Volume kinetics is a branch of pharmacokinetics particularly designed for studying fluid distribution over time.1 An infusion of a fluid bolus causes changes of plasma dilution, which provides information on how the body handles the infused fluid. This concept, similar to pharmacokinetics for drugs, has substantially improved our understanding of how fluids are functionally distributed within the body. By calculating volume kinetic parameters, such as volume of distribution (V) and clearance (Cl) rate, we can predict plasma dilution of an infused fluid over time. This could enhance fluid therapy planning in a clinical setting. However, volume kinetic analysis requires repetitive invasive sampling of blood, which may cause discomfort to patients, it is time consuming, and not clinically practical.2
Pulse CO-oximetry is a recent advancement in patient monitoring that allows for the continuous noninvasive measurement of hemoglobin concentration (SpHb). Pulse CO-oximetry has been shown to provide accurate measurements of hemoglobin (Hb) in healthy volunteers undergoing hemodilution. However, a recent study using an early version of the sensor and software concluded that SpHb was not accurate enough for calculations of volume kinetic parameters.3 In the present study, we examined whether serial measurements of SpHb provide reliable data for the purpose of volume kinetic analysis compared with Hb measurements from invasive venous blood samples, total Hb (tHb).
Our hypothesis was that SpHb monitoring could replace invasive sampling for the purpose of volume kinetic analysis. Any difference in the volume kinetic parameters, as calculated using tHb or SpHb, would indicate that our hypothesis was not correct.
This was an observational study that took place in 2010. The study was approved by the Ethical Board of Stockholm County, Stockholm, Sweden (no. 2009/669-31/3). Thirty patients in 2 age groups admitted for various reasons to the emergency room (ER) in a tertiary care center in Stockholm were included. One group (young) consisted of subjects aged 20 to 39 years (mean, 30 years); the other group (geriatric) consisted of patients 75 to 97 years (mean, 84 years). At the time of enrollment, subjects underwent a medical history and short physical examination. Subjects were excluded if they were terminally ill, classified as New York Heart Association IV, previously diagnosed with renal insufficiency expressed as creatinine clearance <30 mL/min, had cognitive dysfunction expressed as Pfeiffer score <7, or required immediate attention in the ER according to the Adaptive Process Triage (ADAPT) scale.4 Furthermore, they were excluded if they had chest pain, arrhythmias, open fractures, or any medical condition that, in the judgement of the investigator, rendered them unsuitable for participation in the study.
Patients arriving in the ER were triaged according to the ADAPT scale.4 If this triage identified patients who needed emergent care but were hemodynamically stable (ADAPT yellow score), they were eligible for study participation. The investigation was performed while the patients were waiting for the emergency staff to routinely assess them. The patients received information about the study orally as well as in writing and gave their written consent before entering the protocol. IV catheters were placed in both arms of the subjects. In one of the catheters, a room-temperature infusion of a 7 mL/kg buffered crystalloid glucose solution (Rehydrex® with ionic content Na+ 70 mmol, Cl− 45 mmol, Ac− 25 mmol per 1000 mL, glucose content 25 mg/mL = 2.5% = 25 g/L) was given for 15 minutes tvo induce plasma volume expansion. All infusions were administered via an infusion pump (Infusomat FMS®; Braun, Melsungen, Germany). In the other cannula, venous blood samples were taken for the measurement of tHb using a central lab analyzer (Sysmex XE-5000 automated hematology analyzer; Roche Diagnostics, Paris, France). This device was calibrated daily according to manufacturer’s instructions and good laboratory practice, using the cyanide-free, sodium lauryl sulfate method; the coefficient of variation is reported to be <2%. For each patient, a baseline tHb reading (2 samples and their mean) was taken at time 0. Further venous blood samples were obtained at 5, 10, 15, 30, 45, 60, 75, and 90 minutes.
During these interventions, the subjects were simultaneously monitored by a pulse CO-oximetry device (Radical-7® Pulse CO-Oximeter; Masimo, Irvine, CA; SET software version 220.127.116.11) for continuous and noninvasive measurement of SpHb, for peripheral perfusion index (PI) (defined as the ratio between the pulsatile and nonpulsatile absorption of infrared light at the sensor site), oxygen saturation, and pulse rate. This was done by a single spectrophotometric adhesive sensor (Rainbow® ReSposable™ R2-25, revision E; Masimo) attached to either the patient’s ring or middle finger on the noninfusion arm. The adhesive portion of the sensor was placed so that the emitter and detector were precisely aligned on the finger following the manufacturer’s directions for use. Sensors were covered with opaque shields to prevent optical interference. SpHb was recorded at the same time that invasive samples were drawn, which resulted in 9 paired tHb and SpHb values per subject. Data were continuously recorded and downloaded to a computer running data collection software (Physiolog; Masimo). After 90 minutes, the monitor was disconnected and the investigation terminated.
To avoid placement of an arterial line, only venous blood was drawn. If the hand of the subject was cold, warming was provided with heated blankets. The sensitivity of the pulse CO-oximeter was set initially to Adaptive Probe Off Detection (APOD) mode. If continuous values were not acquired with this setting, the pulse CO-oximeter was set to either NORMAL or MAX sensitivity mode. All data were analyzed, including data where the signal sensitivity (SIQ) value was <50%. SIQ is the manufacturer’s proprietary signal quality indicator, which provides a continuous assessment of the reliability of the SpHb measurement. Low SIQ values may be caused by improper placement of the sensor, patient motion, low peripheral perfusion, or a combination of factors.
SpHb accuracy was assessed using Bland-Altman plots for multiple measurements to compare the difference between simultaneous values of SpHb and tHb. To assess whether there was a difference in the response to dilution between the young group and the geriatric group, we calculated the change in Hb: (1) from baseline to the end of the fluid bolus, and (2) from the end of the fluid bolus to the end of the 90-minute test period. A larger decrease in Hb would indicate a more profound dehydration. A volume kinetic analysis was then performed to obtain the distribution volume and clearance of the infused fluid using both tHb and SpHb data for comparison, as described in the Kinetics section below.
Nonlinear regression is the fundamental technique used to analyze physiologic clearance curves. When using “linear” regression to analyze a curve consisting of raw data points, a line is chosen to minimize the sum of the squares of the vertical distances of a series of individual data points from that line.5 Nonlinear regression also uses the least-squares method, which is appropriate if experimental uncertainty is normally distributed.
During hemodilution, the infused fluid is dissolved in the blood, which already contains 80% water. Before being used in further calculations, the hemodilution must be transformed to the corresponding plasma dilution because it is the plasma volume that equilibrates with other fluid spaces in the body. From baseline at time 0 to time t, the dilution of the plasma, which equals the dilution of V (mL) can be expressed as:
Plasma dilution (t) = [(Hb0 – Hbt)/Hbt]/(1 – baseline hematocrit)
The infused fluid is thought to expand a single body fluid space called v (mL), which the body strives to maintain at the target volume, V. Elimination of fluid occurs by baseline urinary excretion and evaporation, Cl0 (mL·min−1), and by a dilution-dependent mechanism governed by a constant, Cl (mL·min−1). The infusion in mL/min is R0. The net volume change a (mL) in the 1-volume model is given by the following differential equation2:
We used Equations (1) and (2) to analyze the data by estimating V, Cl, and Cl0 during and after infusion for both tHb and SpHb data. This analysis was made using MatLab (MatLab® 18.104.22.1685, R14; MathWorks, Inc., Natick, MA). We used fminsearch (Nelder-Mead simplex method) to find initial estimates, and nlinfit (Levenberg-Marquardt algorithm) for the final estimation with error estimates. The differential Equation (1), together with the constraints given in (2), was solved by ode45 (explicit Runge-Kutta method). Data are presented as V (invasive), VNI (noninvasive), Cl (invasive), and ClNI (noninvasive), respectively.
We calculated a post hoc power analysis based on dependent t test, with a significance level of 5% (2-sided) based on the observed standard deviations of the difference between the invasive and noninvasive methods for the V and Cl. This showed that to detect a difference with a reasonable chance (80% power) between the invasive and noninvasive methods with 30 patients, the difference had to be approximately 1000 mL in V between the invasive and noninvasive methods.
Statistics and Calculations
Monitor and patient data collected by the study coordinators were entered into a database for statistical analysis (PASW Statistics 18; SPSS, IBM Corporation, Armonk, NY). For comparison of methods, Bland-Altman graphs were generated for multiple observations from the same subject when the quantity varied over the period of observation (MedCalc version 12.0.0; MedCalc Software, Mariakerke, Belgium).6,7 The bias describes whether there is a systematic deviation between the SpHb and tHb (i.e., SpHb–tHb). Precision was defined as 1 SD of the bias of all points. Limits of agreement (LOA) were defined as ±2 SDs from mean bias. Confidence intervals for bias and LOA were calculated. Data are presented as medians and ranges when appropriate. Statistical comparisons between the invasive and noninvasive V and Cl (V and VNI and Cl and ClNI) were made, respectively. Data for kinetic calculations (V and Cl) were not normally distributed according to Shapiro-Wilk tests and therefore nonparametric tests were used. Wilcoxon signed rank tests were used when appropriate. P < 0.05 was considered significant.
The young group included 14 patients with ages ranging from 25 to 34 years (mean, 30 years). The geriatric group included 16 patients with ages ranging from 78 to 88 years (mean, 84 years). Data from the 2 investigating methods rendered Bland-Altman plots as seen in Figures 1 to 3. There were 242 pairs of tHb and SpHb in total. The average bias and precision of the 242 paired measurements were −0.47 and 1.0 g/dL. When data associated with SpHb SIQ values <50% were omitted, 193 data pairs remained. Data for anemic patients gave slightly less bias. Figure 4 shows that a PI >2 reduces bias. Results for bias, LOA, and their respective confidence intervals are summarized in Table 1. There were no differences in dilution between the age groups (data not shown). The volume kinetic analysis generated V and Cl rate constants, Cl and Cl0, listed in Table 2. Three subjects had to be omitted from the kinetic analysis because the nlinfit method reported the fit as nonidentifiable. For the data from the 27 subjects for which the nlinfit equation could be solved, there were no significant differences in the estimation of the distribution volume, using either tHb value (3070 mL [2601–4208 mL]) or SpHb value (2886 mL [2122–3915 mL]) (P ≤ 0.296, Wilcoxon matched-pair signed rank test). The quality of fit was also similar between tHb and SpHb when examining the standard deviation of the estimates. The elimination rate constant Cl was estimated to be 40.0 mL·min−1 (26.3–50.8 mL·min−1) (Hb) and 18.5 mL·min−1 (0.2–84.0 mL·min−1) (SpHb). Although there was a tendency for a higher clearance in the tHb group, this difference was not statistically significant (P ≤ 0.7276, Wilcoxon matched-pair signed rank test).
The primary objective of this study was to estimate V and Cl in a kinetic model. The basic finding of this study was that there were no significant differences between V and Cl calculated using either invasive (tHb), or noninvasive (SpHb) Hb measurements. This means that noninvasive Hb samples could possibly be used for volume kinetic modeling of infusion of fluids. This could facilitate planning of fluid therapy.
Fluid kinetic models are based on repetitive sampling of an endogenous marker (Hb), which means that a series of Hb measurements are used to derive plasma dilution curves to predict V and Cl. Because kinetic modeling is dependent on a series of sampled values, it should be possible to exchange tHb values for SpHb values provided the overall variability is acceptable.
In this study, the average bias and precision of the 242 paired measurements was −0.47 and 1.0 g/dL. This means that SpHb underestimates tHb values on average by approximately 0.5 g/dL. Even if data with low perfusion (SIQ signal <50 %) are excluded, the bias is slightly negative (−0.24). However, as seen by LOA, there is a wide variation of individual data, which means that a single SpHb value could have difficulty to precisely predict a simultaneous tHb value. For instance, a tHb of 12 could admittedly produce a SpHb of anything between 9 and 14. The algorithm for SpHb SIQ is proprietary to the company but is likely influenced by peripheral perfusion because the precision between the 2 methods appears to improve when PI is >2, as shown in Figure 4. In our study, low perfusion, as indicated by low PI values, occurred in 5 of the 30 subjects (17%). This is marginally different from the study by Hahn et al.,3 in which they reported that 25% of the SpHb data collected were associated with a low PI. However, diseases affected patients in our study whereas there were healthy volunteers in the Hahn et al. study.3 Noninvasive Hb data obviously vary more compared with simultaneous invasive data. Whether this is attributable to differences in capillary equilibration (fluxes of extracellular fluid) or is related to some other aspect of the device performance requires further investigation. One possible explanation could be that the pulse CO-oximetry evaluates a mean value of Hb as a function of signals from a network of capillaries, whereas an invasive sample is based on a more predictable blood flow in an arterial or venous vessel.8 Our results show that SpHb measurements by pulse CO-oximetry can be used to estimate V, which is mainly determined during the infusion phase.
Furthermore, our results show that the elimination constant, Cl, is determined with less accuracy. The clearance factor is mainly determined during the elimination phase, when the infusion is terminated and therefore takes longer to determine.
The large confidence intervals for differences between age groups further support the argument that it is unclear whether or not there was a discrepancy between the young and geriatric groups in terms of fluid responsiveness by examining decreases in invasive versus noninvasive Hb.
Noninvasive CO-oximetry produces a series of SpHb estimations that can be used to calculate the V when buffered glucose solutions are administered to acutely ill but hemodynamically stable patients.
Name: Fredrik Sjöstrand, MD, PhD.
Contribution: This author helped design the study, conduct the study, analyze the data, and write the manuscript.
Attestation: Fredrik Sjöstrand has seen the original study data, reviewed the analysis of the data, approved the final manuscript, and is the author responsible for archiving the study files.
Name: Peter Rodhe, PhD, MSc.
Contribution: This author helped analyze the data and write the manuscript.
Attestation: Peter Rodhe has seen the original study data, reviewed the analysis of the data, and approved the final manuscript.
Name: Ellinor Berglund, RN.
Contribution: This author helped design the study, conduct the study, and analyze the data.
Attestation: Ellinor Berglund has seen the original study data, reviewed the analysis of the data, and approved the final manuscript.
Name: Niclas Lundström, Research Fellow.
Contribution: This author helped design the study, conduct the study, analyze the data, and write the manuscript.
Attestation: Niclas Lundström has seen the original study data, reviewed the analysis of the data, and approved the final manuscript.
Name: Christer Svensen, MD, PhD.
Contribution: This author helped design the study, analyze the data, and write the manuscript.
Attestation: Christer Svensen has seen the original study data, reviewed the analysis of the data, and approved the final manuscript.
Conflicts: Christer Svensen served in 2009 and 2011 as an adviser to Masimo at scientific meetings in connection with the ESA congresses in Milan and Amsterdam for which he received honorarium. This study was also partly funded by an unrestricted grant from Masimo Inc.
This manuscript was handled by: Dwayne R. Westenskow, PhD.
The authors are indebted to Hans Järnbert-Pettersson, Statistician at the Department of Clinical Science and Education, for support. We also thank the staff at the Emergency Room at Södersjukhuset for valuable support and understanding of the importance of research in a difficult environment.