Tissue/blood and tissue/water partition coefficients for propofol in sheep{dagger},{ddagger}

B. M. Q. Weaver1,4, G. E. Staddon2,5 and W. W. Mapleson3

1The Veterinary School, University of Bristol, Langford, Bristol BS18 7DU, UK. 2Department of Medical Physics and Bioengineering, Bristol General Hospital, Bristol BS1 6SY, UK. 3Department of Anaesthetics and Intensive Care Medicine, University of Wales College of Medicine, Cardiff CF14 4XN, UK 4Present address: 79 Sandford Road, Winscombe, North Somerset BS25 1JJ, UK 5Present address: 17 Shaldon Road, Horfield, Bristol BS7 9NN, UK*Corresponding author

{dagger}A provisional analysis of loss of propofol from tissue samples was presented to the Anaesthetic Research Society Meeting, Dundee, July 1998 (Br J Anaesth 1998; 81: 630–1).{ddagger}Declaration of interest: The in vivo work for this study was carried out in the Wellcome Comparative Anaesthetic Laboratory, which was provided by the Wellcome Trust. Financial support was provided by the Perry Foundation and the propofol was generously supplied by Imperial Chemical Industries Limited (now Astra-Zeneca).

Accepted for publication: December 21, 2000


    Abstract
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 Appendix 1
 Appendix 2
 References
 
The primary objective of this study was to determine in vivo tissue/blood partition coefficients of propofol for use in physiological modelling of its pharmacokinetics. The sheep was used as an animal model. In the main series of experiments, crossbred ewes received a bolus of propofol 1% (Diprivan) followed by an infusion during which blood concentrations were measured at intervals. After 2 h, the sheep were killed with an injection of potassium chloride, and tissue samples were taken for storage at –20°C and subsequent analysis. Tissue/blood partition coefficients depend on the amount of triglyceride which accumulates in blood from the propofol vehicle; for blood, free of added triglyceride, the following coefficients were calculated: brain, 3.23; heart, 5.94; kidney, 2.46; spleen, 1.86; semimembranosus muscle, >=1.61; triceps muscle, >=1.47. Calculated tissue/water coefficients were 35 times greater. There was indirect evidence of extraction of propofol by the lungs.

Br J Anaesth 2001; 86: 693–703

Keywords: anaesthetics i.v., propofol; pharmacokinetics, propofol; solubility, partition coefficients; solubility, lipid; sheep


    Introduction
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 Appendix 1
 Appendix 2
 References
 
Physiological, as opposed to empirical, modelling of pharmacokinetics allows prediction of how blood and tissue concentrations of an anaesthetic will change with time in various circumstances but, for this purpose, it is necessary to know how the agent is distributed between the various tissues and organs of the body. Little information is available on this topic with regard to propofol and so, initially, the distribution of propofol within the blood of sheep was studied.1 The present study was carried out (under UK Home Office licences PPL 30/00861 and PIL 30/01083) to determine the uptake of propofol from the blood into the major tissues of the body and to calculate the in vivo tissue/blood partition coefficients.


    Materials and methods
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 Appendix 1
 Appendix 2
 References
 
General plan
Two groups of healthy, adult, crossbred ewes were studied. In group T (six ewes) the objective was to determine tissue concentrations a few minutes after a large bolus dose of propofol. In group P (14 ewes) the objective was to determine tissue/blood partition coefficients. This was done by infusing propofol for 2 h on the assumption that, at the end of the infusion, at least the well-perfused tissues would be nearly in equilibrium with the arterial concentration. Therefore, as well as taking tissue samples, we took blood samples during the infusion to track blood concentration and to determine the arterial concentration at the time of cardiac arrest. All concentration measurements were corrected for loss in storage: the rate of loss from blood samples had already been determined,2 and the rate of loss from the tissue samples was estimated by repeat analyses over a period of time. The tissues sampled were brain, heart, kidney, liver, lungs, spleen and muscle [semimembranosus in all sheep, plus triceps in the partition-coefficient sheep (group P)]. The preparation of propofol used was 1% Diprivan.

Procedure
Caveat
This was an exploratory study so the procedure developed as the investigation progressed, as is evident from the summary in Table 1, where group P has been divided into subgroup Pa, in which arterial cannulation was performed before the propofol was started, and subgroup Pb, in which it was performed during the infusion.


View this table:
[in this window]
[in a new window]
 
Table 1 Summary of the general plan of the investigation
 
Cannulation
In group T, the propofol was injected through a flexible cannula, previously placed into a jugular vein under local anaesthesia.3 In group P, both jugular veins were similarly cannulated, ensuring that the tip of each cannula remained within the cervical region. Propofol was administered through one cannula and venous samples were drawn through the other. An artery was cannulated under general anaesthesia: a digital artery in the first four ewes in subgroup Pa, and a carotid artery in the remaining four and in all six ewes in subgroup Pb.

Euthanasia
In all groups, euthanasia was performed while the animal was still deeply anaesthetized: an intravenous injection of a saturated solution of potassium chloride at the appropriate time (Table 1) rapidly induced cardiac arrest. Within 40 min, the requisite tissue samples were removed, sealed in plastic bags and stored in a freezer at –20°C for subsequent analysis for propofol content.

Group T
A bolus of propofol 10 mg kg–1 induced anaesthesia and ensured that the animals were deeply anaesthetized when, 2–7 min later, the potassium chloride was injected. The tissue samples were analysed on three to five occasions between 6 and 26 months post mortem.

Subgroup Pa
In order to obtain arterial samples throughout the infusion, cannulation was performed under an alternative form of anaesthesia before any propofol was administered. Induction was with thiopental in four animals, halothane in three and isoflurane in one; in each sheep, maintenance was with the same volatile agent, or with isoflurane in the sheep with thiopental induction. Once each animal was anaesthetized, its trachea was intubated with a cuffed tracheal tube which was connected to a circle-absorber breathing system, supplied with about 300 ml min–1 of oxygen and incorporating an in-circle vaporizer. As soon as the cannulations were complete, a bolus of propofol 3 mg kg–1 was administered, followed by an infusion of the agent at 1 ml min–1 for 2 h. The in-circle vaporizer was used to add the same volatile agent as before the propofol, at concentrations required to ensure adequate anaesthesia, as judged clinically, throughout the procedure. Paired sampling of arterial and venous blood was scheduled for 1, 3, 5, 7, 10, 15, 20, 30, 45, 60, 75, 90, 105 and 120 min after the start of the propofol infusion. Euthanasia followed while the infusion continued. Tissue samples were analysed on two or three occasions between 1 and 330 days post mortem; blood samples were analysed once between 1 and 31 days.

Subgroup Pb
Here it was realized that it would be sufficient to obtain arterial samples during only the later part of the infusion; therefore the arterial cannulation was performed after inducing anaesthesia with a larger bolus of propofol 6 mg kg–1 followed by the 2-h infusion. In the first two ewes, the rate of infusion was 1 ml min–1, as in subgroup Pa; in the last four, the rate was 2 ml min–1 to provide additional information on propofol kinetics. As soon as possible after commencing the infusion, the trachea of each animal was intubated and connected to the circle breathing system. In all six sheep, the propofol was supplemented with isoflurane, from the in-circle vaporizer, according to the criterion used for subgroup Pa. Venous sampling was as in subgroup Pa; in addition three to six paired arterial and venous samples were taken in the last 10 min before cardiac arrest was induced at 123–152 min. Tissue samples were analysed once only, 9–106 days post mortem; blood samples were analysed once at 3–50 days.

Liver samples
In all but the first of the 14 ewes in group P, two samples of liver were taken. The ‘early’ sample was taken as soon as possible after cardiac arrest and placed in the freezer within 10–20 min of the arrest; the second, ‘standard’ sample was taken to the freezer 10–25 min later, along with the samples of the other tissues. This was done to provide information on possible loss by continuing metabolism of propofol in the liver before freezing.

Analysis of propofol in blood and tissue
Analysis of propofol in blood samples was as described previously2 with derivatizations4 of two 1-ml subsamples, each analysed in duplicate. For analysis of propofol in tissue, a subsample of each frozen tissue sample was taken by grating and the remainder returned to the freezer. The grated subsample was chopped, and further broken down with a pestle and mortar to facilitate homogenization. It was then weighed and potassium dihydrogen phosphate (0.1 M) was added to produce a 1:2 w/v suspension. This was homogenized in a liquidizer, with only short bursts of activity to avoid heating, which would have denatured the tissue. Each subsample was then treated in the same way as the blood samples, each derivatization using 1.5 ml of homogenate.

Data processing
Processing was commenced only when all the data had been collected; a detailed account is given in Appendix 1. Analysis of variance was used to determine the best way of summarizing the loss rates from tissue samples. The resulting rates were used to estimate the concentration in each tissue sample at day 0, i.e. the day of sampling. Concentrations in liver samples were additionally corrected for metabolic loss before freezing. Regression analysis was used to estimate the arterial concentration at the time of cardiac arrest in each sheep—directly in subgroup Pa, but in subgroup Pb a mean arterial–venous difference was added to the fitted venous concentration. Partition coefficients were obtained by dividing each day-0 tissue concentration by the corresponding arterial concentration at cardiac arrest. P<0.05 was regarded as significant.


    Results
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 Appendix 1
 Appendix 2
 References
 
Loss from tissue samples during storage at –20°C
The data analyses in Appendix 1 showed that the loss from tissue samples was better represented by a linear decrease in log concentration rather than in actual concentration, implying an exponential decay of propofol content. It also emerged that separate slopes were needed for loss after a bolus (group T) and after an infusion (group P), and for loss from spleen and non-spleen tissues after a bolus (Table 2). Only non-spleen tissue after a bolus showed a significant loss rate (the confidence limits excluded zero). The rates for spleen, and for all tissues after an infusion, were small and non-significant; they were assumed to be zero for calculating the day-0 concentrations.


View this table:
[in this window]
[in a new window]
 
Table 2 Loss of propofol from tissue samples stored at –20°C. Concentration in µg g–1. CL=95% confidence limits; *data from subgroup Pa, assumed to apply to whole of group P
 
Loss from liver samples before storage at –20°C
The mean rate of change of log10(concentration in µg g–1) between the early and standard samples was –0.0134 min–1 (SE 0.0037), or 3% min–1 compound (95% confidence limits 1.2, 4.8% min–1).

Tissue concentrations
Estimated concentrations at the time of cardiac arrest are listed in Table 3 for each tissue and for each method of administration. The estimates are in terms of geometric mean with asymmetrical confidence limits. Estimates for liver are after correcting for loss before freezing.


View this table:
[in this window]
[in a new window]
 
Table 3 Tissue concentrations of propofol (µg g–1) in each tissue for each method of administration, geometric mean (95% confidence limits); ratios of the means for the two rates of infusion; pooled SD of log10(concentration in µg g–1); perfusion of each tissue derived from Pearson.5 concn=concentration; Semimem=semimembranosus; *pooled between the two infusion rates only; the SD of log10(concentration) for muscle in the bolus group was 0.585
 
Blood concentrations
In subgroup Pa sheep, infused at 1 ml min–1, arterial concentrations generally increased slowly after the transient produced by the initial bolus, but relatively rapidly in ewe 37 (Fig. 1, top). However, to maintain deep anaesthesia, inhalation supplementation was required in all cases (including ewe 37) after an average of 10 min. When propofol was infused at 2 ml min–1, the venous concentrations (the only ones measured until near the end of the infusions in subgroup Pb) were greater than those at 1 ml min–1 (Fig. 1, bottom) and increased much more rapidly. Despite this, supplementation with an inhalational agent was required in the 2 ml min–1 infusions after approximately 1 h.



View larger version (22K):
[in this window]
[in a new window]
 
Fig 1 Concentrations of propofol in blood, corrected for loss in storage, plotted against time. Top: arterial concentrations in individual sheep of subgroup Pa, infused at 1 ml min–1. Bottom: mean of regression lines fitted (from 15 min onwards) to arterial (continuous line) and venous (broken lines) concentrations in each sheep of subgroup Pa, infused at 1 ml min–1, and to venous concentrations in each of those subgroup Pb sheep infused at 2 ml min–1. The slopes of the fitted lines are [mean (SE)]: for subgroup Pa sheep infused at 1 ml min–1 (n=7, i.e. excluding ewe 37) 0.55 (0.21) and 0.55 (0.25) µg ml–1 h–1 for arterial and venous concentrations respectively; and, for subgroup Pb sheep infused at 2 ml min–1 (n=4, venous concentrations only), 3.4 (0.70) µg ml–1 h–1. The corresponding residual standard deviations were 0.72, 0.65 and 1.54 µg ml–1.

 
Partition coefficients
From this point onwards the appropriate split of the group P sheep is not into the procedure-based subgroups Pa and Pb, but according to the two rates of infusion: 10 sheep at 1 ml min–1 and four at 2 ml min–1. Corresponding partition coefficients for each tissue, for each infusion rate, are listed in terms of geometric mean and asymmetrical confidence limits in Table 4. It should be noted that, although many tissue concentrations in the 2 ml min–1 sheep were nearly double those in the 1 ml min–1 sheep (Table 3), the arterial concentrations after 2 h were more than double (Fig. 1, bottom), so the partition coefficients are systematically, and significantly (P<0.001), somewhat smaller in the 2 ml min–1 sheep.


View this table:
[in this window]
[in a new window]
 
Table 4 Raw tissue/blood partition coefficients ({lambda}) for each tissue for each infusion rate, geometric mean (95% confidence limits); ratios of the two means for each tissue; pooled standard deviation of log10({lambda}); tissue densities. Tissue densities, used to convert tissue concentrations from µg g–1 to µg ml–1, are means (SDs) of those for 2–7 mammalian species other than sheep6 because no values could be traced for this species. Semimem=semimembranosus. Coefficients for the muscles should be regarded as lower limits (see Discussion)
 

    Discussion
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 Appendix 1
 Appendix 2
 References
 
Tissue storage
The pattern of loss of propofol from tissues during storage (Table 2) is surprising: substantial loss from non-spleen bolus tissues, but small and non-significant changes in the remainder. It does not seem possible to attribute the pattern to experimental errors because the methods of storage and analysis were the same throughout the study, and the analyses of bolus and infusion tissues overlapped in time. Propofol is known to evaporate but its boiling point is 242.3°C (J. B. Glen, personal communication) and, even if evaporation were the method of loss in storage, half-life should be proportional to the solubility of propofol in the tissues; yet there was nothing exceptional about spleen in terms of partition coefficient (Table 4).

Another possibility is degradation of propofol. Some evidence can be adduced that the distribution of the propofol vehicle may match the pattern of loss in Table 2. An earlier paper3 noted that the vehicle is very similar to Intralipid: a major component of both is soya bean oil. For convenience, the vehicle will be referred to below as ‘intralipid’, with a lower-case ‘i’. The earlier paper3 reported plasma concentrations of triglyceride. Those concentrations arose from measurements in some of the present sheep and in other crossbred ewes subjected to similar boluses and infusions of propofol. The initial increases in plasma concentration shortly after the injection of a bolus of propofol were sufficient to account for all the injected intralipid. Therefore, the amount of intralipid in each tissue would depend only on the amount of plasma it contained. Again there is nothing exceptional about the spleen when it is contracted but, when maximally engorged it contains about 12 times as much plasma as any other tissue (Table 5), and five times as much at the degree of engorgement estimated for the present studies after a bolus of propofol (Table 5). This implies that, in group T, the spleen contained five times as much intralipid as any other tissue.


View this table:
[in this window]
[in a new window]
 
Table 5 Residual plasma mass in sheep tissues as percentage of blood-free tissue mass.5 *Estimated from an equation5 relating plasma and spleen packed-cell fractions to the sequestration of red cells in spleen, and using measurements of the decrease of packed-cell fractions in the first minute after a bolus of propofol in group P and in other1 sheep
 
At the end of a 2-h infusion, the plasma triglyceride concentrations had increased greatly3 but accounted for only half the total amount injected in the infusions. This is consistent with tissue samples in the infusion studies containing substantial amounts of at least some components of intralipid. Therefore perhaps the pattern of loss rates in Table 2 can be attributed to some component of intralipid having a preservative action on propofol.

This hypothesis is clearly speculative but, when these findings on loss rate were presented to the Anaesthetic Research Society,7 none of the alternative hypotheses elicited from the members was consistent with all the observations.

The loss rate from liver during storage at –20°C was no different from that for other non-spleen tissues. Therefore, there was no liver-specific metabolism once the samples were frozen.

Tissue distribution
There is no ready means of knowing whether the concentrations for the bolus-only method (Table 3) were before, at or after the postinjection peak, except that the huge perfusion of the lung will ensure that the measured concentration relates to well after its first-pass peak. However, the concentrations were generally greater for the well-perfused tissues than for muscle and the concentrations in the former were greater than the corresponding concentrations in the infusion experiments.

After a 2-h infusion, well-perfused tissues should all be nearly in equilibrium with the arterial concentration at the time of cardiac arrest. The highest concentration was then found in the heart and was about double that in any other tissue. The smallest concentration was found in the lungs, where ‘extraction’ of propofol has been reported in sheep;8 but the concentration in liver, where metabolism is known to occur, was comparable to that in other tissues (Table 3). Perhaps there was an even greater loss from lung tissue before freezing. Support for extraction by the lungs is possibly indicated by the pooled standard deviation for lungs (Table 3) being even greater than that for liver, and three to four times greater than for any other tissue. Further evidence of extraction by the lung may be hidden in the anomalous behaviour of ewe 37 (see below).

Partition coefficients
Raw tissue/blood coefficients
The partition coefficients in Table 4 arise from dividing all the fitted day-0 tissue concentrations (converted to µg ml–1) for each sheep by the corresponding estimated arterial concentration at the time of cardiac arrest. The differences between tissues therefore follow the same pattern as the tissue concentrations in Table 3. The systematic differences between the two infusion rates are related to the triglycerides from the propofol vehicle, which accumulate in blood during the infusions—more so at 2 ml min–1 leading to smaller partition coefficients. Therefore, the coefficients in Table 4 are relevant only to the particular circumstances of those infusions.

Tissue/water and tissue/normal-blood coefficients
A set of more generally applicable tissue/water coefficients was derived as follows. From data previously reported1 3 or obtained from published work,911 it was found possible to derive a general equation (Appendix 2, equation 12) for the blood/water partition coefficient as a function of the triglyceride concentration in plasma and of packed-cell fraction. This was then quantified for the present study in equations 16 and 17, to give the blood/water coefficient as a function of time for each sheep. Tissue/water coefficients could then be derived by multiplying each tissue/blood coefficient by the blood/water coefficient for the end of the corresponding infusion. Unlike in Table 4, differences between the mean tissue/water partition coefficients for the two infusion rates were not significant, so overall geometric means and confidence limits were calculated. These are displayed in Table 6, together with a set of tissue/normal-blood partition coefficients. These were obtained by dividing each tissue/water coefficient by the blood/water coefficient for ‘normal’ sheep blood (35.0; Appendix 2, equation 13).


View this table:
[in this window]
[in a new window]
 
Table 6 Tissue/water partition coefficients as geometric mean (95% confidence limits) derived by multiplying each tissue/blood coefficient by the estimated blood/water partition coefficient applicable at the time of cardiac arrest in each sheep (Appendix 2, equations 16, 17); coefficients for the partitioning between tissues and normal sheep blood derived by dividing each tissue/water coefficient by the blood/water coefficient for normal blood (35.03, Appendix 2, equation 13); pooled standard deviation of log10({lambda}). {lambda}=partition coefficient (tissue/water or tissue/normal-blood). Values apply to both infusion rates; coefficients for the muscles should be regarded as lower limits (see Discussion)
 
These tissue/normal-blood coefficients are somewhat artificial because they will be changed as soon as any propofol is administered, but they may be useful for comparison with corresponding values for other drugs: they are broadly comparable to values for inhalation anaesthetics.12 They are mostly about 30% higher than those in Table 4 for 1 ml min–1, and about twice those for 2 ml min–1.

Possible systematic errors
Both sets of coefficients (Tables 4 and 6) are susceptible to systematic errors arising from incomplete equilibration between tissue and blood. A cup-and-bucket simulation method was devised for correcting for this (details are available from the authors). It was applied to each individual coefficient, and a geometric mean was calculated for each tissue. For the well-perfused tissues, the incompleteness was no more than 1% except for liver (4% because of one very large tissue/blood coefficient). Therefore, these coefficients would be little influenced by errors in the assumed perfusions (Table 3). Muscles showed about 20% incompleteness, but this value is somewhat uncertain because the results were then critically dependent on the perfusion assumed. Therefore, the coefficients given for muscle should be regarded simply as lower limits. In addition, the coefficients for liver and lungs will always be low because in vivo metabolism and extraction8 respectively prevent these tissues ever reaching equilibrium with blood.

There are three further sources of uncertainty in the partition coefficients. First, volatile anaesthetics can affect the binding of drugs to proteins, but for halothane and isoflurane the effect is small (0–16%) at 1 minimum alveolar concentration (MAC)1315 and probably smaller at the <1 MAC needed in our study to supplement the propofol infusion. Secondly, non-linearity of binding of propofol has been reported, but only for low concentrations (1%) of human serum albumin and haemoglobin and not for human plasma.16 In the present study, despite the nearly twofold differences in mean tissue concentrations between the two infusion rates (Table 3) the corresponding differences in mean tissue/water coefficients were not significant and no more than 14% for the non-metabolizing, well-perfused tissues (brain, heart, kidney, spleen). This suggests little non-linearity in the present circumstances. Finally, we estimate that uncertainties in some of the parameters used in Appendix 2, taken from earlier work,1 might change the tissue/water coefficients by ±20% and the tissue/normal blood coefficients by ±10%.

Other sources of partition coefficients
Ludbrook and colleagues17 give a plot of calculated brain concentrations of propofol in sheep against measured sagittal-sinus concentration during and after a 45-min infusion of 1 ml min–1 of propofol. The slope of the plot corresponds to a brain/blood partition coefficient of about 3.5. This is comparable to our estimates of 3.2 (Table 6) for before any propofol and 2.4 for the end of a 2-h infusion of 1 ml min–1 (Table 4). Brain/blood concentration ratios of 3.5 and 2.9 have been found18 for propofol in the rat after 15- and 30-min infusions respectively. Perhaps this decrease with duration of the infusion reflects increasing lipid accumulation in the blood.

The anomalous ewe 37
In this sheep the arterial concentration increased unusually rapidly (Fig. 1, top). This was not associated with an unusually steep increase in plasma triglyceride concentration (ewe 37 was one of those in which triglyceride concentrations were measured).3 In addition, for every tissue in ewe 37, the propofol concentration was greater than in any other sheep: as much as eight times greater than the next largest for lung. On the other hand, each partition coefficient for ewe 37 was within, or almost within, the corresponding range of values for other sheep infused at 1 ml min–1, except for lung, where the coefficient for ewe 37 was three times the next largest, reaching a value of 0.95. This approaches the mean coefficients in Table 4 for many other tissues. This pattern of results suggests that extraction by the lungs was much less in ewe 37 than in the other sheep. It is consistent with the finding that extraction by the lungs was very variable between sheep.8

Possible mechanisms of extraction by the lungs
It seems very unlikely that this extraction is a result of elimination of propofol vapour in the expired gas because of the high boiling point of propofol (242.3°C) (J. B. Glen, personal communication). The alternative of some enzymatic transformation of propofol to a substance not detected by our analysis method, with a near absence of the necessary enzyme in ewe 37, seems more plausible. The non-linear nature of the extraction process19 also points to an enzyme, not evaporation. The presence of an enzyme in the lungs would be ‘logical’ if it converted the propofol to a more volatile form (perhaps propane or propene, boiling points –42 and –47°C respectively), which would then be excreted in the expired gas. Although extrahepatic extraction has been reported in humans,20 21 the only evidence of the lungs being involved is of 10–20% extraction in two of eight patients.22 This may be an indication that enzymatic processes are less transferable between species than are partition coefficients.

Implications for target-controlled infusions of propofol
Target-controlled infusion anaesthesia aims to provide and maintain a preset blood concentration of propofol. Therefore, as the blood/water coefficient increases with increasing triglyceride concentration during an infusion, the concentration in the water phase of blood will decrease progressively; and it is this concentration with which the site of action of propofol equilibrates.

The more recently introduced 2%, and 6%,23 preparations of propofol should lead to less accumulation of triglyceride in blood and hence to smaller blood/water and larger tissue/blood partition coefficients. This will facilitate uptake from blood to tissue, so that a given mg kg–1 min–1 infusion regimen should give smaller blood concentrations than intended but greater tissue concentrations.

In conclusion, this study provides a set of tissue/blood and tissue/water partition coefficients for propofol for some of the major organs—an essential ingredient of the physiological modelling of pharmacokinetics. The study also shows that the physiological modelling of any agent requiring a lipid carrier will be more complex than for other agents. In parallel with this, empirical models may need to be fitted to new observational data obtained when using each new formulation. The study demonstrates that, logically, target-controlled infusion should aim for a constant concentration not in whole blood but in the aqueous phase of blood.


    Appendix 1
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 Appendix 1
 Appendix 2
 References
 
Data processing
Loss of propofol from tissue samples during storage
The investigational unit for loss at –20°C is the sample, so a separate straight line was fitted to the measurements on each standard sample in order to obtain a slope for loss rate in each sample. This was done in a single regression analysis and therefore with a single residual variance. Initially, only those samples with three or more measurements were used. This analysis showed that, with log concentration, as opposed to actual concentration, the fit was better (a straight line accounted for a larger fraction of the variance around the mean for the sample); also the residuals were then more nearly normally distributed.

A second, similar analysis used only log concentration but incorporated all samples with two or more measurements. The resulting individual slopes had large standard errors, about equal to each slope, and so required averaging. A first analysis of variance of the slopes, weighted inversely to the squares of the standard errors of the individual slopes, showed that this needed to be done separately for bolus and infusion data. In a similar weighted analysis of the bolus-only data, there was a significant difference of slope between spleen and the other tissues (P<0.0001), but no significant difference between the other tissues (P=0.7) nor between sheep (P=0.3). The splenic tissue showed a non-significant increase in propofol content during storage (P=0.3). In the infusion data, there was no significant difference between any of the tissues (P=0.8) nor between sheep (P=0.8). The loss rates are summarized in Table 2.

Estimated log concentration in each tissue sample at day 0
For each sample, the log concentration at day 0 was taken to be the intercept of a line fitted to the log measured concentrations for that sample. For non-spleen bolus tissues, the common, fitted slope was used, but for all the other samples zero slope was used. The fitting process was done separately for each tissue sample.

Correction of liver concentrations for loss before freezing
Loss rates from liver samples before freezing, obtained from measurements on the early and standard samples, were equally consistent between sheep whether expressed in terms of log concentration or actual concentration. For consistency with loss at –20°C, a mean loss rate of log concentration was calculated and assumed to apply to each liver sample during the time from cardiac arrest to the freezing of that sample. This led to a set of liver concentrations corrected to the time of cardiac arrest.

Blood concentrations
The measured concentrations were first corrected for an exponentially declining loss in storage of 0.7% per day compound.2

In the infusion experiments, the tissues would be at or approaching equilibrium with the arterial concentration at the time of cardiac arrest. In subgroup Pa (Fig. 1, top), the arterial concentration at cardiac arrest was estimated for each sheep by fitting a straight line to all the measurements (from 15 min onwards to avoid transients after the initial bolus) (Fig. 1, bottom) and extrapolating the line by a few minutes to the time of cardiac arrest (Table 1). In subgroup Pb, the venous concentration at cardiac arrest was similarly estimated for each sheep, and then corrected to an estimated arterial concentration by using the arteriovenous pairs of measurements in the last 12 min of each infusion.

An analysis of variance of the arterial–venous differences in subgroup Pb showed that there were no significant differences between the means for the two infusion rates used in that subgroup, but that there were significant differences between sheep. Therefore, for each sheep, the fitted venous concentration at cardiac arrest was corrected by the addition of the mean arterial–venous difference for that sheep.

Partition coefficients
The above processing provided, for each infusion sheep, the estimated concentration in each tissue at the time of sampling, and the estimated concentration in arterial blood at the time of cardiac arrest. Therefore, partition coefficients were obtained by dividing each tissue concentration by the corresponding arterial concentration. Each tissue concentration was first converted from µg g–1 to µg ml–1 by multiplying by the corresponding tissue density (Table 4).

Summarizing tissue concentrations and partition coefficients
Analysis of variance of tissue concentrations and of partition coefficients (in terms of either actual or log values) showed clear differences between tissues (P<0.0001) for each method of administration. Examination of the residuals showed that, after allowing for different standard deviations in different tissues, the distribution of the residuals was consistent with normality for log concentration and log partition coefficient (P>0.2) but not for actual concentration or partition coefficient (P<<0.001). Within each tissue, variances of log concentration and of log partition coefficient were sufficiently consistent to justify pooling the standard deviations of the log values between methods of administration, except that a separate, larger standard deviation was needed for muscle in the bolus group. Means and confidence limits were then calculated in terms of log values (using the pooled standard deviation) and converted to geometric means with asymmetrical limits (Table 3).

Statistical analyses
These were done with the statistical package GLIM.24 25 The Shapiro-Francia W' test26 was used to test the normality of distributions. All logarithms were to the base 10.


    Appendix 2
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 Appendix 1
 Appendix 2
 References
 
Theory for the calculation of blood/water partition coefficients
Estimation of the concentration of triglyceride in plasma as a function of time
Let C be the concentration of triglyceride in plasma in mmol litre–1. Measurements of such concentrations were made in crossbred ewes, similar to the present ones, injected with a bolus of propofol 6 mg kg–1 followed by a 2-h infusion of 1 or 2 ml min–1.3 The results showed that the normal, pre-bolus concentration of triglyceride was

C0=0.222.

Subsequently the measured concentration could be well represented by an instantaneous step increase followed by a linear increase with time. For the 1 and 2 ml min–1 infusion rates the step increases were respectively

{Delta}Cbol1=1.151

{Delta}Cbol2=1.918.

The slope of the linear trend, in mmol litre–1 h–1, could be represented by

dC/dt=–0.26+0.0151xx

where x is the infusion rate in mg kg–1 h–1.

Diprivan 1% contains 100 mg triglyceride (soya-bean oil)11 per ml. Therefore, an infusion of 1 ml min–1 Diprivan contains 100x60/M mg kg–1 h–1 triglyceride, where M is body mass in kg. Therefore, in terms of mmol litre–1 min–1,

dC/dt=–0.004 33+1.51xI·/M

where I· is the infusion rate of propofol in ml min–1.

Therefore, the total concentration of triglyceride at any time, in mmol litre–1, is given by

C=C0+{Delta}Cbol+(–0.004 33+1.51xI·/M) xt(1)

where t is time in minutes.

In fact, the step increase in the 1 ml min–1 sheep was much the same as in the 2 ml min–1 sheep (1.92 mmol litre–1), but the concentration then decreased rapidly to join the line defined by equation 1. On the other hand, the step increase to be expected from the bolus of 3 mg kg–1 used in most of the present 1 ml min–1 sheep can be expected to produce a step increase of a little less than 1.15 mmol litre–1, followed by a drift up to the line of equation 1. That the equation is indeed reasonably applicable to all the present sheep is shown by the fact that the tissue/water partition coefficients (Table 6), which depend on equation 1, have much the same pooled standard deviation between sheep as the raw tissue/blood coefficients (Table 4).

To convert to fractional concentration (litre litre–1) of triglyceride in plasma, ftr, multiply by 0.861 (mean millimolar mass, g mmol–1) and divide by 877 (mean density, g litre–1):10

ftr=0.000 982x{C0+{Delta}Cbol+(–0.004 33+1.51xI·/M) xt}. (2)

Estimation of the triglyceride/water partition coefficient, {lambda}tr/w, for propofol
Let vx, mx, {rho}x be the volume, mass and density respectively of component x in 1 ml of Diprivan, where x can be p=propofol, tr=triglyceride or a=aqueous phase. Also let mp,x and Cp,x be the mass and mass/volume concentration respectively of propofol in phase x. Then

mp,a=Cp,axva

where

va=1–vtrvp.

Also

Cp,tr=(mpmp,a)/vtr

={mpCp,ax(1–vtrvp)}/vtr.

Therefore the triglyceride/water partition coefficient for propofol is given by

{lambda}tr/w=Cp,tr/Cp,a={mp/Cp,a–(1–vtrvp)}/vtr

where, using published values for mp, mtr,11 Cp,a,9 and {rho}tr,10 and assuming 1 g ml–1 for the density of propofol (it has very little influence on the outcome),

mp=10 mg

Cp,a=0.018 57 mg ml–1

vtr=mtr/{rho}tr=0.1 g/0.877 g ml–1

vp=mp/{rho}p=0.01 g/1 g ml–1.

Therefore

{lambda}tr/w=4715.(3)

Derivation of the blood/water partition coefficient of propofol, {lambda}b/w, as a function of the partition coefficients and fractional concentrations of the components of blood
Let Cn=mass/volume concentration of propofol in n, fm,n=volume fraction of m in n (in plasma where n is omitted), and {lambda}m/n=volume/volume partition coefficient of propofol between m and n, where m and n can be b=blood, c=(red) cell, pl=plasma, pr=protein, tr=triglyceride, w=water and npl=non-protein liquid (water plus triglyceride). Then


which reduces to

{lambda}pl/w=fprx({lambda}pr/w–1)+ftrx({lambda}tr/w–1)+1.(4)

Also

{lambda}b/w={lambda}pl/wx(1–fc,b)+{lambda}c/wxfc,b.

Therefore

{lambda}b/w={fprx({lambda}pr/w–1)+ftrx({lambda}tr/w–1)+1}x        (1–fc,b)+{lambda}c/wxfc,b.(5)

Quantification of the equation for the blood/water partition coefficient
In similar crossbred ewes, measurements on samples of normal blood, spiked with 5 µg propofol ml–1,1 gave fpr=0.0427 (76.4 g protein per litre of plasma ÷ 1.79 g protein per ml protein), {lambda}pr/npl=283 and {lambda}c/pl=1.42. From these data, other terms in equation 5 can be derived as follows:

{lambda}pr/w={lambda}pr/nplx{lambda}npl/w(6)

and

{lambda}c/w={lambda}c/pl x{lambda}pl/w(7)

where


which reduces to


The concentration of triglyceride in plasma from normal sheep has been estimated above: C0=0.222 mmol litre–1=0.191 g litre–1; however, in the spiked blood,1 the 5 µg ml–1 of propofol would be accompanied by 50 µg ml–1 (= 0.050 g litre–1) of triglyceride.11 Therefore, for the spiked blood, the concentration would be 0.191+0.050= 0.241 g litre–1; and, in litre litre–1

ftr=0.000 275.(9)

Then, using equations 9 and 3 in equation 8 gives


and, from equation 6,

{lambda}pr/w=283x2.35=666.(10)

Therefore, in the spiked blood, using equations 10, 9 and 3 in equation 4,

{lambda}pl/w=0.0427x665+0.000 275x4714+1=30.69.

Then equation 7 leads to the generally applicable cell/water partition coefficient,

{lambda}c/w=1.42x30.69=43.58.(11)

Substituting equations 10, 3 and 11 in equation 5 gives

{lambda}b/w=(0.0427x665+ftrx4714+1)x(1–fc,b)+43.58xfc,b

=(29.40+4714x ftr)x(1– fc,b)+43.58xfc,b.(12)

For normal (unspiked) sheep blood, equation 2 reduces to

ftr=0.000 982xC0=0.000 218

and5

fc,b=0.35

whence equation 12 becomes

{lambda}b/w=30.42x(1–0.35)+43.58x0.35=35.03.(13)

During the infusions, the pattern of measurements of packed-cell fraction (fc,b) in all the present sheep can be adequately represented by a very rapid drop from normal values to a constant:

fc,b=0.26.(14)

Substituting equation 14 in equation 12 gives

{lambda}b/w=(29.40x0.74+43.58x0.26)+4714x0.74xftr

=33.09+3488xftr.(15)

Finally, substituting equation 2 in equation 15 gives {lambda}b/w as a function of time for each sheep in the 1 and 2 ml min–1 infusion groups respectively:

{lambda}b/w1=37.79+(–0.0148+5.17/M)xt(16)

{lambda}b/w2=40.42+(–0.0148+10.35/M)xt.(17)


    References
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 Appendix 1
 Appendix 2
 References
 
1 Weaver BMQ, Staddon GE, Raptopoulos D, Mapleson WW. Partitioning of propofol between blood cells, plasma and deproteinised plasma in sheep. J Vet Anaesth 1998; 25: 19–23

2 Weaver BMQ, Staddon GE, Raptopoulos D, Mapleson WW. Measurement of propofol concentration in sheep blood and plasma: effect of storage at different temperatures. J Pharmacol Toxicol Methods 1995; 34: 199–202[ISI][Medline]

3 Weaver BMQ, Raptopoulos D, Staddon GE, Mapleson WW. Plasma triglyceride concentration during intravenous infusions of propofol and Intralipid in sheep. J Vet Anaesth 1996; 23: 15–9

4 Adam HK, Douglas EJ, Plummer GF, Cosgrove MB. Estimation of ICI 35,868 (Diprivan®) in blood by high-performance liquid chromatography, following coupling with Gibbs’ reagent. J Chromatogr 1981; 223: 232–7[Medline]

5 Pearson MRB. The pharmacokinetics of inhaled anaesthetics in the sheep and ox. Thesis, University of Bristol, 1988: 47, 104, 154, 174, 324–8

6 Webb AI, Weaver BMQ. The density of equine tissue at 37°C. Res Vet Sci 1979; 26: 71–5[ISI][Medline]

7 Weaver BMQ, Staddon GE, Mapleson WW. A puzzle in propofol loss rates from tissue samples in sheep. Br J Anaesth 1998; 81: 630P–1P[ISI][Medline]

8 Mather LE, Selby DG, Runciman WB, McLean CF. Propofol: assay and regional mass balance in the sheep. Xenobiotica 1989; 19: 1337–47[ISI][Medline]

9 Babl J, Doenicke A, Mönch V. New formulation of propofol in an LCT/MCT emulsion: approach to reduce pain on injection. Eur Hosp Pharm 1995; 1: 15–22

10 Lide DR, editor. Handbook of Chemistry and Physics, 77th Edn. Boca Raton: CRC Press, 1996; 3–184, 227, 229

11 Tan CH, Onsiong MK. Pain on injection of propofol. Anaesthesia 1998; 53: 468–76[ISI][Medline]

12 Steward A, Allott PR, Cowles AL, Mapleson WW. Solubility coefficients for inhaled anaesthetics for water, oil and biological media. Br J Anaesth 1973; 45: 282–93; 1974; 46: 310(erratum)[ISI][Medline]

13 Büch U, Isenberg JC, Baldauf J, Altmayer P, Büch HP. Interactions between thiopental and volatile anesthetics (halothane and isoflurane) in isolated heart preparations of the rat. Methods Find Exp Clin Pharmacol 1992; 14: 499–505

14 Dale O, Jenssen U. Interaction of isoflurane with the binding of drugs to proteins in serum and liver cell cytosol: an in vitro study. Br J Anaesth 1986; 58: 1022–6[Abstract]

15 Dale O, Nilsen OG. Displacement of some basic drugs from human serum proteins by enflurane, halothane and their major metabolites: an in vitro study. Br J Anaesth 1984; 56: 535–42[Abstract]

16 Altmayer P, Büch U, Büch HP. Propofol binding to human blood proteins. Arzneim-Forsch 1995; 45: 1053–6[Medline]

17 Ludbrook GL, Upton RN, Grant C, Martinez A. Prolonged dysequilibrium between blood and brain concentrations of propofol during infusions in sheep. Acta Anaesthesiol Scand 1999; 43: 206–11[ISI][Medline]

18 Shyr M-H, Tsai T-H, Tan PPC, Chen C-F, Chan SHH. Concentration and regional distribution of propofol in brain and spinal cord during propofol anesthesia in the rat. Neurosci Lett 1995; 184: 212–5[ISI][Medline]

19 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[Abstract/Free Full Text]

20 Simons PJ, Cockshott ID, Douglas EJ, Gordon EA, Hopkins K, Rowland M. Disposition in male volunteers of a subanaesthetic intravenous dose of an oil in water emulsion of 14C-propofol. Xenobiotica 1988; 18: 429–40[ISI][Medline]

21 Lange H, Stephan H, Rieke H, Kellermann M, Sonntag H, Bircher J. Hepatic and extrahepatic disposition of propofol in patients undergoing coronary bypass surgery. Br J Anaesth 1990; 64: 563–70[Abstract]

22 Dogra S, Isaac PA, Cockshott ID, Foy JM. Pulmonary extraction of propofol in post-cardiopulmonary by-pass patients. J Drug Dev 1989; 2 (Suppl. 2): 133

23 Knibbe CAJ, Koster VS, Aarts LPHJ, Langemeijer HJM, Danhof M. Pilot study in humans on the pharmacokinetics and safety of propofol 6% SAZN. Pharm World Sci 1999; 21: 239–40[ISI][Medline]

24 Francis B, Green M, Payne C, eds. The GLIM System—Release 4 Manual. Oxford: Clarendon Press, 1993

25 Healy MJR. GLIM: An Introduction. Oxford: Clarendon Press, 1988

26 Altman DG. Practical Statistics for Medical Research. London: Chapman and Hall, 1991