- Research article
- Open Access
Pharmacokinetic-pharmacodynamic modelling of the cardiovascular effects of drugs – method development and application to magnesium in sheep
© Upton and Ludbrook; licensee BioMed Central Ltd. 2005
- Received: 28 October 2004
- Accepted: 10 March 2005
- Published: 10 March 2005
There have been few reports of pharmacokinetic models that have been linked to models of the cardiovascular system. Such models could predict the cardiovascular effects of a drug under a variety of circumstances. Limiting factors may be the lack of a suitably simple cardiovascular model, the difficulty in managing extensive cardiovascular data sets, and the lack of physiologically based pharmacokinetic models that can account for blood flow changes that may be caused by a drug. An approach for addressing these limitations is proposed, and illustrated using data on the cardiovascular effects of magnesium given intravenously to sheep.
The cardiovascular model was based on compartments for venous and arterial blood. Blood flowed from arterial to venous compartments via a passive flow through a systemic vascular resistance. Blood flowed from venous to arterial via a pump (the heart-lung system), the pumping rate was governed by the venous pressure (Frank-Starling mechanism). Heart rate was controlled via the difference between arterial blood pressure and a set point (Baroreceptor control). Constraints were made to pressure-volume relationships, pressure-stroke volume relationships, and physical limits were imposed to produce plausible cardiac function curves and baseline cardiovascular variables. "Cardiovascular radar plots" were developed for concisely displaying the cardiovascular status. A recirculatory kinetic model of magnesium was developed that could account for the large changes in cardiac output caused by this drug. Arterial concentrations predicted by the kinetic model were linked to the systemic vascular resistance and venous compliance terms of the cardiovascular model. The kinetic-dynamic model based on a training data set (30 mmol over 2 min) was used to predict the results for a separate validation data set (30 mmol over 5 min).
The kinetic-dynamic model was able to describe the training data set. A recirculatory kinetic model was a good description of the acute kinetics of magnesium in sheep. The volume of distribution of magnesium in the lungs was 0.89 L, and in the body was 4.02 L. A permeability term (0.59 L min-1) described the distribution of magnesium into a deeper (probably intracellular) compartment. The final kinetic-dynamic model was able to predict the validation data set. The mean prediction error for the arterial magnesium concentrations, cardiac output and mean arterial blood pressure for the validation data set were 0.02, 3.0 and 6.1%, respectively.
The combination of a recirculatory model and a simple two-compartment cardiovascular model was able to describe and predict the kinetics and cardiovascular effects of magnesium in sheep.
- Cardiac Output
- Arterial Blood Pressure
- Pharmacokinetic Model
- Systemic Vascular Resistance
- Pharmacodynamic Model
The effective use of some drugs can be limited by their adverse effects on the cardiovascular system, particularly when they are used intravenously in relatively high doses. There have been many studies documenting the cardio-vascular effects of drugs. Similarly, many mathematical models of the cardiovascular system, of varying complexity, have been presented in the literature [1, 2]. In pioneering work, models of the cardiovascular system have been linked to pharmacokinetic models of volatile anaesthetic disposition [3–5]. These kinetic-dynamic models have since been developed into mannequin based anaesthesia simulators, which now have a pivotal role in the training of anaesthetists. This approach has been facilitated by the fact that models of volatile anaesthetic disposition have traditionally been physiologically based (e.g. using representations of tissue:blood partition coefficients and blood flows for individual organs or groups of organs). It is therefore possible to equate blood flow in the cardiovascular model to blood flow in the pharmacokinetic model. Nevertheless, a limiting factor in the implementation of this approach is the availability of experimental data on concentration-effect relationships .
In contrast, for traditional ("non-volatile") drugs, there have been very few instances in which kinetic models of a drug have been linked to cardiovascular pharmacodynamic models. The work of Francheteau et al. is an important exception , but even this early work was restricted to analysis of only one aspect of the cardiovascular system (i.e. accounting for heart rate mediated control of blood pressure but not Frank-Starling control of cardiac output). However, it is clear this approach has the potential to provide a more rational basis for designing dose regimens of cardio-active drugs, and could provide insight into strategies for controlling their cardio-vascular effects. It maybe possible to predict a priori the cardiovascular consequences of, for example, a change in dose regimen of a drug.
There are a number of difficulties in implementing this approach for traditional drugs. One problem is that most drugs do not cause changes in one single cardiovascular variable (such as blood pressure) that can be described in the usual manner using a simple semi-empirical dynamic model (e.g. an Emax model). Rather, a number of cardiovascular variables may be altered simultaneously in a manner that is complex and interrelated. Thus, any dynamic model used must account for these intrinsic relationships between cardiovascular variables. Another problem is that changes in the cardiovascular system (in particular blood flow distribution) invariably alter the kinetics of the drug under study. Therefore, the kinetic model of the drug must be able to account for the effects of blood flow changes on the disposition of the drug. This requires the kinetic model to have a physiological basis, and importantly excludes the widely used mamillary compartmental pharmacokinetic model.
The general aims of this study were threefold. First, to develop a simple dynamic model of the cardiovascular system that was of sufficient complexity to account for the major mechanisms by which drugs can alter cardiovascular variables. Second, to examine whether recirculatory kinetic models  have sufficient physiological basis to account for drug related blood flow changes. Third, to examine approaches for identifying the important concentrations (and their sites in the body) that can be used to link the kinetic and dynamic models.
The specific aim was to use previously published data collected using a chronically instrumented sheep preparation [8, 9] to develop a kinetic-dynamic model for the cardiovascular effects of magnesium. Magnesium is given intravenously to treat a number of diseases, including pre-eclampsia. It relaxes smooth muscles in blood vessels thereby lowering systemic vascular resistance, with a consequent decrease of mean arterial blood pressure and increase in cardiac output. It provides a useful drug for initial analysis as its kinetics and dynamics are relatively simple and well understood.
The overall hypothesis of this work is that it is possible to construct a faithful model of the cardiovascular effects of drugs such as magnesium. While doing so requires more assumptions and estimates of parameter values than normally associated with semi-empirical pharmacokinetic-pharmacodynamic modelling, a physiological approach greatly increases the utility of the resulting models. It is proposed that the general methods presented here could be applied to the development of similar models for other drugs with acute cardiovascular effects.
With respect to devising a pharmacodynamic model of the cardiovascular system, the important steps are:
1. Identifying which cardiovascular variables (e.g. heart rate, blood pressure) are important. This depends on the drug and the intended use of the model, but it is proposed that there is a minimum set of variables that is needed for a basic description of cardiovascular status.
2. Devising a way of conveniently presenting the output of the dynamic model for a range of cardiovascular variables for comparison with data.
3. Identifying a cardiovascular model of the appropriate complexity. Ideally the model must be of the minimum complexity that includes the cardiovascular variables identified above, and the major sites of action of the drugs.
4. Identifying which parameters of the cardiovascular model can be estimated by curve-fitting, and which require prior estimates or measurements of physiological values. Most cardiovascular models are stiff numerical systems with many parameters, and only a small number can be estimated by curve-fitting the data in the traditional way.
With respect to the pharmacokinetic model, there remains one crucial step:
5. Constructing a kinetic model with a physiological basis that is sufficiently realistic to describe and predict the concentration of the drug in the key target organs controlling the cardiovascular system. On first principles, these could be expected to include:
a. the myocardial concentrations when the drug has a direct myocardial effect (e.g. causes myocardial depression);
b. the CNS concentrations when the drug affects the cardio-respiratory control centre of the brain;
c. the arterial blood concentration when the drug affects baro-receptors or smooth muscle in the walls of the arterial vascular system;
d. the venous blood concentration when the drug affects smooth muscle in the wall of the capacitance vessels of the venous vascular system.
It is known that these concentrations can follow different time-courses, particularly after bolus administration or a change of infusion rate [10, 11]. However, it may not be necessary to know the time-course of these concentrations for every drug, depending on its mechanism of action.
Data sets and software
The data used to construct the model were collected in the same laboratory using a conscious chronically instrumented sheep preparation and have been published previously [8, 9]. This facilitated the model building process, as the effect of differences in species and measurement methods could be discounted.
Data set 1  (for model development) was a detailed set of cardiovascular measurements made after the administration of 30 mmol of magnesium over 2 min to 5 sheep. Measurements included arterial and coronary sinus (effluent from the heart) magnesium concentration, cardiac output, mean arterial blood pressure, heart rate, an index of myocardial contractility (Maximum positive rate of change of left ventricular Pressure, dp/dt) and an index of filling pressure (Left ventricular end diastolic pressure) and myocardial blood flow. These were made until 25 min after the start of administration.
Data set 2  (for model validation) was a less comprehensive set of cardiovascular measurements made after the administration of 30 mmol of magnesium over 5 min to 5 sheep (not the same sheep as Data set 1). Measurements included arterial magnesium concentration, cardiac output, and mean arterial blood pressure, and were made until 25 min after the start of administration. The blood pressure data for one animal in this data set was excluded, as it was idiosyncratically low.
The time-course of the data averaged across sheep were used for all modelling – the resultant model therefore represents the response of the average sheep. Inter- and intra-animal variability were not considered, although it is noted that the final model may provide insight into sources of kinetic and dynamic variability for later study.
The software used was the Scientist for Windows program (Version 2.01, Micromath, Salt Lake City, Utah, USA), predominantly for curve-fitting. The R language, Version 1.9.0,  was used for graphical data analysis, data handling and simulations. Coding the same model in the two different programs provided a useful check for errors.
All data points were weighted equally. A parameter was arbitrarily defined as non-identifiable if the standard deviation of the parameter returned by the fitting program was greater than the parameter estimate (i.e. the coefficient of variation was greater than 100%). A model with non-identifiable parameters means that the data do not contain sufficient information to estimate the parameter with precision.
The symbols used throughout have been based on standards for the pharmacokinetic literature. Unfortunately, the use of C for both concentration (in pharmacokinetics) and compliance (in cardiovascular physiology) creates of conflict for this paper. To avoid confusion, CPL will be used for compliance here.
Pharmacodynamic model of the cardiovascular system
Identification of important cardiovascular variables
The choice of the cardiovascular variables used in the model is clearly dependent on the drug under study and the intended purpose of the model. However, we propose that a minimum of 7 fundamental cardiovascular variables is sufficient for most pharmacological purposes. These variables are shown with their default unit of measurement in the model: Central venous pressure (CVP, mmHg), Myocardial contractility (CNT, L mmHg-1), Stroke volume (SV, L), Heart rate (HR, min-1), Cardiac output (CO, L min-1), Systemic vascular resistance (SVR, Resistance units, RU) and mean arterial blood pressure (MAP, mmHg).
This choice of variables requires several assumptions:
All variables are time averaged in that beat to beat variation is ignored (e.g. mean arterial blood pressure is used rather than systolic and diastolic blood pressures).
That the function of the left and right side of the heart is the same, and there are no abnormalities in the pulmonary vasculature so that the heart-lung system can be treated as one pump.
Long time-scale events such as fluid shifts and renal mechanisms controlling blood pressure are ignored.
Mathematically, CNT must therefore have the units of volume / pressure. However, contractility is difficult to measure in vivo, and that there are a number of surrogate measures including dp/dt. These can also be used with appropriate scaling factors.
That there are no factors affecting the relationship between myocardial fibre length (the true determinant of stroke volume) and central venous pressure (e.g. changes in myocardial compliance). CVP is therefore used as an easily measured index for myocardial fibre length – the assumption is that the two are related using a scaling factor. Left ventricular end diastolic pressure is also as an alternative index when data are presented as percent change from baseline.
SVR therefore has the units of pressure over flow. In this paper, the resistance units (RU) are therefore mmHg L-1 min.
Presentation of relationships between cardiovascular variables
The effect of a drug on one or two variables can usually be summarised on a plot of the variable (drug effect) against time. However, it is more difficult to summarise the dynamic effect of a drug on the cardiovascular system for the following reasons: First, the large number of variables required in the summary, where the seven described above could be considered a minimum. Second, the fixed inter-relationships (e.g. Eqns. 2–4) between these variables that should be revealed by the summary (e.g. if SV increased by 25% and all else remains the same, then CO should also increase by 25% (Eqn. 3)). Third, usually an analysis requires comparing one cardiovascular state (e.g. pre-drug) with another (e.g. post-drug), or examining the time-course of drug effects.
Cardiovascular model – Structure and parameter estimation
There are many published models of the cardiovascular system of various levels of complexity and intended for various tasks . However, in this paper, the cardiovascular model was constructed progressively from first principles, with adaptations and increases in complexity as dictated by the requirements of the modelling process and the data. This ensured the model was the minimum that was needed for the task at hand.
In vivo, the cardiovascular system has two major control systems; control of cardiac output via the Frank-Starling mechanism, and control of blood pressure via baroreceptor control of heart rate. These were added progressively to the model.
A simple Frank-Starling model
The solution to the simple Frank-Starling model can be found algebraically, but for consistency is shown in Additional File 1 as differential equations.
Baseline (pre-drug) cardiovascular variables. A set of target values that was representative of the sheep studied in our laboratory was compiled from previous measurements and literature values as indicated. A set of parameter values for the final (Constrained-Frank-Starling-Baroreceptor) was derived (Table 2) that produced an internally consistent model that closely replicated these target values (also shown for comparison).
Target value origin
Model derived Value
Central venous pressure
unpublished previous measurements and literature 
inferred from Vblood & CVP
Mean arterial pressure
previous measurements 
Systemic vascular resistance
calculated from CO & MAP
previous measurements 
previous measurements 
calculated from CO & HR
previous measurements [9, 21]
Sympathetic tone – chronotropy
scaling factor only
Sympathetic tone – Contractility
scaling factor only
Frank-Starling and Baroreceptor model
Constraining the model to increase physiological plausibility
Baseline values for the cardiovascular model
Baseline model parameters The parameters chosen as those producing representative baseline (pre-drug) cardiovascular variables (Table 1). The co-efficient of variation (CV (%)) of these parameter values as determined by the Monte-Carlo sensitivity analysis is also shown.
Ratio of venous over arterial compliance
Systemic vascular resistance
Mean arterial pressure set point
Gain for heart rate control
The sensitivity of the baseline cardio-vascular model to changes in parameter values was determined via Monte-Carlo simulation . Multi-variate normally distributed noise was added to the parameter values for a series of 10,000 simulations of the resulting cardiovascular variables. Those parameter sets that produced a set of cardiovascular variables within 10% of the target set were selected and analysed for with respect to parameter variability and correlation.
Fitting the cardiovascular model to the magnesium data
Changes in cardiovascular variables with the administration of magnesium were analysed as percentage change from baseline. This removed the contribution of inter-animal variability in baseline cardiovascular variables (which was nevertheless minor [8, 9]) to variability in the cardiovascular effects of magnesium. The analysis involved fitting cardiovascular radar plots to the measured magnesium data (Data set 1) for key time-points (1, 2, 4, 10 and 25 min) during and after magnesium administration. The cardiovascular model was parameterised in terms of primary cardiovascular variables that could be directly influenced by magnesium. These were SVR, CPLv, CPLratio, CNT, S1 and S2. Vblood could also be considered a primary variable, but it was considered unlikely that magnesium could change the blood volume. The remaining cardiovascular variables were considered secondary in that they would change in response to changes in the primary variables as given by Eqns 2 to 4.
Initially, the only primary parameter fitted to the data for each time point was SVR while the other parameters were held constant. This was based on the prior knowledge that this was the primary mechanism of action of magnesium. If the MSC was low and the cardiovascular radar plot showed a poor fit between model predictions and the data, an additional parameter was fitted one at a time from the remaining parameters listed above. A parameter was removed from the fit if it produced an undefined estimate. The parameter was kept in the fit if it improved the MSC and the fidelity of observed vs. predicted plots on the cardiovascular radar. By this process, the values of the primary cardiovascular parameter at each key time point required to describe the observed data were determined.
Recirculatory pharmacokinetic model of magnesium disposition
Key points during the model development process were: 1) The representation of the lungs as a single compartment. 2) The representation of the cardiac output change as an empirical forcing function (see Fig. 10, this would later be replaced by the predictions of the final cardiovascular model). 3) The representation of the body as extracellular and intracellular spaces connected by a permeability term, in keeping with the known slow cellular uptake of magnesium. 4) The clearance of magnesium is renal, but it can be reabsorbed or excreted in the tubules, as dictated by homeostatic requirements . Thus, renal clearance may be variable.
To confirm that the kinetics of magnesium were cardiac output dependent, the final kinetic model was subjected to a sensitivity analysis for this parameter. Cardiac output was assigned values of 2, 4, 6, 8 or 10 L min-1 while the other parameters were fixed at their best fit value. The time-course of the arterial magnesium concentration was recorded in each case.
Linking the pharmacokinetic and pharmacodynamic models
The relationship between the key cardiovascular parameters (effects) and the concentrations of magnesium in arterial and coronary sinus blood were examined using hysteresis plots (effect vs. concentration). A concentration-effect relationship was considered plausible if produced a predictable relationship with minimal hysteresis that was consistent with the known mechanisms of action of the drug.
The fitted primary cardiovascular parameters for Magnesium data set 1 Units are as for Table 2. The parameter estimates are given with the standard deviation returned by the curve-fitting program. S1 could not be reliably fitted to the data.
Magnesium had little effect on myocardial contractility (Fig. 11C), and the linking function assumed that CNT remained at baseline values. Magnesium appeared to increase the sympathetic tone coefficient for contractility (S2) by approximately 25% at between concentrations of 2 and 4 mmol L-1 (Fig. 11D). However, this rise in S2 only occurred late in the study (Table 3). It indicates subtle changes in the relationship between the filling pressure index (LVEDP) and the contractility index (dp/dt). This may reflect measurement error in these variables, non-stationarity in the experimental preparation or subtle delayed changes in myocardial compliance caused by magnesium. However, it was found that a link function assuming S2 remained at baseline values (Fig. 11D) was an adequate account of the data and did not compromise the predictive power of the model in the validation stage.
Validation of the final model
The final kinetic-dynamic model developed using Data set 1 was used to predict the arterial magnesium concentrations, cardiac output and mean arterial blood pressure for Data set 2. Data set 2 differed from Data set 1 in that the dose of magnesium was given over 5 min instead of 2 min. Consequently, although the dose was the same, the cardiovascular effects were less pronounced. For example, the lowest blood mean arterial pressure for Data set 1 was 76% of baseline, while for Data set 2 this was 86% of baseline. The only change made to the parameters of the final model was to alter the duration of infusion of the magnesium.
Parameter sensitivity of cardiovascular model (baseline conditions)
The baseline cardiovascular variables and the parameters that produced them are summarised in Tables 1 and 2, respectively. Of the 10,000 random parameter sets examined in the Monte-Carlo sensitivity analysis, only 37 produced a set of cardiovascular variables that was within 10% of the target cardiovascular variables. The variability of these successful parameter values was low (Table 2), and the spread of each parameter showed a unimodal, approximately normal distribution. This suggests that there was a unique set of parameter values for the model that was consistent with normal baseline physiology. Visual inspection showed no obvious correlation between parameter values, except for CPLv and CPLratio (correlation coefficient = 0.83). This suggests that specifying the value for one of these parameters significantly constrains the value that can be taken for the other, as would be expected on physiological grounds. It can be concluded that each parameter had an important role to play in the model, and that each could only take a limited range of values to be consistent with the required baseline physiology. By extension, the assumptions regarding the values of these parameters are likely to be appropriate. Furthermore, the changes in these parameters observed following magnesium administration therefore reflect the effects of this drug rather than uncertainty in the parameter space of the model.
Parameter estimates – cardiovascular data
Parameter estimates – pharmacokinetic data
The fitted pharmacokinetic parameters for the Magnesium data set 1 The parameter estimates are given with the standard deviation returned by the curve-fitting program.
A feature of recirculatory pharmacokinetic models is that their initial kinetics are governed by first-pass passage of drug through the lungs, and the dilution of the injected drug with the cardiac output . The cardiac output sensitivity analysis for the model confirmed this behaviour for magnesium (Fig. 10, bottom). This reinforces the need for a common cardiac output term for the cardiovascular and recirculatory kinetic model (Fig. 12). The resultant final model therefore accounts for the fact that by altering cardiac output, magnesium alters its own kinetics.
Relating the estimated cardiovascular parameters in Table 3 to the concurrent arterial concentrations produced the concentration-effect curves shown in Fig. 11. Link functions were established for SVR and CPLv, but not CNT or S2. The overall role of the link functions is summarised in Fig. 12.
Model validation – pharmacokinetic component
Model validation – pharmacodynamic component
The final pharmacodynamic model was able to accurately predict the time-course of the cardiac output changes observed for the validation Data set 2 (Fig. 13). The mean prediction error was 3.0%.
The dynamic model captured the general trend of the mean arterial blood pressure for the validation data (Fig. 13), but some systematic deviations were evident. The model was accurate until the end of the infusion, but thereafter slightly over-estimated the rate of recovery of blood pressure. However, the model did predict that the drop in blood pressure would be considerably less for a 5 min versus 2 min infusion, and the overall magnitude of the changes in blood pressure for the 5 min infusion were small (less than 10% change). The mean prediction error was 6.1%.
Concentration-effect relationships and recirculatory models
In this paper, all cardiovascular effects were related to the arterial concentration of magnesium. As covered in the introduction, there may be other sites in the body that have a theoretical claim to being the most appropriate link concentration for certain cardiovascular dynamic effects. For example, the reductions in myocardial contractility caused by thiopental have been shown to have a better temporal relationship to the thiopental concentrations in the myocardium itself rather than in arterial blood . This consistent with a direct thiopental effect on the myocardium.
In recirculatory models, it is possible to add a "target organ" to represent organs such as the heart . The fact that this was not necessary for magnesium may be the exception rather than the rule. As magnesium has small volumes of distribution, there is little difference in the time-course of the arterial and regional venous concentrations. Furthermore, the predominant effects of magnesium were directly on blood vessels (arterioles for SVR and large veins for capacitance) in direct equilibrium with blood rather than organs such as the heart or brain. Thus, a "systemic" recirculatory model was sufficient for magnesium. As other drugs are studied using this method, data on target organ kinetics and their incorporation into the kinetic model may be necessary.
There are a number of limitations of this modelling approach, many of which are inherent in the assumptions made in the construction of the model. Other limitations may become apparent if the model is used outside of the range of the data used to develop the model. For example, the CL term in the kinetic model was very low (Table 4). This may reflect extensive tubular re-absorption, but may also reflect the fact that the concentrations were followed for only 25 min in the original paper (the time by which most cardiovascular variables had returned to baseline). Studies of a longer duration would help to define this clearance term better.
The cardiovascular model also assumes an instantaneous baroreceptor response. While it is relatively easy (in modelling terms) to add a delay to this response, this was not supported by the data. However, if the model is extended to situations with very rapid blood pressure changes (e.g. orthostatic hypotension) this deficiency may become significant.
Constructing physiologically based models, even of the simplicity presented here, requires crossing many decisions points where a choice must be made from multiple options – sometimes the choices are data driven, sometimes theory driven, sometimes the subjective experience of the model maker must be called upon. While a "wrong" model is evident because it does no match the data, there is clearly no "right" model of the cardiovascular system. It is anticipated that more limitations of the cardiovascular dynamic model will become apparent when model is rigorously compared to data for other drugs, and for other cardiovascular scenarios. It is should be expected that the model will continue to evolve as these data are collected and analysed.
The combination of the recirculatory kinetic model and the simple cardiovascular dynamic model was able to describe and predict the concentrations and cardiovascular effects of magnesium in sheep. It is proposed that the general methods used here could be applied to other drugs with cardiovascular effects. The authors are currently applying the method to intravenous anaesthetics.
Cardiovascular term Description Frank-Starling model
Pv Central venous pressure
Pa Mean arterial pressure
Mean circulatory pressure
Volume of blood in arterial compartment
Volume of blood in venous compartment
Volume of blood in arterial compartment at MCP
Volume of blood in venous compartment at MCP
Ratio of venous over arterial compliance
Systemic vascular resistance
Cardiac output (left side)
Cardiac output (right side)
unit conversion factor – contractility
additional for Frank-Starling-Baroreceptor model
Mean arterial pressure set point
Gain for heart rate control
additional for Constrained-Frank-Starling-Baroreceptor model
Pressure in arterial compartment when stressed
Pressure in venous compartment when stressed
Volume in arterial compartment when stressed
Volume in venous compartment when stressed
slope for arterial pressure-volume relationship
intercept for arterial pressure-volume relationship
slope for venous pressure-volume relationship
intercept for venous pressure-volume relationship
Sympathetic tone coefficient – Chronotropy
Sympathetic tone coefficient – Contractility
maximum for stroke volume-CVP relationship
half-volume for stroke volume-CVP relationship
"Hill factor" for stroke volume-CVP relationship
Pharmacokinetic term Description Description Recirculatory model
doserate of zero order infusion
duration of zero order infusion
Arterial magnesium concentration (total)
Venous magnesium concentration (total)
Apparent distribution volume of the lung
Apparent distribution volume of the body compartment
Permeability-surface area product of deep compartment
Apparent distribution volume of the deep compartment
Supported by the National Health and Medical Research Council of Australia (Project Grant 157952)
- Coleman TG: Mathematical analysis of cardiovascular function. IEEE Transactions on Biomedical Engineering. 1985, 32: 289-294.View ArticlePubMedGoogle Scholar
- Kappel F, Peer RO: A mathematical model for fundamental regulation processes in the cardiovascular system. J Math Biol. 1993, 31: 611-631. 10.1007/BF00161201.View ArticlePubMedGoogle Scholar
- Smith NT, Zwart A, Beneken JEW: Interaction between the circulatory effects and the uptake and distribution of halothane: Use of a multiple model. Anesthesiology. 1972, 37: 47-58.View ArticlePubMedGoogle Scholar
- Fukui Y, Smith NT: Interactions among ventilation, the circulation, and the uptake and distribution of halothane – use of a hybrid computer multiple model: I. The basic model. Anesthesiology. 1981, 54: 107-118.View ArticlePubMedGoogle Scholar
- Tham RQY, Sasse FJ: Large-scale multiple model for the simulation of anesthesia. Advanced simulation in biomedicine. Edited by: Moller DPF. 1990, New York: Springer-Verlag, 173-195.View ArticleGoogle Scholar
- Francheteau P, Steimer JL, Merdjan H, Guerret M, Dubray C: A mathematical model for dynamics of cardiovascular drug action: application to intravenous dihydropyridines in healthy volunteers. J Pharmacokinet Biopharm. 1993, 21: 489-514. 10.1007/BF01059111.View ArticlePubMedGoogle Scholar
- Upton RN: The two-compartment recirculatory model – an introduction to recirculatory pharmacokinetic concepts. Br J Anaesth. 2004, 92: 475-484. 10.1093/bja/aeh089.View ArticlePubMedGoogle Scholar
- Ludbrook GL, James MF, Upton RN: The effect of magnesium sulfate on cerebral blood flow velocity, cardiovascular variables, and arterial carbon dioxide tension in awake sheep. J Neurosurg Anesthesiol. 1999, 11: 96-101.View ArticlePubMedGoogle Scholar
- Zheng D, Upton RN, Ludbrook GL, Martinez A: Acute cardiovascular effects of magnesium and their relationship to systemic and myocardial magnesium concentrations after short infusion in awake sheep. J Pharmacol Exp Ther. 2001, 297: 1176-1183.PubMedGoogle Scholar
- Huang YF, Upton RN, Runciman WB: IV bolus administration of subconvulsive doses of lignocaine to conscious sheep: relationships between myocardial pharmacokinetics and pharmacodynamics. Br J Anaesth. 1993, 70: 556-561.View ArticlePubMedGoogle Scholar
- Ludbrook GL, Upton RN, Grant C, Gray EC: Brain and blood concentrations of propofol after rapid intravenous injection in sheep, and their relationships to cerebral effects. Anaesth Intens Care. 1996, 24: 445-452.Google Scholar
- Ihaka R, Gentleman R: R: A language for data analysis and graphics. J Comput Graphical Stat. 1996, 5: 299-314. [http://www.r-project.org/]Google Scholar
- Torrington KG, McNeil JS, Phillips YY, Ripple GR: Blood volume determinations in sheep before and after splenectomy. Lab Anim Sci. 1989, 39: 598-602.PubMedGoogle Scholar
- Bernillon P, Bois FY: Statistical issues in toxicokinetic modeling: A bayesian perspective. Environ Health Perspect. 2000, 108: 883-893.View ArticlePubMedGoogle Scholar
- Upton RN, Ludbrook GL: A physiological model of induction of anaesthesia with propofol in sheep. 1. Structure and estimation of variables. Br J Anaesth. 1997, 79: 497-504.View ArticlePubMedGoogle Scholar
- Yucha C, Dungan J: Renal handling of phosphorus and magnesium. Nephrol Nurs J. 2004, 31: 33-37.PubMedGoogle Scholar
- Huang YF, Upton RN, Gray EC, Grant C, Zheng D, Ludbrook GL: The effects of short intravenous infusions of thiopentone on myocardial function, blood flow and oxygen consumption in sheep. Anaesth Intensive Care. 1997, 25: 627-633.PubMedGoogle Scholar
- Huang YF, Upton RN, Zheng D, McLean C, Gray EC, Grant C: The enantiomer-specific kinetics and dynamics of verapamil after rapid intravenous administration to sheep: physiological analysis and modeling. J Pharmacol Exp Ther. 1998, 284: 1048-1057.PubMedGoogle Scholar
- Guyton AC, Hall JE: Cardiovascular physiology IV. Edited by: Arthur C Guyton, John E Hall. 1982, Baltimore: University Park PressGoogle Scholar
- Stalberg HP, Hahn RG, Hjelmqvist H, Ullman J, Rundgren M: Haemodynamics and fluid balance after intravenous infusion of 1.5% glycine in sheep. Acta Anaesthesiol Scand. 1993, 37: 281-287.View ArticlePubMedGoogle Scholar
- Huang YF, Upton RN, Rutten AJ: Adverse haemodynamic effects of the rapid intravenous injection of hypotonic solutions in sheep. Res Vet Sci. 1996, 60: 209-212. 10.1016/S0034-5288(96)90040-5.View ArticlePubMedGoogle Scholar
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.