British Journal of Anaesthesia 106 (4): 558–65 (2011)

Advance Access publication 9 February 2011 . doi:10.1093/bja/aer001

PAEDIATRICS

Population pharmacokinetics of nalbuphine after

surgery in children
F. Bressolle 1*, S. Khier 1, A. Rochette 2, J. M. Kinowski 3, C. Dadure 2 and X. Capdevila 2
1
Pharmacokinetic Laboratory, Faculty of Pharmacy, BP 14491, Montpellier I University, 15 Avenue Ch. Flahault, 34093 Montpellier Cedex 5,
France
2
Department of Anesthesia and Critical Care ‘A’, Lapeyronie Hospital, 371 av du doyen G. Giraud, CHU de Montpellier, 34295 Montpellier
Cedex 5, France
3
Pharmacy Service, Carémeau Hospital, Rue du Prof. Janbreau, Nimes, France
* Corresponding author. E-mail: fbressolle@aol.com

Background. Nalbuphine is an opioid analgesic agent widely used for control of mild-to-severe
Editor’s key points pain. However, limited data are available on the pharmacokinetics of this drug in children. The
† This study reveals aim of this study was to characterize the population pharmacokinetics of nalbuphine in
paediatric population patients with ages ranging from 1 to 11 yr and to identify patient characteristics partially
data for nalbuphine, an explaining inter-individual variability in nalbuphine pharmacokinetic parameters.
opioid analgesic agent Methods. Twenty-two children were included in this study. They received nalbuphine after
widely used for control of surgery by continuous infusion (loading dose, 0.2 mg kg21 over 10 min followed by
mild-to-severe pain. continuous infusion of 0.8 mg kg21 over 24 h). If pain relief was not adequate, 0.1 mg kg21
† A two-compartment bolus doses were allowed in 10 min. Eleven blood samples were collected per patient. The
allometric power model data were analysed by non-linear mixed-effect modelling with the use of a two-
developed in this study compartment structural model.
best described the data. Results. Twenty patients completed the study. In the final model, the parameter values were
† Allometric models in standardized for a body weight of 70 kg using an allometric model. Population parameter
children well described estimates were: clearance 130 litre h21 70 kg21, inter-compartment clearance 75.6 litre h21
the relationships 70 kg21, central volume of distribution 210 litre 70 kg21, and peripheral volume of
between clearances, distribution 151 litre 70 kg21. In the children of this study, total clearance expressed in litre
volumes of distribution, h21 kg21 decreased significantly with increasing age and the elimination half-life
and weight and might be significantly increased.
useful for dose Conclusions. The allometric power model developed in this study best reflected the data and
adjustments. may be useful for dose adjustment.
Keywords: population pharmacokinetics; surgery, laparoscopy; children
Accepted for publication: 17 December 2010

Nalbuphine hydrochloride [(2)-17(cyclobutylmethyl)-4,5a- The estimated hepatic extraction ratio of nalbuphine is

epoxymorphinan-3,6a,14-triol hydrochloride] is a synthetic 0.5 – 0.7; thus, its hepatic clearance will be mainly depen-
narcotic agonist –antagonist analgesic of the phenanthrene dent on hepatic blood flow.4 5 Pharmacokinetic data of
series with a duration of analgesia of 4–5 h.1 It is structurally this drug are limited; studies have been carried out in
related to narcotic antagonist, naloxone, and to the potent adults, children, and neonates.4 6 – 8 It is well known
narcotic analgesic, oxymorphone. Nalbuphine is used to that the pharmacokinetics of most drugs are age-
treat and prevent moderate to severe pain; it can also be dependent.9 – 11 Maturation of metabolic pathways takes
used for pain relief before and after surgery and during place at a different rate; the metabolic clearance of drugs
childbirth. In adults with acute pain, its analgesic potency is usually very low at birth and then increases to reach a
is equivalent to that of morphine on a milligram basis.2 In maximum at about the age of 1 yr when it can exceed
regard to morphine, nalbuphine could induce less respiratory that of adults. Simultaneously, water compartments are
depression at high doses and less effect on arterial pressure.3 significantly larger in children than in adults. Thus, the
This drug undergoes an important hepatic metabolism in treatment of postoperative pain by drugs extensively
humans giving N-hydroxycetocyclobutylmethyl nornalbu- metabolized in the liver raises a challenge to the clinician
phine, the major metabolite, and hydroxylated derivatives.1 who takes care of them.

& The Author [2011]. Published by Oxford University Press on behalf of the British Journal of Anaesthesia. All rights reserved.
For Permissions, please email: journals.permissions@oup.com
Nalbuphine pharmacokinetics in children BJA
Although nalbuphine has been approved for clinical use in intubation, anaesthesia was maintained with sevoflurane
children, pharmacokinetic data of the drug remain very (1 MAC end-tidal concentration in 50% oxygen and air) and
limited. Nicolle and colleagues7 reported data for a few neo- remifentanil (0.4 mg kg21 min21). Neuromuscular block was
nates whose mothers had received nalbuphine during labour. obtained with atracurium (0.5 mg kg21 to maintain a
Jacqz-Aigrain and colleagues8 have reported data of a popu- train-of-four count of 1) at the beginning of surgery. A
lation pharmacokinetic study carried out in neonates. These second i.v. line was inserted into a femoral vein to facilitate
authors found that birth weight is a major determinant of subsequent blood sampling. Central temperature was care-
variability in nalbuphine disposition. Jaillon and colleagues4 fully maintained above 35.58C using a forced air system. Ven-
have shown that the elimination half-life of nalbuphine is tilation was adjusted to maintain end-tidal CO2 between 30
shorter in infants of 1.5–5 yr of age than in those of 5–8.5 and 35 mm Hg at the beginning of surgery. No further
yr and that systemic clearance per kilogram of body weight change in ventilation was allowed during the study period.
decreased with age. However, only seven infants have been At skin closure, remifentanil and sevoflurane were stopped,
included in each group. The present study was designed to end of neuromuscular block was assessed, and children
provide data on pharmacokinetics of nalbuphine in a paedia- were allowed to breathe 100% O2 spontaneously. After
tric population aged 1–11 yr old. The purpose of this study surgery, children received niflumic acid, a selective inhibitor
was (i) to determine accurate population pharmacokinetic of cyclooxygenase 2, by rectal route (20 mg kg21 twice a
parameters by using a two-compartment open model; this day), acetaminophen i.v. (30 mg kg21, four times a day),
model was parameterized in terms of total clearance, and nalbuphine i.v. according to the following protocol: the
central volume of distribution, inter-compartment clearance, loading dose, 0.2 mg kg21, over 10 min was given at
and peripheral volume of distribution, (ii) to accurately esti- wound closure followed by continuous infusion of 0.8 mg
mate both inter- and residual variability in pharmacokinetic kg21 over 24 h. In addition, 0.1 mg kg21 bolus doses were
parameters, and (iii) to identify which of the patient physio- allowed in 10 min if pain relief was not adequate, but no
logical parameters could have influenced drug disposition. more than twice within 1 h and no more than five times
These data provide further valuable information regarding for the 24 h study period. If pain remained unacceptable,
appropriate paediatric dosing. nalbuphine would be stopped and rescue analgesia would
be provided using i.v. morphine. Pain, evaluated every hour
Methods and 30 min after each additional dose of nalbuphine, was
considered unacceptable when the Children and Infants
Study design Postoperative Pain Score (CHIPPS)12 exceeded 2 on a
The study protocol was reviewed and approved by the insti- maximum of 10.
tutional review board. It was performed in accordance with
the legal requirements and the Declaration of Helsinki, and Blood sampling
with current European Community and US Food and Drug
It was planned to collect heparinized blood samples (2.5 ml)
Administration guidelines for good clinical practice. Written
from each patient at the following times: (i) immediately
informed consent was obtained from the parents or legal
before and (ii) at the end of the loading dose, (iii) 12 and
guardians.
24 h after the beginning of i.v. infusion, and (iv) 0.25, 0.5,
Twenty-two infants undergoing laparoscopic fundoplica-
1, 2, 4, 6, and 12 h after the end of infusion. Immediately
tion for gastro-oesophageal reflux aged 1–11 yr were
after collection, plasma was separated by centrifugation
included in this study. They were admitted in the department
(1500g) within 10 min and was frozen at 2308C until
of paediatric surgery of the Lapeyronie Hospital (Montpellier,
assayed.
France). For all children, pre-anaesthetic data, anamnestic
data, physical examination, and standard laboratory tests
which included haematological and biochemical tests were Nalbuphine assay
performed before and at the end of the study. Each subject’s Nalbuphine was quantified in human plasma by high-
measurements of alanine aminotransferase, aspartate ami- performance liquid chromatography using tandem-mass
notransferase, bilirubin (direct bilirubinaemia ,2 mg dl21), spectrometry detection (LC-MS/MS). The internal standard
creatinine (creatininaemia ,1 mg dl21), red blood cell used was morphine. The LC system equipped with an auto-
count, white blood cell count, platelet count, haematocrit, sampler set at 48C was coupled to a PE Sciex API 365 quadru-
and haemoglobin were within normal ranges. Also children pole MS (Applied Biosystem MDS Sciex, Courtaboeuf, France)
receiving analgesics or anti-inflammatory drugs the week with a turbo electrospray ion source that was operated in a
before surgery were excluded. Anaesthesia was standardized positive ionization mode with the nebulizer and TurboIon-
for all patients. Patients were fasting for 6 h. One hour before Spray gases (nitrogen) set at 12 and 30 pounds per square
surgery, they received midazolam (0.4 mg kg21) as rectal inch (psi), respectively. The voltage and temperature were
premedication. Upon arrival in the operating theatre, stan- maintained at 4500 V and 2508C, respectively. Nitrogen gas
dard monitoring equipment was applied, an i.v. line was was used in a collision-induced dissociation at a back pressure
established, and anaesthesia was started with propofol of approximately level 6. Nalbuphine was quantified via a mul-
(4 mg kg21) and remifentanil (1 mg kg21). After tracheal tiple reaction monitoring mode of the transitions at m/z

559
BJA Bressolle et al.

358.2340.2 for nalbuphine and m/z 286.2286.2 for mor- where Pj is the pharmacokinetic parameter of the subject j, tij
phine, with a dwell time of 500 ms per transition. Optimized the time of the ith measurement, Dj the dosing history of the
collision energy of 10 eV was used for nalbuphine and 14 eV subject j, f the pharmacokinetic model, and 1ij the residual
for morphine. The chromatographic separation was carried deviation of the model from the observations and contains
out on a 30×5 mm (5 mm) reversed-phase Lichrospher contributions from intra-individual variability, assay error,
ODS2 LC column operating at room temperature (218C). and model misspecification for the dependent variable. 1 is
An isocratic mobile phase consisting of ammonium formate assumed to be a random Gaussian variable with mean zero
(1 mM) –acetonitrile (80:20, v/v) containing 1% formic acid and variance of v21 . Inter-individual variability in pharmacoki-
was used at a flow rate of 0.3 ml min21. A 10 ml sample was netic parameters was assessed according to an exponential
injected onto the column. Samples were extracted using the error model; the Pj parameter of the jth subject was
solid-phase extraction (SPE) automate Aspec XL4 on Bond described by the relationship:
Elut C18 (100 mg) cartridges. SPE cartridges were first con-
ditioned with 1 ml of methanol, 1 ml of distilled water, and Pj = Pmean × exp(hP ) (2)
1 ml of Tris-buffered saline solution (pH 7.5) and then
plasma samples were loaded onto the cartridges. The where Pj is the pharmacokinetic parameter of subject j, Pmean
column was then rinsed with 1 ml of distilled water. The the population pharmacokinetic parameter, and hP a
elution was carried out with 1 ml of a mixture of aceto- Gaussian random variable with mean zero and variance of v2hP .
nitrile –water (80:20, v/v) containing 1% formic acid. The Individual parameters were calculated as empirical Bayes
organic phase was evaporated under a stream of nitrogen estimates using the POSTHOC option in NONMEM. Several
at 408C. The peak area ratios (nalbuphine/internal standard) secondary pharmacokinetic parameters were calculated
varied linearly with concentration over the range of 1–100 from the individual primary pharmacokinetic parameters:
mg litre21. All calibration curves were weighted according to the volume of distribution at steady-state (Vss) and the half-
the 1/x 2 weighting scheme. The method was precise (pre- life (t1/2) of the terminal part of the curves.
cision, ≤12%) and accurate (recovery, 91–100%). Mean After selection of the best basic pharmacostatistical
extraction efficiencies .80% for each analyte were obtained. model, both a traditional approach and an allometric
No significant matrix effects occurred. Dilution has no influ- scaling were used to test the influence of covariates. In a
ence on the performance of the method. The lower limit of first step, the relationships between the individual pharma-
quantitation (LLOQ) was 1 mg litre21. cokinetic parameters and the above-mentioned covariates
were investigated graphically. Covariates showing a strong
Pharmacokinetic analysis correlation with a pharmacokinetic parameter were then
Pharmacokinetic model-building analyses were performed separately incorporated in the population model and tested
using the non-linear mixed-effects modelling (NONMEM) for statistical significance. Both linear functions and power
software (version 5.1.1, Globomax LLC, Hanover, MD, USA)13 models were tested (covariates being centred or not
through the Visual-NM graphical interface.14 The following around the mean values). The effect of each covariate was
covariates were considered pertinent to this study: patients’ assessed by the likelihood ratio test, based on the difference
age, body surface area, weight, height, gender, and the in the objective function values between hierarchical models.
ASA score (physical status classification system before The forward inclusion and backward elimination method was
surgery). A two-compartment model fitted data better than applied for covariate model development.
a one- or a three-compartment pharmacokinetic model. In a second step, allometric scaling15 – 17 was also tested
This model was parameterized in terms of total clearance to assess the influence of weight on CL, Q, V1, and V2:
(CL), central volume of distribution (V1), inter-compartment  g
WTj
clearance (Q), and peripheral volume of distribution (V2). Paramj = Paramstd × (3)
WT70 kg
First-order conditional estimation was used to fit the
models because individual data sets were extensive.13 The
with g ¼0.75 for clearances and g ¼1 for volumes of distri-
structural model was chosen on the basis of changes in
bution. In this equation, Paramj is the parameter in the jth
22 log-likelihood and on graphical analyses of the goodness
individual with a weight of WTj and Paramstd the parameter
of fit. Because 22 log-likelihood is approximately x2 distribu-
in an individual with a weight of 70 kg (WT70 kg).
ted and the addition of one compartment increases the
degree of freedom by a factor of 4, a change of 9.49 in 22
log-likelihood was required at the 5% significance level to
Quality of fit
select the more complex model. Several error models were Criteria for model selection included visual inspection of
compared: additive, exponential, or combined (additive+ goodness-of-fit plots [i.e. measured concentrations (DV) vs
exponential). It was found that residual variability was best population (PRED) and individual predictions (IPRED);
described by an exponential error model given below: weighted residuals (WRES) vs PRED; and WRES vs time after
administration]. The performance of Bayesian estimation
Cij (t) = f (Pj , Dij , tij ) × exp(1ij ) (1) was assessed by examining the prediction error (PE); PE
was defined as [(DV– IPRED or PRED)/IPRED or

560
Nalbuphine pharmacokinetics in children BJA
PRED]×100%. Both MDPE (median of all PEs) and MDAPE
(median of all absolute PEs) were calculated. Table 1 Patient characteristics. CI, confidence interval
Using the final covariate model, the visual predictive Age (yr) Weight Height (cm) Body surface
checks (VPCs) were carried out by simulating 1000 virtual (kg) area (m2)
data sets to assess the performance of the model. This Mean 4.2 15.6 97.6 0.64
analysis was performed using the R program. The 5th, 50th
95% CI 3.0 –5.4 12.4 –18.7 90.3 –104.9 0.56 – 0.72
(population median response), and 95th percentile concen- Maximum 11 34.0 137 1.0
trations were plotted against time post-dose and the Minimum 1 7.2 74 0.39
patients’ data were superimposed.

Simulations
We selected an average concentration of 12 mg litre21 as the A decrease in the objective function of 60.6 is associated
therapeutic concentration for the design of an effective with the use of a two-compartment model compared with a
dosing regimen for nalbuphine. This concentration corre- one-compartment model. The use of a three-compartment
sponds to the mean steady-state plasma concentration model led to failure in model convergence. Before covariate
observed in this study. Simulations were performed using inclusion (Step 1), population pharmacokinetic parameters
parameters of the final model to determine the optimal are summarized in Table 2. During covariate analysis, two
treatment schedule of nalbuphine needed to maintain this models gave good results: (i) the addition of weight and age
on CL [CL¼weight×(2age× u1 + u2)] and age on V1
concentration. Typical patients of weight 10 kg (1–2 yr old),  u
13 kg (2–5 yr old), and 24 kg (5–11 yr old) were assumed [V1 = u3 × age/4.2 4 ] to the base model produced a decrease
for simulation purposes. Moreover, we have characterized in objective function of 44.6 units; and (ii) the use of allometric
the pharmacokinetics of elimination of nalbuphine in terms scaling results in an improvement of the objective function of
of the 20%, 50%, and 80% context-sensitive decrement 39.3 units. For both models, the fits to the data were excellent
times. and the differences were relatively small. As the first model
has two more parameters than the allometric model, it is
Validation of the final model not statistically better. Moreover, criteria on the quality of fit
were slightly better for the allometric model. Thus, this
The bootstrap resampling procedure was used for evaluating
model was selected as the final model.
the stability and robustness of the final model.18 The boot-
The estimated population parameters of nalbuphine in
strap resampling was repeated 1000 times to evaluate
the final model are shown in Table 2. It was not possible to
whether an appreciable discrepancy existed between the
estimate population parameter variability on V2. The ratios
parameter values estimated from the original data and the
of the between-subject variance predictable from covariates
estimated bootstrap mean values. Final population par-
to the total population parameter variance obtained without
ameters were compared with those obtained from the
covariate analysis are presented in Table 3. Inclusion of body
1000 bootstrap analyses.
weight decreased the variance of CL and V1, but increased
the variance of Q. A reason may be that there is no v for
Results V2 (having a large variance of 0.473 in the basic model) in
All patients received planned continuous analgesia and no the allometric model, so that the high variance of Q may in
therapeutic failure (i.e. shift to morphine) was observed. part be attributable to V2. Results presented in Table 3 indi-
There were 14 patients (63.6%) who required at least one cate that 39% and 65% of the overall variability in V1 and
additional bolus of nalbuphine. This first bolus was given at CL are predictable from covariate information, respectively.19
a mean time of 2.2 h after surgery (SD: 2.3 h). Nine patients Mean pharmacokinetic parameters of nalbuphine in the 20
(41.0%) received a second bolus at 7.4 h, seven received a children who completed this study are presented in Table 4.
third bolus at 10.0 h, three a fourth bolus at 12.6 h, and The elimination half-life of nalbuphine increased with age
one a fifth bolus at 19.8 h. All resulted in a decrease in (from 1.7 h in 1- to 2-yr-old children to 3.5 h in 7–11 yr
CHIPPS from 6.5 down to 0.7 within 30 min. Drowsiness old) and the CL in litre h21 kg21 decreased (from 3.3 to 2.2
was observed in seven patients (31.8%), nausea in three litre h21 kg21, in the two groups, respectively; Fig. 1).
(13.6%), and urinary retention in three. Out of 22 children The goodness of fit has been evaluated by comparing the
enrolled in the study, two were excluded for lack of blood regression line estimated on the DV vs IPRED values [slope:
samples. Baseline patient characteristics of the 20 children 1.04, 95% confidence interval (CI): 0.99–1.06; intercept:
(six girls and 14 boys) who completed the study are pre- 20.57 mg litre21, 95% CI: 21.4 to 0.023] to the reference
sented in Table 1. Eight children had an ASA score of I, 10 line of slope¼1 and intercept¼0, and no significant differ-
had an ASA score of II, and two had an ASA score of III. ence occurred (Supplementary Fig. S1A). Likewise, the slope
Depending on the patient, 4–6 h after the end of infusion, of the regression line DV vs PRED was not statistically differ-
nalbuphine concentrations were below the LLOQ of the ent from 1 and the intercept was not statistically different
analytical method. Therefore, a total of 157 plasma concen- from 0 (slope: 1.04, 95% CI: 0.97 –1.10; intercept: 20.20 mg
tration measurements were included in the analysis. litre21, 95% CI: 21.63 to 1.22) (Supplementary Fig. S1B).

561
BJA Bressolle et al.

Table 2 Population pharmacokinetic parameters of nalbuphine. IIV, inter-individual variability; SE, standard error of estimate expressed as
coefficient of variation; s, residual variability; CL, total clearance; V1, initial volume of distribution; V2, peripheral volume of distribution; Q,
inter-compartment clearance; PEs, prediction errors; MDPE, median of all PEs; MDAPE, median of all absolute PEs; DV, observed
concentrations; IPRED, individual predicted concentrations; PRED, population predicted concentrations

Basic model Final covariate model (allometric model)

Mean (SE, %) IIV, % (SE, %), shrinkage, % Mean (SE, %) IIV, % (SE, %), shrinkage, %
Population parameters
CL (litre h21) 41.6 (9.4) 31.2 (39.9), 13.6
CL (litre h21 70 kg21) 130 (5.1) 18.4 (36.7), 7.86
V1 (litre) 39.9 (8.5) 39.1 (38.3), 37.0
V1 (litre 70 kg21) 210 (8.7) 30.7 (53.1), 28.5
V2 (litre) 27.9 (19.7) 68.8 (57.5), 28.6
V2 (litre 70 kg21) 151 (14.0) Not estimated
Q (litre h21) 15.7 (36.6) 22.4 (28.7), 44.7
Q (litre h21 70 kg21) 75.6 (35.2) 34.8 (53.5), 34.4
s (%) 24.7 (18.0) 24.1 (21.4)
MDPE (%)
(DV vs IPRED) 21.78 0.17
(DV vs PRED) 3.70 20.40
MDAPE (%)
(DV vs IPRED) 10.7 11.3
(DV vs PRED) 28.3 22.8

Table 3 Effect of covariate analysis on variance (v2). PPV2, total

population parameter variance obtained without covariate 4
CL/weight (litre h–1 kg–1)

analysis; BSVP2, between subject variance predictable from

covariates; BSVR, random between subject variance estimated 3
when covariate analysis is included; CL, total clearance; V1, initial
volume of distribution; V2, peripheral volume of distribution; Q, 2
inter-compartment clearance
Slope: –0.15
1 Intercept: 3.40
PPV2 BSVP2 BSVR BSVP2/PPV2
P=7.2×10–5; r =−0.78
CL 0.0979 0.0641 0.0338 0.65 0
V1 0.153 0.059 0.0941 0.39 0 2 4 6 8 10 12
Age (yr)
V2 0.473 Not estimated
Q 0.050 20.071 0.121 21.42
Fig 1 Scatter plots showing the relationship between individual
weight-normalized CL values and age.

Table 4 Mean pharmacokinetic parameters of nalbuphine in the Moreover, adequate plots were observed in the final model
20 children of this study undergoing laparoscopic fundoplication. between weighted residuals and predicted concentrations
IIV, inter-individual variability; CL, total clearance; V1, initial
(Supplementary Fig. S1C). The vast majority of the weighted
volume of distribution; V2, peripheral volume of distribution; Q,
inter-compartment clearance; Vss, steady-state volume of residuals lay within 2 units of perfect agreement and were
distribution; t1/2, half-life of the terminal part of the curve symmetrically distributed around the zero ordinate; no sys-
tematic deviations were observed. For the final model, the
Mean IIV (%) MDPE and MDAPE of the population predictions PRED were
CL (litre h 21
) 41.0 35.1 20.40% and 22.8%, respectively (Table 2). These values
V1 (litre) 47.3 51.1 were in a typical range for pharmacokinetic models. The
V2 (litre) 34.0 48.3 plot of PEs vs time shows random distribution around the
Q (litre h21) 24.1 33.5 zero ordinate (data not shown). The VPC plot (Fig. 2) confirms
Vss (litre) 81.3 48.0 the adequacy of model predictions, showing no apparent
t1/2 (h) 2.7 18.3 deviations between model and data. About 95% of the
data fit well within the 5th –95th percentiles band and the

562
Nalbuphine pharmacokinetics in children BJA
data were symmetrically distributed around the median. The Table 5 presents (i) the simulation results for different
time– concentration profile (with 90% CI) for a 4-yr-old child ages and weights to obtain a target concentration of 12 mg
weighing 16 kg is presented in Figure 3. litre21 and (ii) context-sensitive decrement times for 20%,
50%, and 80% concentration decrement. These dosing sche-
dules result in 85% of simulated subjects having concen-
120 trations above 12 mg litre21 at steady state (between 12
and 15 mg litre21). Three children of 1, 1.5, and 7.2 yr have
100 steady-state nalbuphine concentrations of 9.7, 8.8, and 8.3
mg litre21, respectively.
Nalbuphine (µg litre–1)

80 The final model was fitted repeatedly to 1000

bootstrap-resampled data sets. Less than 6% of bootstrap
60 runs were unsuccessful. The average parameter values
obtained from the bootstrap analyses and the final estimates
40 from the original data set are compared in Table 6. These
results indicated that the reliability and robustness of the
20
parameter estimates and thus the population pharmacoki-
netic model was acceptable.
0
0 5 10 15 20 25 30
Time (h) Discussion
The present study was undertaken in the light of the limited
Fig 2 VPC plot for the studied population. The middle line depicts information regarding pharmacokinetics of nalbuphine in
the model predicted median. The other lines (above and below
children.4 7 8 With the exception of neonates,8 these pub-
the median) present the 5th and 95th percentiles. The grey
lished studies have been performed in a limited number of
area depicts the range between the 1th and 99th percentiles.
Black dots are observations. subjects.4 7 Both NONMEM and non-parametric expectation
maximization methods have been used for population phar-
macokinetic modelling in children. In this study, nalbuphine
population characteristics were estimated using NONMEM.
100 This population modelling method described nalbuphine
data well. Different pharmacokinetic models were tested;
Nalbuphine (µg litre–1)

the structural model was chosen on the basis of the

changes in 22 log-likelihood and qualitative assessment of
10 diagnostic plots. Contrary to the findings reported by
Typical child: 4 yr, 16 kg Jacqz-Aigrain and colleagues,8 a two-compartment model
Loading dose, 4 mg was found to fit the data satisfactorily. This model was in
Maintenance dose, 16 mg
Css, 13.7 µg litre–1 agreement with that published by Jaillon and colleagues.4
1 The number of patients finally included in this study was
0 5 10 15 20 25 30 35
Time (h) 20. In infants having about the same age, there are
marked inter-patient variations in weight (e.g. in patients
between 3 and 4 yr of age, the body weight ranged from 8
Fig 3 Time –concentration profile for a child aged of 5 yr and
weighing 20 kg. The dashed lines represent the 90% CIs. Css,
to 15 kg). These results were in agreement with those
steady-state plasma concentration reported by Jacqz-Aigrain and colleagues8 in neonates. It
has been suggested that the use of allometric models in

Table 5 Dosing regimens that attain a target concentration of 12 mg litre21 at steady-state and context-sensitive decrement times

Age (yr) Mean weight (kg) Loading dose over 10 min (mg) Maintenance dose over 24 h (mg) Context-sensitive
decrement times
(min)
% concentration
decrement
20 50 80
Range, 122; mean, 1.5 10 0.40 (0.040 mg kg21) 9.6 (0.960 mg kg21) 19 54 168
Range, .225; mean, 3.5 13 0.65 (0.050 mg kg21) 12.1 (0.931 mg kg21) 22 62 193
Range, .5211; mean, 7.5 24 1.10 (0.046 mg kg21) 16.0 (0.666 mg kg21) 25 80 246

563
BJA Bressolle et al.

Patients who completed the study received administered

Table 6 Bootstrap validation of the estimated population doses of nalbuphine ranging from 1 to 1.4 mg kg21 day21
pharmacokinetic parameters in the final model
(mean, 1.12 mg kg21 day21) including one to four additional
Parameters Original data 1000 bootstrap bolus doses required in 14 children to maintain adequate
replicates pain relief. The bolus dose at the initiation of treatment
Mean 2.5% Mean 2.5% ranged between 0.18 and 0.21 mg kg21 (mean, 0.2 mg
estimate quantile, estimate quantile, kg21), reaching a mean nalbuphine plasma concentration
97.5% 97.5% of 63.3 mg litre21. The maintenance dose ranged between
quantile quantile
0.73 and 0.83 mg kg21 (mean, 0.8 mg kg21). The
CL (litre h21) 130 115, 145 131 118, 144 maximum dose of 1.4 mg kg21 day21 was administered to
V1 (litre) 210 166, 255 208 164, 251 two children of 4.1 and 5.5 yr old. For simulations, we have
V2 (litre) 151 129, 173 153 128, 180 selected an average concentration of 12 mg litre21 as the
Q (litre h21) 75.6 55.0, 96.2 75.8 54.6, 96.9 therapeutic concentration for nalbuphine. It is remarkable
Inter-individual variability that in most of the patients who required additional bolus
v 2CL 0.0338 0.0140, 0.0324 0.0128, doses, the nalbuphine concentrations were lower than 10
0.0536 0.0613
mg litre21. Moreover, the majority of these patients had
v 2V 1 0.0941 0.0405, 0.0929 0.0436,
maintenance doses lower than those predicted from simu-
0.148 0.142
lated data. For a typical 3.5-yr-old child, weighing 13 kg, an
v 2V 2 Not estimated
initial bolus dose of 0.050 mg kg21 nalbuphine and a main-
v 2Q 0.121 0.0280, 0.123 0.0272,
0.215 0.218 tenance dose of 0.931 mg kg21 over 24 h achieve and main-
Residual 0.0580 0.0353, 0.0563 0.0340, tain the steady-state plasma concentration of 12 mg litre21.
variability, s 2 0.0807 0.0841 Overall, the treatment was well tolerated by the children and
none of them required morphine rescue analgesia.
Patients of this study received co-administration of mida-
zolam, propofol, remifentanil, sevoflurane, and acetamino-
children well described the relationships between clearances phen. Nalbuphine is metabolized via cytochrome P-450 3A4
and weight and volumes of distribution and weight.15 – 17 19 (CYP3A4) and 2C19.20 CYP3A4 is responsible for the metab-
In the current study, different covariate models were exam- olism of numerous therapeutic drugs including midazolam
ined: the traditional approach using a linear or power model and acetaminophen.21 22 Anaesthetic agents—remifentanil,
and the allometric model. The best results were obtained propofol, and sevoflurane—were selected to avoid clinical
using the allometric model. Such a parameterization avoids interference with pain evaluation and proved rapid elimin-
additional parameters in the model and allows a comparison ation. Midazolam also had a short elimination half-life,
of children parameter estimates with those reported in between 1.5 and 2.5 h and was given at least 3 h before nal-
adults and neonates. A large proportion of the parameter buphine.23 Acetaminophen was administered after surgery,
variability (39% and 65% of the overall variability in V1 and four times a day; it is likely that therapeutic blood levels
CL, respectively) can be attributable to the inclusion of are much higher than those of nalbuphine. Thus, a risk of
weight in the model. In our population of children, the non-competitive inhibition of the metabolism of nalbuphine
mean CL was 41.0 litre h21, mean Vss was 81.3 litre, and by acetaminophen, as previously reported for fentanyl,24
mean t1/2 elimination was 2.7 h. After weight adjustment, might occur. However, in clinical practice, acetaminophen is
CL was 2.78 litre h21 kg21 and Vss was 5.5 litre kg21. routinely associated with opioid analgesics as part of multi-
Similar CL values were found by Jaillon and colleagues4 in modal analgesic procedures; nalbuphine– acetaminophen
infants (1.5 –8.5 yr). As reported by these authors, systemic drug interactions have never been reported. Concerning the
clearance of nalbuphine expressed in litre h21 kg21 other co-administered drugs, there is no risk of pharmacoki-
decreased significantly with age (Fig. 1) and the elimination netic interactions.
half-life significantly increased. At the age of 11 yr, children In conclusion, we reported for the first time the results of
have a CL about two times lower than at the age of 1 yr. a population pharmacokinetic analysis carried out in children
In the present study, Vss and the elimination half-life were to estimate individual pharmacokinetic parameters of nalbu-
1.5–2 times higher than that reported by Jaillon and phine. This study demonstrates the importance of consider-
colleagues.4 ing and incorporating weight as a covariate in order to
The population model developed in this study predicts nal- adequately describe the drug behaviour. The allometric
buphine plasma concentrations accurately and with good power model developed in this study best reflected the
precision as evidenced by small MDPE and MDAPE values data and may be useful for dose adjustment.
(20.4% and 22.8%, respectively) and the results of the VPC
plots. The result of bootstrap analysis validation indicated
that the reliability and robustness of the parameter esti- Supplementary material
mates and thus the population pharmacokinetic model was Supplementary material is available at British Journal of
acceptable. Anaesthesia online.

564
Nalbuphine pharmacokinetics in children BJA
Conflict of interest 12 Büttner W, Finke W. Analysis of behavioural and physiological
parameters for the assessment of postoperative analgesic
None declared. demand in newborns, infants and young children: a comprehen-
sive report on seven consecutive studies. Paediatr Anaesth 2000;
References 10: 303–18
1 Errick JK, Heel RC. Nalbuphine: a preliminary review of its 13 Beal SL, Sheiner LB. NONMEM User’s Guide. San Francisco: Univer-
pharmacological properties and therapeutic efficacy. Drugs sity of California at San Francisco, 1994
1983; 26: 191– 211 14 Research Development Population Pharmacokinetics. Visual-NM
2 Beaver WT, Feise GA. A comparison of the analgesic effect of User’s Manual, version 5.1. Montpellier: Research Development
intramuscular nalbuphine and morphine in patients with post- Population Pharmacokinetics, 1998
operative pain. J Pharmacol Exp Ther 1978; 204: 487 –96 15 Anderson BJ, Meakin GH. Scaling for size: some implications for
3 Miller RR. Evaluation of nalbuphine hydrochloride. Am J Hosp paediatric anaesthesia dosing. Paediatr Anaesth 2002; 12: 205– 19
Pharm 1980; 37: 942– 9 16 Tod M, Jullien V, Pons G. Facilitation of drug evaluation in children
4 Jaillon P, Gardin ME, Lecocq B, et al. Pharmacokinetics of nalbu- by population methods and modelling. Clin Pharmacokinet 2008;
phine in infants, young healthy volunteers and elderly patients. 47: 231–43
Clin Pharmacol Ther 1989; 46: 226 –33 17 Anderson BJ, Holford NH. Mechanism-based concepts of size and
5 Nies AS, Shand DG, Wilkinson GR. Altered hepatic blood flow and maturity in pharmacokinetics. Annu Rev Pharmacol Toxicol 2008;
drug disposition. Clin Pharmacokinet 1976; 1: 135 –55 48: 303–32
6 Aitkenhead AR, Lin ES, Achola KJ. The pharmacokinetics of oral 18 Efron B, Tibshirani R. Bootstrap methods for standard errors, con-
and intravenous nalbuphine in healthy volunteers. Br J Clin Phar- fidence intervals, and other measures of statistical accuracy. Stat
macol 1988; 25: 264 –8 Sci 1986; 1: 54– 77
7 Nicolle E, Devillier P, Delanoy B, Durand C, Bessard G. Therapeutic 19 Anderson BJ, Pons G, Autret-Leca E, Allegaert K, Boccard E.
monitoring of nalbuphine: transplacental transfer and estimated Pediatric intravenous paracetamol (propacetamol) pharmacoki-
pharmacokinetics in the neonate. Eur J Clin Pharmacol 1996; 49: netics: a population analysis. Paediatr Anaesth 2005; 15:
485– 9 282– 92
8 Jacqz-Aigrain E, Debillon T, Daoud P, et al. Population pharmaco- 20 Stringer RA, Strain-Damerell C, Nicklin P, Houston JB. Evaluation
kinetics of nalbuphine in neonates. Paediatr Perinat Drug Ther of recombinant cytochrome P450 enzymes as an in vitro
2003; 5: 190– 8 system for metabolic clearance predictions. Drug Metab Dispos
2009; 37: 1025– 34
9 Bouwmeester NJ, Van Den Anker JN, Hop WC, Anand KJ,
Tibboel D. Age- and therapy-related effects on morphine require- 21 Emoto C, Murayama N, Yamazaki H. Effects of enzyme sources on
ments and plasma concentrations of morphine and its metab- midazolam 1′ -hydroxylation activity catalyzed by recombinant
olites in postoperative infants. Br J Anaesth 2003; 90: 642– 52 cytochrome P450 3A4 in combination with NADPH-cytochrome
P450 reductase. Drug Metab Lett 2008; 2: 190–2
10 de Hoog M, Mouton JW, van den Anker JN. Thoughts on ‘Popu-
lation pharmacokinetics and relationship between demographic 22 Available from http://www.medscape.com/viearticle/410825_4
and clinical variables and pharmacokinetics of gentamicin in (accessed 21 September 2010)
neonates’. Ther Drug Monit 2003; 25: 256– 7 23 Rey E, Delaunay L, Pons G, et al. Pharmacokinetics of midazolam
11 Singh J, Burr J, Stringham D, Arrieta A. Commonly used antibac- in children: comparative study of intranasal and intravenous
terial and antifungal agents for hospitalized paediatric patients. administration. Eur J Clin Pharmacol 1991; 41: 4355– 7
Implications for therapy with an emphasis on clinical pharmaco- 24 Feierman DE. The effect of paracetamol (acetaminophen) on fenta-
kinetics. Paediatr Drugs 2001; 3: 733 –61 nyl metabolism in vitro. Acta Anaesthesiol Scand 2000; 44: 560–3

565