Pharmacokinetics and effects of propofol 6% for short-term sedation in paediatric patients following cardiac surgery

Abstract

Aims

This paper describes the pharmacokinetics and effects of propofol in short-term sedated paediatric patients.

Methods

Six mechanically ventilated children aged 1–5 years received a 6 h continuous infusion of propofol 6% at the rate of 2 or 3 mg kg⁻¹ h⁻¹ for sedation following cardiac surgery. A total of seven arterial blood samples was collected at various time points during and after the infusion in each patient. Pharmacokinetic modelling was performed using NONMEM. Effects were assessed on the basis of the Ramsay sedation score as well as a subjective sedation scale.

Results

The data were best described by a two-compartment pharmacokinetic model. In the model, body weight was a significant covariate for clearance. Pharmacokinetic parameters in the weight-proportional model were clearance (CL) = 35 ml kg⁻¹ min⁻¹, volume of central compartment (V₁) = 12 l, intercompartmental clearance (Q) = 0.35 l min⁻¹ and volume of peripheral compartment (V₂) = 24 l. The interindividual variabilities for these parameters were 8%, < 1%, 11% and 35%, respectively. Compared with the population pharmacokinetics in adults following cardiac surgery and when normalized for body weight, statistically significant differences were observed for the parameters CL and V₁ (35 vs 29 ml kg⁻¹ min⁻¹ and 0.78 vs 0.26 l kg⁻¹P < 0.05), whereas the values for Q and V₂ were similar (23 vs 18 ml kg⁻¹ min⁻¹ and 1.6 vs 1.8 l kg⁻¹, P > 0.05). In children, the percentage of adequately sedated patients was similar compared with adults (50% vs 67%) despite considerably higher propofol concentrations (1.3 ± 0.10 vs 0.51 ± 0.035 mg l⁻¹, mean ± s.e. mean), suggesting a lower pharmacodynamic sensitivity to propofol in children.

Conclusions

In children aged 1–5 years, a pharmacokinetic model for propofol was described using sparse data. In contrast to adults, body weight was a significant covariate for clearance in children. The model may serve as a useful basis to study the role of covariates in the pharmacokinetics and pharmacodynamics of propofol in paediatric patients of different ages.

Introduction

Propofol is widely used for sedation in mechanically ventilated adult patients in the Intensive Care Unit (ICU). Favourable properties for this indication are the rapid onset of action, the easy titratability and the rapid reversal of sedation without signs of accumulation, and its safety in patients with renal and hepatic disease. Propofol is often used for the sedation of children in the paediatric intensive care unit (PICU), although this is subject to debate and limited data are available [1, 2]. Since 1992, several fatal cases have been described following long-term sedation (> 3 days) with high doses of propofol (> 6 mg kg⁻¹ h⁻¹) in children in the PICU, associated with myocardial failure, metabolic acidosis, lipaemic serum and multiorgan failure [3, 4]. The US Food and Drug Administration found no direct link between propofol and death in PICUs [5], but these cases led in Europe to the recommendation that it be abandoned for paediatric sedation. Nevertheless, propofol is used in many PICUs for the sedation of children in North America as well as in Europe, albeit with strict controls on dose and duration of infusion [2, 5–8].

Recently two studies of propofol for sedation in children have been reported [9, 10]. In a retrospective study in 106 children in a PICU, propofol at a mean dose of 3.39 mg kg⁻¹ h⁻¹ compared favourably with other sedatives [9]. The incidence of metabolic acidosis as well as mortality was similar in the propofol and the control group [9]. In another study, Martin et al. [10] found no evidence of metabolic acidosis or significant biochemical changes in nine children sedated by propofol infusion of 1–4 mg kg⁻¹ h⁻¹ for 48 h after cardiac surgery.

Very limited information is available on the pharmacokinetics and pharmacodynamcs of propofol in children [7]. These patients generally require higher doses of propofol than adults [12, 13], which may be due to differences in pharmacokinetics, pharmacodynamics or both. Therefore, the objectives of the present study were to investigate the pharmacokinetics as well as the effects of propofol for short-term sedation in children following cardiac surgery, and to compare the results with those in adults. Due to its lipophilic characteristics, propofol is formulated in an intravenous fat emulsion (typically propofol 1% and more recently also propofol 2%). As it has been suggested that the fat content of propofol formulations may have contributed to fatalities, propofol 6% [14–16] was used in the present investigation instead of propofol 1%, the former containing 83% less fat.

Methods

The study protocol was approved by the local ethics committee of the Department of Paediatrics of the University Medical Centre of Utrecht, the Netherlands. The parents of the patients gave their consent in writing after being informed of the aims, nature and procedures of the study.

Patients

Six children scheduled to undergo cardiac surgery and who were expected to require postoperative mechanical ventilation for a minimum of 6 h participated in the study. Two children had (atrio)ventricular septal defect repair, two had repair of tetralogy of Fallot and two had repair of aortic stenosis or Shone's complex. The patients (four male and two female) were aged between 1 and 5 years, weighed between 10 and 21 kg, had normal renal (serum creatinine concentration < 2 times the upper limit of normal for their age) and hepatic function (liver function tests < 3 times the upper limit of normal for their age), were not suffering from severe or uncontrolled noncardiac disease, respiratory infections or impairment of lipid metabolism (serum triglycerides < 2 mmol l⁻¹ and/or serum cholesterol < 2 times the upper limit of normal for their age).

Study design and procedure

In all children, the anaesthetic technique was standardized. Before surgery, diazepam was given orally as premedication. Anaesthesia was induced with sufentanil and midazolam, and patients were paralysed with pancuronium. Thereafter patients were intubated and mechanically ventilated. Anaesthesia was maintained with sevoflurane and midazolam, and sufentanil was given as needed. After the surgical procedure mechanical ventilation was continued and the patient was transferred to the PICU. A continuous infusion of propofol 6% (prepared in the Department of Clinical Pharmacy, St Antonius Hospital, Nieuwegein, the Netherlands) [17] was started for a duration of 6 h in a dose of 2 or 3 mg propofol kg⁻¹ body weight h⁻¹. The propofol emulsion was administered into a running saline infusion by an IVAC P3000 infusion pump (Alaris Medical Systems, Hampshire, United Kingdom). An infusion of morphine was also given (0.25 mg kg⁻¹ body weight 24 h⁻¹). During surgery and the postoperative sedation period in the PICU the electrocardiogram, heart rate, central venous pressure, arterial blood pressure, rectal temperature, the amounts of all drugs, fluids and blood administered, blood loss and urine output were measured continuously or at intervals of 15 min. Serum triglyceride, cholesterol and liver function test measurements were taken before and after the propofol infusion. Any adverse events that occurred during surgery or postoperative sedation in the PICU were recorded.

Blood sampling

Samples of arterial blood (250 µl) were collected from an indwelling radial arterial cannula prior to the start of the propofol infusion. For three patients, blood samples were collected at approximately 15, 60, 240 and 360 min after the start of the propofol infusion and at 30, 60 and 120 min after the termination of the infusion, the exact time being recorded for each sample. For the other three patients, samples were collected at approximately 30, 120, 300 and 360 min after the start, and at 15, 45 and 120 min after termination of the infusion. Samples were collected in oxalate tubes, mixed thoroughly and were stored at 4 °C until analysis.

Drug assay

Propofol concentrations in whole blood samples were measured using high-performance liquid chromatography with fluorescence detection [18], with the following modifications. A calibration curve ranging from 0.05 to 1.0 mg l⁻¹ was prepared. To 100 µl of undiluted blood 20 µl of methanol were added and the mixture vortexed. Then 250 µl of internal standard solution were added, the mixture vortexed and, after centrifugation, 15 µl of the supernatant were injected onto the chromatograph. The limit of quantification was 0.03 mg l⁻¹. The coefficients of variation for the intra- and interassay precision were less than 5.0% and 5.8%, respectively, over the concentration range studied.

Pharmacokinetics

A two-compartment model was used with the pharmacokinetic parameters elimination clearance (CL), intercompartmental clearance (Q), volume of the central compartment (V₁) and volume of peripheral compartment (V₂). The pharmacokinetic parameters of the mixed-effects population model, in which both intraindividual and interindividual variability were considered, were estimated using NONMEM (version V) [19].

NONMEM minimizes an objective function in performing nonlinear regression analysis. The weighted residuals (WR) were used as the primary measure of goodness of fit. The median population values of WR (MDWR) was used as an estimate of model bias, and the median population value of the absolute weighted residual (MDAWR) was used as an estimate of model accuracy [20].

For estimation of interindividual variability of all pharmacokinetic parameters, log-normal distributions of the volumes and clearances in the population were assumed. Therefore for the ith individual:

(1)

where θ_i is the individual value of the parameter θ, θ_TV is the mean population value of this parameter (‘typical value’), and η_i is a random variable normally distributed about zero.

The intraindividual variability that describes the residual errors, was calculated using two different residual error models, after investigating different error models which were tested according to the visual assessment of the weighted residuals and the NONMEM objective function: an additive model was used during the propofol infusion and a constant coefficient of variation model for the post infusion data. This means for the jth measured concentration of the ith individual (c_ij) during the propofol infusion:

(2)

which assumes that the residual error is of the same magnitude over the range of measurements, and where c_pred,ij is the corresponding predicted concentration.

After the infusion:

(3)

which assumes that the residual error is proportional to the propofol concentration.

In equation 2 and 3, ɛ_1,ij and ɛ_2,ij are random variables normally distributed around zero.

Covariate analysis was also performed using NONMEM. Individual estimates of the pharmacokinetic parameters of each individual were obtained using a two-compartment model without covariates (simple model). The estimated pharmacokinetic parameters were plotted independently against the covariates age, weight or body mass index to identify their influence. Using the simple model without covariates, each covariate was separately incorporated and both proportional functions and proportional functions with added constants were explored. The effects of covariates were tested for statistical significance using the NONMEM objective function (P < 0.01, χ² distribution). In addition, the plots of the residuals (measured-over-predicted concentration values), and the residual intraindividual variability which corresponds to unexplained variability, and the goodness-of-fit measures WR, MDWR and MDAWR were evaluated.

Level of sedation

The level of sedation was assessed by the PICU nursing staff using the Ramsay six-point scale [21] which considers the following levels: (1) patient anxious, agitated, restless, (2) patient cooperative, orientated and tranquil, (3) patient drowsy or asleep, responds easily to commands, (4) patient asleep, brisk response to a light glabellar tap, (5) patient asleep, sluggish response to a light glabellar tap and (6) patient asleep, no response to a light glabellar tap.

In addition, a subjective scale was used: (1) too deep, (2) adequate (patient tranquil/asleep), (3) moderate (patient restless/asleep) and (4) inadequate (patient agitated/awake).

At a sedation level of 1 on the Ramsay scale and/or subjective score 3 or 4, midazolam could be administered. Overall, the level of sedation was considered adequate when, with no additional sedative medication given, the Ramsay score was between 2 and 6 and the subjective score was 2 during all 6 h of the propofol infusion. In all other cases, the patients were considered to be inadequately sedated [15]. The corresponding mean propofol blood concentration was defined as the average of the propofol concentration at 240 and 360 min after the start of the infusion.

Comparison of pharmacokinetics and pharmacodynamics with adults

The data from a similarly designed short-term sedation study with propofol in adults following coronary artery bypass surgery [15] were reanalysed on the basis of the two-compartment population model as described in the section Pharmacokinetics. In short, 24 adult patients received a continuous 5 h infusion of propofol (formulated as propofol 6% or propofol 1%) at the rate of 1 mg kg⁻¹ h⁻¹ for sedation following cardiac surgery (standardized anaesthesia with fentanyl and midazolam). A total of 24 arterial blood samples was collected from each patient during and after the infusion (ranging from 5 min after the start of the infusion until 150 min after the termination of the infusion). During the propofol infusion, the depth of sedation was assessed using the Ramsay sedation score as well as the subjective sedation scale [15], and the same criteria as described under section Level of sedation were used.

Statistical analysis

The difference between pharmacokinetic parameters obtained in children vs adults were tested by analysing the posthoc individual pharmacokinetic parameter estimates [19] using the unpaired t-test (with Welch correction where needed).

Results

The pharmacokinetics of propofol in sedated children were adequately described by a two-compartment model. Table 1 shows the pharmacokinetic parameter estimates and performance measures when no covariates were incorporated into the two-compartment model (‘simple model’). Covariate analysis revealed that clearance correlated with the body weight of the children (r² = 0,90). When body weight was incorporated into the model as a proportional function for clearance (weight-proportional model), the objective function decreased from −69 to −102 (P < 0.001), and the measures for bias (MDWR) and precision (MDAWR) and the variabilities were reduced compared with the simple model in children (Table 1). For the body weight-proportional model, mean popluation values (± s.e. mean) for clearance (CL), volume of central compartment (V₁), intercompartmental clearance (Q) and volume of peripheral compartment (V₂) were 35 ± 1.9 l kg⁻¹ min⁻¹, 12 ± 2.9 l, 0.35 ± 0.06 l min⁻¹ and 24 ± 4.4 l, respectively. Additional exponents or constants to the proportional function did not statistically improve the performance of the model. Figure 1 shows the concentration–time profiles of the best, median and worst performance of the simple model and the corresponding fits of the weight-proportional model, with lines predicted by the population (mean) model and by the individualized population model. No influence of covariates on other pharmacokinetic parameters was found.

Table 1.

Pharmacokinetic parameters for propofol and performance measures of the simple model with no covariates and the body weight-proportional model for children and adults undergoing open heart surgery.

	Simple model		Children BW-model		Simple model		Adults*BW-model
Parameter	Value	% CV	Value	% CV	Value	% CV	Value	%CV
Estimated parameters
CL (l min−1)	0.49	39%			2.3	20%
(ml kg−1 min−1)			35	8%	(29)		28	17%
V1 (l)	7.1	86%	12	< 1%	21	50%	20	15%
(l kg−1)			(0.78)		(0.26)
Q (l min−1)	0.6	< 1%	0.35	11%	1.4	29%	1.5	31%
(ml kg−1 min−1)			(23)		(18)
V2 (l)	29	< 1%	24	35%	139	22%	141	25%
(l kg−1)			(1.6)		(1.8)
Intra-individual variability
Variance (ɛ1) (mg l−1)	0.07		0.07		0.01		0.01
Variance (ɛ2) (%CV)	20%		15%		12%		11%
Performance measures
−2LL	−69		−102		−2721		−2718
MDWR	−21%		−3.5%		0.4%		−3.8%
MDAWR	60%		52%		51%		54%

Inter-individual variability is expressed as %CV and equals the square root of the exponential variance of ηminus 1. Intra individual variability during the infusion is expressed in mg l ⁻¹ (variance (ɛ₁)) and the intraindividual variability after the infusion is expressed as %CV and equals the square root of the variance of ɛ₂. CL = clearance, V₁ = volume of central compartment, Q = intercompartmental clearance, V₂ = volume of peripheral compartment, −2LL = objective function, MDWR=median weighted residual, MDAWR = median absolute weighted residual, BW = body weight (kg). Values in parentheses are derived by dividing the population mean value by mean body weight.

^*Data from Knibbe et al. [15], reanalysed on the basis of a population pharmacokinetic model.

Figure 1.

Propofol concentration over time for the best, median and worst performances of the simple two-compartment model without covariates (left column) and the weight-proportional model (right column). The solid circles represent the measured propofol concentrations, the thin grey lines the concentrations predicted by the population (mean) model and the dark grey lines the concentrations predicted by the individualized population model.

In adults who were sedated following coronary artery surgery during 5 h, the simple two-compartment population model adequately described the pharmacokinetics of propofol. Mean population values (± s.e. mean) for clearance (CL), volume of central compartment (V₁), intercompartmental clearance (Q) and volume of peripheral compartment (V₂) were 2.3 ± 0.09 l min⁻¹, 21 ± 2.1 l, 1.4 ± 0.11 l min⁻¹ and 139 ± 8.1 l, respectively. The incorporation of physiologic characteristics into the model did not improve the performance measures (Table 1).

Figure 2 presents plots of observed vs predicted propofol concentrations for the weight-proportional model in children and the simple model in adults. The figure shows that the final models were able to describe the pharmacokinetics for both data sets with similar bias and precision, despite differences in number of patients (6 children vs 24 adults) and in number of observations per patient (sparse data set in children and rich data set in adults).

Figure 2.

Plot of the propofol concentrations predicted by the individual estimates of the final population model vs the observed propofol concentrations in children (weight-proportional model) and in adults (simple model) with lines of identity.

Comparison of pharmacokinetic parameters between children and adults when normalized for body weight revealed that clearance and volume of central compartment were larger in children than in adults (35 vs 29 ml kg⁻¹ min⁻¹ and 0.78 vs 0.26 l kg⁻¹, respectively, P < 0.05), whereas values for Q and V₂ were similar (23 vs 18 ml kg⁻¹ min⁻¹ and 1.6 vs 1.8 l kg⁻¹, respectively, P > 0.05). This is further demonstrated in Figure 3, in which the distribution of the pharmacokinetic parameter estimates of all patients is presented. For each parameter, the difference between the children and the adults was 6.7 (2.7, 10.8) ml kg⁻¹ min⁻¹ for CL, 7.3 (−2.7, 17.4) ml kg⁻¹ min⁻¹ for Q, 0.60 (0.26, 0.93) l kg⁻¹ for V₁, and −0.19 (−0.79, 0.41) l kg⁻¹ for V₂ (value (95% confidence interval)).

Figure 3.

Individual estimates (•) and population mean of clearance, intercompartmental clearance, volume of central compartment and volume of peripheral compartment of the final population models vs body weight of the children (n = 6) and adults (n = 24).

Compared with the level of sedation in adults, the percentage of adequately sedated patients was similar in children (50% vs 67%), despite considerably higher propofol concentrations (1.3 ± 0.10 vs 0.51 ± 0.035 mg l⁻¹, mean ± s.e. mean) reflecting the higher propofol doses (3 mg kg⁻¹ h⁻¹vs 1 mg kg⁻¹ h⁻¹, children vs adults, respectively) (Table 2). Two children were adequately sedated according to the criteria defined previously (Table 2), whereas two other children required a midazolam infusion of 0.1 or 0.2 mg kg⁻¹ h⁻¹ in addition to the propofol infusion of 3 mg kg⁻¹ h⁻¹. The first patient received 2 mg kg⁻¹ h⁻¹ propofol and required two additional bolus doses of midazolam of 0.1 mg kg⁻¹ and another child received by mistake a midazolam infusion of 0.2 mg kg⁻¹ h⁻¹ in addition to the propofol infusion of 3 mg kg⁻¹ h⁻¹. These two patients were not included in Table 2.

Table 2.

Level of sedation and mean propofol concentration in children and adults undergoing open heart surgery (mean ± s.e. mean).

	Children	Adults*
Number of patients	4	24
Propofol dose (mg kg−1 h−1)	3	1
Adequately sedated	50%	67%
Inadequately sedated	50%	33%
Mean propofol concentration (mg l−1)	1.3 ± 0.10	0.51 ± 0.035

^*Data from Knibbe et al. [15], reanalysed on the basis of a population pharmacokinetic model.

During and after the propofol infusion, the serum triglyceride and cholesterol concentrations, liver function tests, serum creatinine and body temperature were all within normal ranges. There were no adverse events in any of the children.

Discussion

Fatal cases following the use of propofol in children in PICU [3, 4] indicate that it may be inappropriate to use the drug in this patient group in the same way as in adults [9]. The adverse events in children occurred during the uncontrolled use of propofol following its successful use in adult critically ill patients. Because of its attractive features such as rapid onset, short duration of action, and ease of titratability, the use of propofol in sedated children has been re-evaluated [6–11]. However, a pharmacokinetic and pharmacodynamic evaluation of propofol given to children in the PICU is needed to explain observed differences in dose requirement between children and adults, and to identify covariates that may lead to more rational dosing schemes for propofol in this particular patient group.

In this paper a pharmacokinetic model for the sedation of children aged 1–5 years following cardiac surgery is described based on sparse data and using a population approach. In our view, the latter is preferred over the standard two-stage approach in paediatric settings, because it allows one to model the entire population studied together, so that the maximum amount of information using the minimum number of samples is obtained. In the present study we show that, using a population approach and with the limited number of seven samples per child, a model with acceptable intra- and interindividual variabilities and adequate performance measures can be developed, and in which the influence of covariates can be identified.

In the pharmacokinetic model, body weight was a significant covariate for elimination clearance. A previous pharmacokinetic study in 19 sedated children in the PICU by Reed et al. [7] used the standard two-stage approach and found similar pharmacokinetic parameters estimates although with larger unexplained interpatient variability. Extensive sampling led to the characterization of the data on the basis of a three compartment model. However, despite the wide range in age (0.13–182 months) and in body weight (2.6–60 kg) in their study population, they could not explain this large interpatient variation. We think that this may be caused by the use of the standard two-stage approach instead of the population approach.

The estimates for CL and V₁ we found in sedated children in PICU are similar to values described in otherwise healthy children after general anaesthesia, although in those studies a three-compartment model was applied after prolonged sampling [20, 22–24].

The pharmacokinetic parameters CL and V₁, when normalized for body weight, were larger in children than in adults. The exact nature of the relationship between body weight and clearance needs to be clarified. It has been suggested that values of pharmacokinetic parameters in humans should be standardized for body weight on the basis of a power function, because of the evidence that body weight is a good predictor of body function over a wide range of body weights and species [25, 26]. The widely used alternatives, linear and per kilogram models are, according to these authors, inferior in comparison with the power law [25, 26]. In the present investigation we were unable to identify such a power function, which may be explained by the limited number of subjects in this study and the relatively narrow body weight range. The objective of future investigations should be to identify the exact relationship between body weight and the values of the pharmacokinetic parameters by studying children of different ranges of body weight.

Although the study is relatively small to examine the efficacy of sedation, there was a large difference in the concentration required for a given level of sedation between children and adults. Whereas in children propofol concentrations were 2.5-fold higher than in adults, the percentage of adequately sedated children was similar, which suggests a difference in pharmacodynamic sensitivity to propofol between children and adults. In our view, this relative insensitivity to propofol in children may have led to the high doses which were needed to sedate children, who subsequently died [3, 4]. The very large difference in the depth of sedation in relation to the propofol concentration between children and adults indicates that further data on the pharmacodynamics in children of varying age groups are needed. To this end, the present study design and model may provide a useful starting point.

In conclusion, a pharmacokinetic model of propofol used to sedate children aged 1-5 years was described using sparse data. In contrast to adults, body weight was a significant covariate for clearance in children. The model may serve as a useful basis for further investigations in different age groups to study the role of covariates in the pharmacokinetics and pharmacodynamics of propofol in paediatric patients.