the arterial pressure greatly reduces the capillary filtration until a new Starling equilibrium is established, which has been achieved perhaps 15 min later. ^28^ However, in the present study, no fluid was provided during the induction. What then happens is that fluid spontaneously translocates from Vt1 to Vc to equilibrate the decreased hydrostatic pressure in Vc. ^27^ This fluid increases the dilution of Vc, which the model interprets as a low Vc. ^41^ With stable Starling forces the size of Vc correlates well with the estimated plasma volume, but it is operationally a calibration factor between plasma volume and plasma dilution. Hence, a low Vc during general anesthesia does not necessarily indicate hypovolemia.

The low value of Vc during the first 20 min of the infusion (Figure 5a) is probably due to auto‐transfusion. Translocated fluid volume might amount to 500 mL during a short period of an operation but averages 200 mL during the entire procedure ^41^ which corresponds to an anesthesia‐induced decrease of the mean arterial pressure of 30 mmHg. ^27^ Such fluid transfer can partially explain the difference between awake and anesthetized subjects regarding the fluid volume needed to open up Vt2 in Figure 4. It also explains why plasma dilution was greater, while the volume expansion of Vc was lower, during general anesthesia as compared to in the awake state (cf. Figure S3 with Figure 3).

Auto‐transfusion during general anesthesia does not confuse the flow rates between the fluid compartments but slightly modifies their volumes and the rate constants. The volume expansion of Vc should probably be 200 mL greater during anesthesia than shown in Figure 3a, whereby k12 would be twice as high instead of three times higher than in the awake state; this modification receives support from the flow rate calculations. Similarly, Vt2 should expand three times more instead of four times more during anesthesia than in the awake state. Finally, anesthesia would likely show negative covariance with k21, which can be expected due to the inhibitory effect on lymphatic flow. ^24^ , ^25^ This inhibition is probably disclosed by the twice as great difference between inflow and outflow to Vt1 during anesthesia compared to the awake state (98‐78 vs 53‐42 mL/min).

Dilution Kinetics

A benefit with volume kinetics, as compared to the use of radioisotope tracers, is that dynamic events in fluid distribution can be studied and simulated, just as in drug pharmacokinetics. The data is obtained with minimal invasiveness and can be used to answer physiological questions in living humans. The collection of data used for the present calculations has been made possible by maintaining the same strategy for the setup and blood sampling during fluid studies performed during several decades.

An unused possibility is to use volume kinetics to better understand the kinetics of drugs that are given together with fluid, that is, when the fluid serves as carrier. A setting where this occurs include antibiotics administered in intensive care. In general, the plasma and interstitial concentrations of water‐soluble drugs can be expected to vary with fluid volume changes in the three body fluid compartments, which may undergo dramatic shifts in severe disease.

Areas where the method has already been employed is to identify the level of fluid maldistribution in disease states, such as preeclampsia, ^42^ and physiological conditions including dehydration and hemorrhage, ^7^ and to describe the fluid kinetics in special settings, such as pediatric anesthesia ^43^ and gastrectomy in adults. ^44^ The obtained kinetic data can be used to recommend infusion rates that results in a steady‐state plasma volume expansion and avoid abrupt variations in plasma volume expansion that are at risk of causing adverse effects.

A pharmacokinetic model is not usually interpreted in physiological terms, but much evidence supports the idea that volume kinetics reflects the distribution of infused fluid between physiological body fluid compartments. The size of Vc is routinely close to the plasma volume obtained by anthropometry, and changes in hematocrit correspond well to isotope‐measured changes in plasma and blood volumes. ^45^ , ^46^ , ^47^ Elimination is quantified by collection of the excreted urine, and the k21‐generated flow agrees with the flow pattern in the thoracic duct. ^48^ Partition coefficients and metabolism are not issues but experiments in non‐steady‐state settings should be avoided, or the altered physiology should be considered by covariates. Hence, in the current presentation, k12 is used as the rate constant for capillary filtration and k21 for lymphatic flow, whereas physiological correlates to k23 and k32 are still unclear.

Limitations

An overarching question is if the general relationships between the flows of fluid in the body in the anesthetized state differ from those in the awake state. The literature provides little guidance on this issue, which probably stems from the methodological challenge of providing an overview of fluid turnover in the whole body. Despite the role of volume kinetics to fill this gap, the current presentation has several shortcomings.

The kinetic constants and the simulated flow rates express changes in flow rates induced by the infusion of Ringer´s solution. The capillary leakage rate of fluid at baseline is not included, but it amounts to approximately 7 mL/min, which can be inferred from the known capillary leakage of albumin and the albumin content of lymph (40%‐50%). ^49^ , ^50^

The analysis is based on several previous studies, where the data had been collected in the same way but with different study questions than the one raised here.

Both inhaled and intravenous anesthesia could be used.

Two types of solutions were used, both lactated and acetated Ringer's, but a previous analysis shows that the choice of solution has a negligible influence on the fluid kinetics. ^51^ Both solutions are slightly hypotonic (osmolality 270 mosmol/kg) and would therefore be expected to pass into the body cells and cause edema. However, the sodium concentration of the first urine portion excreted in response to the fluid load normally is much lower than that of the infused fluid, so it counteracts this withdrawal. ^52^ A few subjects were given isotonic (0.9%) saline, which is known to reduce the urine output ^32^ but otherwise shares the kinetics with the Ringer solutions.

The program for population kinetics used could only include 15 parameters, which is not enough to cover all aspects of the kinetics of crystalloid fluid. These include the time dependencies indicated in Figure 5 and hemodynamic variables. The entrance of fluid to Vt2 is typically subject to a time delay, ^36^ , ^37^ although this did not reach statistical significance in the present study. Moreover, the arterial pressure can explain much of the reduction of k10 by general anesthesia. ^51^ The k21 is inhibited as long as the infusion continues, which can be explained by the rise in venous pressure that accompanies volume loading. ^37^ Finally, old age slightly decreases k10. ^51^

Conclusions

Analysis of fluid volume kinetics following infusion of Ringer's solution showed acceleration of the rate of distribution of fluid from the plasma to the extravascular space during general anesthesia as compared to the awake state. The rate increase amounted to approximately 100% which should ensure adequate nutritional flow to the cells despite decreased blood flow. Fluid also slowly accumulated in a second extravascular fluid compartment with very slow exchange with the plasma, which might explain why edema and an increase in body weight develop and can last several days after surgery.

Author Contributions

Robert G. Hahn initiated and supervised the data collection in the included studies, performed the analyses, and authored the manuscript.

Conflicts of Interest

Robert G. Hahn has received a grant from Grifols for studies of 20% albumin.

Funding

There was no specific funding for the present work.

Supporting information