Population pharmacokinetics of ketamine in children with heart disease

International Journal of Pharmaceutics 478 (2015) 223–231

Contents lists available at ScienceDirect

International Journal of Pharmaceutics journal homepage: www.elsevier.com/locate/ijpharm

Personalised medicine

Population pharmacokinetics of ketamine in children with heart disease Mohammed H. Elkomy a,b , David R. Drover a, * , Gregory B. Hammer a , Jeffery L. Galinkin c, Chandra Ramamoorthy a a b c

Department of Anesthesia, 300 Pasteur Drive, Stanford University, Stanford, CA 94305-5640, USA Department of Pharmaceutics and Industrial Pharmacy, Beni Suef University, Beni Suef, Egypt University of Colorado at Denver Health Science Center, Aurora, CO 80045, USA

A R T I C L E I N F O

A B S T R A C T

Article history: Received 14 July 2014 Received in revised form 16 October 2014 Accepted 12 November 2014 Available online 13 November 2014

This study aims at developing a population pharmacokinetic model for ketamine in children with cardiac diseases in order to rationalize an effective 2-h anesthetic medication, personalized based on cardiac function and age. Twenty-one children (6 months to 18 years old) were enrolled in this prospective, open label study. Ketamine 2 mg/kg IV was administered and blood samples were then collected over 8 h for ketamine assay. Pharmacokinetic data analysis using NONMEM, was undertaken. Ketamine pharmacokinetics was adequately described by a two-compartment linear disposition model. Typical population parameters were: total clearance: 60.6
(weight/70)0.75 L/h, intercompartmental clearance: 73.2
(weight/70)0.75 L/h, central distribution volume: 57.3
(weight/70) L, and peripheral distribution volume: 152
(weight/70) L. Ketamine clearance in children with pre-existing congenital heart disease was comparable to values reported in healthy subjects. Computer simulations indicated that an initial loading dose of ketamine 2 mg/kg IV over 1 min followed by a constant rate infusion of 6.3 mg/kg/h for 29 min, 4.5 mg/kg/h from 30 to 80 min, and 3.9 mg/kg/h from 80 to 120 min achieves and maintains anesthetic plasma level for 2 h in children 1 year or older (weight 10 kg). ã 2014 Published by Elsevier B.V.

Keywords: Children Ketamine Pharmacokinetics NONMEM

1. Introduction Ketamine differs from most anesthetic agents in that it stimulates the sympathetic nervous system, often resulting in increased systemic vascular resistance, cardiac output, and blood pressure (Haas and Harper, 1992). Accordingly, ketamine has been a popular agent for induction of anesthesia in young children with pre-existing congenital heart disease (Greeley et al., 2001). Ketamine was reported to be a safe alternative for anesthesia maintenance in children with both cyanotic (Tugrul et al., 2000), and non-cyanotic structural heart disease (Ulke et al., 2008). Ketamine has been used to provide anesthesia during pediatric cardiac catheterization procedures (Oklu et al., 2003), and to achieve sedation in children after cardiac surgery (Hartvig et al., 1993; Tobias et al., 1990).

* Corresponding author. Tel.: +1 650 725 0364; fax: +1 650 725 8544. E-mail address: [email protected] (D.R. Drover). http://dx.doi.org/10.1016/j.ijpharm.2014.11.026 0378-5173/ ã 2014 Published by Elsevier B.V.

Ketamine is a highly lipid soluble drug (Cohen and Trevor, 1974), that is associated with large steady-state volume of distribution (1.5–3.7 L/kg) and rapid clearance (0.7–2 L/h/kg). (Brunette et al., 2011; Clements and Nimmo, 1981; Clements et al., 1982; Domino et al., 1984, 1982; Geisslinger et al., 1993; Grant et al., 1983; Hartvig et al., 1993; Herd and Anderson, 2007; Herd et al., 2007; Hijazi et al., 2003; Malinovsky et al., 1996). Ketamine is eliminated by hepatic metabolism through N-demethylation to Nor-ketamine via CYP 3A4, CYP 2B6, and CYP 2C9 enzyme systems (Hijazi and Boulieu, 2002). However, CYP 3A4 has been shown to be the major contributor to ketamine metabolism (Hijazi and Boulieu, 2002). The high ketamine clearance rate suggests that its elimination is susceptible to factors affecting hepatic blood flow (Brunette et al., 2011; Hartvig et al., 1993; Malinovsky et al., 1996). Despite the common use of ketamine in children with cardiac disease, pharmacokinetic (PK) data in this population are sparse. To the best of our knowledge, there is only one study that investigated ketamine disposition from long-term infusions in children after cardiac surgery (Hartvig et al., 1993). The current study was undertaken to better define the PK of ketamine in children with

224

M.H. Elkomy et al. / International Journal of Pharmaceutics 478 (2015) 223–231

pre-existing congenital heart disease following a single dose of ketamine in order to rationalize an effective 2-h anesthetic medication, personalized based on cardiac function and age. 2. Material and methods 2.1. Patients, ketamine dosing and monitoring Following approval of the Institutional Review Board (IRB), twenty-one children between the ages of 6 months and 18 years were enrolled in this prospective, open label study, after informed written consent of their parents. All patients were fasted prior to surgery and received premedication according to institutional guidelines. Upon arrival in the operating room, following standard monitoring and inhalational induction of anesthesia with sevoflurane, an IV catheter was inserted and a neuromuscular blocking agent was administered. The trachea was intubated, then arterial and central venous catheters were inserted as per institutional protocol. A baseline (T0) venous blood sample was obtained, following which ketamine 2 mg/kg IV was administered as a zero order infusion over 5 min using a Baxter AS50 infusion pump (Baxter Inc., Deerfield, IL, USA). Venous blood samples were drawn at 1, 5, 10, 15, 20, 30, 45, 60 min; then at 2, 3, 4, 5, 6 and 8 h later. The blood samples were drawn into EDTA tubes that were gently mixed to mix EDTA with blood and put on ice. Samples were immediately centrifuged following collection at 1500 g for 10 min at 4  C and plasma was separated into propylene tubes that were frozen for batched assay. After a 6:1 (acetonitrile:plasma) protein precipitation, plasma samples were measured by iC42 clinical research (Aurora, CO) using a Waters Acquity UPLC-MS/MS system. The injection volume was 10 mL of supernatant. The analytical column was an Acquity UPLC HSS T3 1.8 mm 2.1
100 mm. The assay had a lower limit of quantitation (LLQ) of 0.003 mg/mL and the range of reliable response was from 0.003 to 2.5 mg/mL (r2 > 0.999), inter-day accuracy was 2.5% and inter-day precision was 3.4%. Intra-day accuracy was 4.2% and intra-day precision was 1.1%. The analytical run time was 2.5 min. 2.2. Population PK modeling Ketamine plasma concentration–time data were analyzed using the non-linear mixed effects modeling software program NONMEM (version VII; Icon Development Solutions, Ellicott City, MD). A two-compartment linear disposition model parameterized in terms of total clearance (CL), inter-compartmental clearance (Q), central (VC) and peripheral (VP) volume of distribution was used as the structural PK model. The model was specified using PREDPP subroutines ADVAN3 TRANS4. The first-order conditional estimation (FOCE) with h–e interaction was used for the estimation of the model parameters. Under the assumption that the PK parameters are log-normally distributed, the inter-individual variability was described by an exponential variance model (Eq. (1)): Pi ¼ u
expðhI Þ

(1)

where Pi is the value of the respective PK parameter in the ith individual, u is the typical (population) value of the parameter, and hI is the random (unexplained) difference between u and Pi. Values of hI are assumed to follow a multivariate normal distribution with mean zero and variance–covariance matrix, V. Covariance was permitted between each of the pharmacokinetic parameters. Proportional (Eq. (2)) and combined additive and proportional (Eq. (3)) models of residual (intra-individual) variability for PK observations were evaluated:

Y ij ¼ F ij
1 þ eij





Y ij ¼ F ij
1 þ e1ij þ e2ij

(2)

(3)

where Yij is the jth observed ketamine concentration in the ith subject, Fij is the model-predicted concentration, and e is the residual error assumed to be independently and normally distributed with mean zero and variance s 2. In addition to the two compartment model, one-compartment model (ADVAN1 TRANS2) and three-compartment model (ADVAN6 TOL5) were initially evaluated as structural models. Selection between the competing models was based on (1) the Akaike information criterion (AIC) (Beal et al., 1994), computed as two times the number of model estimated parameters added to the NONMEM minimum objective function (2 log likelihood) value (OFV); (2) the basic goodness-of-fit plots, including observed versus predicted concentrations and conditional weighted residuals versus population predictions; and (3) the precision of parameter estimates, expressed as the relative standard error (% SE) and calculated as the percentage of the standard error provided by NONMEM $COVARIANCE step to the parameter estimate. After selecting the best structural model, likelihood ratio testing under the assumption of x2-distributed difference in the NONMEM OFV (Sheiner and Ludden, 1992), was used to discriminate between several hierarchical models (e.g., diagonal/nondiagonal V, proportional/mixed e,etc.). The significance level (a) was set to 0.01, which means that a decrease in the OFV of 6.63 was necessary to consider model improvement due to an added parameter (1 degree of freedom). When more than one parameter was added, a decrease in OFV of 9.21, 11.35, or 13.28 was needed for 2, 3, and 4 degrees of freedom, respectively. Following principles of pediatric clinical pharmacology and previous population PK models in infants and children for ketamine (Brunette et al., 2011; Herd and Anderson, 2007; Herd et al., 2007), an allometric body weight (WT)-based model scaled to a 70 kg adult (Anderson and Meakin, 2002; Holford, 1996) was first implemented to account for the influence of body size on the PK parameters:   WTi PWR Pi ¼ Pstd
(4) 70kg where Pstd is the parameter value in an individual weighting 70-kg. The exponent PWR was fixed to 0.75 for clearances and 1.0 for volumes. The allometrically scaled model was the base model for further covariate model building. Besides weight, other covariates evaluated were post-natal age (AGE), gender, hypoxia, and patient physical status (ASA scores <3 and 3). Relationship between a continuous covariate and a PK parameter was modeled as linear (Eq. (5)), exponential (Eq. (6)), or power (Eq. (7)) functions:

u ¼ u0
½1 þ uCOVAR
ðCOVAR  mean COVARÞ

(5)

u ¼ u0
exp½uCOVAR
ðCOVAR  mean COVARÞ

(6)

u ¼ u0




uCOVAR COVAR mean COVAR

(7)

where u0 is the typical value of u for a subject with mean covariate value and uCOVAR is the estimated effect for the covariate on parameter value. For categorical covariates, the following model was used (Eq. (8)):

M.H. Elkomy et al. / International Journal of Pharmaceutics 478 (2015) 223–231 Table 1a Patient demographics summary. Patient characteristic

Mean (SD)

Range

21 Number Age (years) 6.44 (4.59) 0.67–16.0 Weight (kg) 24.6 (17.2) 9.05–67.0 Height (cm) 115 (29.7) 72.0–175 Body surface area (m2) 0.878 (0.393) 0.430–1.67 Gender (%) Male/female 66.7/33.3 93.4 (9.1) 71.0–100 Oxygen saturation (%) Physical status (American Society of Anesthesiologists score – ASA) (%) Normal patient (ASA = 1) 4.76 Patient with mild systemic disease 14.3 (ASA = 2) 61.9 Patient with severe systemic disease (ASA = 3) 19.0 Patient with life threatening disease (ASA = 4)

u ¼ u0
½uCOVAR COVAR

(8)

where u0 is the typical value of u for a subject belonging to the covariate reference (zero-coded) category, and uCOVAR is the estimated fractional change in u0 for the non-reference (one-coded) category. Using the basic PK model, each potential covariate was separately included and the model was tested. A covariate was considered to significantly improve the model if the decline in NONMEM OFV was 6.63 (corresponding to a likelihood ratio test at significance level a = 0.01 and 1 degree of freedom). If more than one significant covariate was found, the covariate with the greatest reduction in the OFV was added to the base model and the entire procedure was repeated until no further improvement could be achieved (stepwise forward addition approach). Quality of fit using the final model was evaluated by visual inspection of the observed versus predicted concentrations, conditional weighted residuals versus population predictions and time, and observed/predicted concentrations versus time. The precision of the final model parameters was assessed by non-parametric bootstrapping using Perl-speaks-NONMEM (Lindbom et al., 2005). The original dataset was randomly sampled with replacement, using the individual as the sampling unit, to create 1000 replicate datasets, each with the same number of

225

subjects as the original dataset. Subsequently, NONMEM was used to estimate the population parameters for each dataset. Parameter estimates from bootstrap runs with successful convergence were used to construct empirical 95% confidence interval by recording the 2.5th and 97.5th percentilesof the resulting parameter distributions. Visual predictive check (VPC) was conducted to evaluate how well the final model predicted the distribution of measured ketamine concentrations. The approach adopted in this work was the confidence interval VPC as described by Karlsson and Holford (2008). The final model was used to simulate a dataset having the same characteristics as the observed dataset (i.e., the number of subject, sampling times, doses, and body weights) using NONMEM. For each individual in the dataset a vector of parameters was drawn from a log-normal distribution with a variance corresponding to the between-subject variability previously estimated. For each simulated concentration a simulated residual error was added. Simulations were replicated 1000 times. For calculation of summary statistics and graphing purposes, observed and simulated concentrations were grouped into similar size bins. The 5th, 50th, and 95th percentiles (prediction intervals) of the simulated concentrations were calculated for each replicate. Then for these percentiles, the 95% confidence intervals across the 1000 replicates were computed and compared graphically to the 5th, 50th, and 95th percentiles of the observed data. 2.3. Simulation A simulation study was performed to determine the optimal loading dose and infusion needed to maintain anesthesia in pediatric cardiac patients for 2 h. Targeted pseudo-steady state ketamine concentration was 2 mg/mL. The final population model parameter estimates (fixed effects, between-subject and residual variability) were utilized to simulate individual plasma concentration-time profiles using NONMEM. The virtual dataset consisted of 1000 children with two different weights, 10 and 50 kg, corresponding to ages of 1 and 14 years, respectively. 3. Results Data from 21 pediatric cardiac patients were included for the population PK analysis. Tables 1a and 1b summarize patients’

Table 1b Patient demographics, cardiac diagnosis and procedure. Subject number

Cardiac diagnosis

Procedure

ASA status

Room air SpO2

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21

PA/IVS/MAPCAS Ebstein’s anomaly Heart Block TOF, PA, MAPCAS TOF, MAPCAS Ventricular tachycardia TOF, PA, MAPCAS D-TGA HLHS TOF, PA, MAPCAS s/p complete repair TOF, PA, MAPCAS Supraventricular tachycardia Supraventricular tachycardia Heart block Atrial septal defect TOF, MAPCAS HLHS, Atrial septal defect TOF, MAPCAS Congenital subaortic stenosis Corrected TGA

Unifoccalization Electrophysiology study Pacemaker lead change Unifocalization Unifocalization. Electrophysiology study Unifocalization Pulmonary artery banding Pulmonary arterio plasty RV-PA conduit change, pulmonary arterioplasty Unifocalization revision Electrophysiology study Electrophysiology study Pacemaker placement Device closure Unifocalization Fontan procedure Septal repair Unifocalization Aortic valve repair Arterial switch

3 3 3 3 3 3 4 3 3 3 3 2 2 3 1 3 4 2 4 3 3

82 95 100 78 100 100 81 100 94 99 95 100 100 100 100 88 71 99 82 97 100

PA/IVS: pulmonary atresia, intact ventricular septum, MAPCAS: multiple aortopulmonary collaterals, TOF: tetralogy of Fallot, TGA: transposition of great vessels, HLHS: hypoplastic left heart syndrome.

226

M.H. Elkomy et al. / International Journal of Pharmaceutics 478 (2015) 223–231

Fig. 1. Ketamine plasma logarithmic concentration–time profiles for different age groups: 6 months infant (A), 1–5 years children (B), 7–10 years children (C), and 11–16 years children (D).

demographic characteristics. A total of 232 plasma concentrations were used for the analysis. The median observed concentration was 0.267 (range: 0.013–3.23 mg/mL) following an IV 2 mg/kg dose administered over 5 min. No measured samples were below the lower quantification limit. Pharmacokinetic profiles for different age groups are shown in Fig. 1. The Akaike information criterion (AIC) for one-, two- and threecompartment models was 2645, 2217, and 2201, respectively. Despite that the AIC value was in favor of a three- rather than a two-compartment model (16 points difference), visual inspection of observed versus predicted ketamine concentrations as well as the conditional weighted residuals versus population predictions (data not shown) indicated similar quality of fit using the two rival models. The relative standard error (%SE) for the three-compartment model fixed effect parameters, except for clearance, was relatively large, >25% (the %SE of the central compartment volume was 48%), suggesting that the current data are not sufficient to

precisely estimate the model extra-parameters. Taking these findings into account, the two-compartment model was selected to be the structural model for ketamine PK. The NONMEM minimum objective function value (OFV) did not change significantly by using a combined proportional and additive residual error model (DOFV = 2) and the additive error variance was <0.0001. Therefore, a proportional model was used. Using body weight to allometrically scale model clearances and volumes decreased the between subject variability for CL, Q, and VP by 57.4, 39, and 19.5%, respectively. Plotting the random effects for CL as function of body weight (Fig. 2) revealed strong correlation that was significantly reduced upon inclusion of the allometric models. Additionally, using a power model with the exponent as an estimable parameter resulted in an unstable model, where the final parameter estimates were considerably affected by initial values. This may be attributed to the relatively small number of subjects included in the study. The allometric model was the

Fig. 2. Pattern of between subject variability in clearance, CL (hCL) versus body weight, before (left panel) and after (right panel) including allometric covariate relationships of weight on CL, intercompartmental clearance (Q), central (VC) and peripheral (VP) volumes of distribution. Solid line is a lowess smoother.

M.H. Elkomy et al. / International Journal of Pharmaceutics 478 (2015) 223–231

227

Table 2 Ketamine population pharmacokinetic parameter estimates and bootstrap statistics using the final pharmacokinetic model. Parameters

Estimate (% SE)

% BSV

Bootstrap median [95% C.I.]

CLstd (L/h/70 kg) VCstd (L/70 kg) Qstd (L/h/70 kg) VPstd (L/70 kg) Residual error (proportional %)

60.6 (8.2) 57.3 (12.0) 73.2 (18.1) 152 (13.8) 16.1 (15.5)

34.2 53.5 74.4 54.2

60.0 [51.4, 72.0] 57.6 [45.0, 82.6] 70.8 [49.7, 106] 149 [114,206] 16.1 [13.4, 18.7]

–

% SE: relative standard error; % BSV: between-subject variability (square root of v multiplied by 100%); C.I.: confidence interval; CLstd, VCstd, Qstd, and VPstd are total clearance, intercompartmental clearance, central distribution volume, and peripheral distribution volume standardized to a 70-kg person using allometric models; proportional residual error %: calculated as square root of s 2 multiplied by 100%. 2

Table 3 Correlation of between subject variability for the final pharmacokinetic model parameters.

CL VC Q VP

CL

VC

Q

VP

1 0.608 0.423 0.481

1 0.209 0.571

1 0.563

1

CL: total clearance; Q: intercompartmental clearance; VC and VP: central and peripheral volume of distribution.

base model for further covariate testing. Post-natal age, gender, oxygen saturation level or ASA scores failed to demonstrate significant effects on any of the model parameters. Table 2 lists the final population PK model parameter estimates with their bootstrap median and non-parametric 95% confidence

intervals. The point estimates from the original dataset were very similar to the median values obtained from the bootstrap analysis, indicating stability and robustness of the population PK model. The model fixed effects parameters were estimated with reasonable precision (%SE <19%). Furthermore, the 95% confidence intervals for the standardized clearance, CLstd and central distribution volume, VCstd were relatively narrow, confirming their high estimation precision. Parameter precisions corresponding to the between-subject variability of CL, VC, and Q were relatively low, with %SE >40%. Between-subject random effects exhibited nonGaussian distribution around the point estimates. Allowing covariance between CL, Q, VC, and VP significantly improved the model fit (DOFV = 57). The correlation values (Table 3) for most cases were >0.4, suggesting lack of independence. Basic goodness-of-fit plots of the final population PK model are depicted in Fig. 3. The near-symmetric distribution of observed

Fig. 3. Basic goodness-of-fit plots for the final model of ketamine population pharmacokinetics showing observed versus population and individual predicted concentrations on logarithmic scales (upper left and right panels); and conditional weighted residuals versus population predicted concentrations and time (lower left and right panels).

228

M.H. Elkomy et al. / International Journal of Pharmaceutics 478 (2015) 223–231

ketamine concentrations as a function of population predictions around the identity line (Fig. 3, upper left panel), as well as the good agreement between observed and individual predicted concentrations (Fig. 3, upper right panel) indicate that the model fit the data well at the population and subject levels. The plot of conditional weighted residuals versus population predicted concentrations (Fig. 3, lower left panel), or over time (Fig. 3, lower right panel), showed random distribution around the line of zero, indicating lack of any patterns in the model fit across the plasma concentration or time ranges in this study. Goodness-of-fit over the study period was further confirmed in Fig. 4. Ratio of observed to population predicted concentrations fell between 0.2 and 2 for 97% of the observations, and 100% of the observed/individual predictions were between 0.5 and 1.44. The visual predictive check of the final population PK model (Fig. 5) shows that the observed concentrations 5th, 50th, and 95th percentiles were consistently within the 95% confidence intervals

of the simulated concentrations percentiles. This finding suggests that the model can adequately replicate the features of the observed data and confirms that the variability parameters were well estimated. Ketamine plasma concentration-time profiles in 1000 virtual patients, following different dosing regimens were simulated using the final population PK model. Fig. 6 shows that an initial loading dose of 2 mg/kg over 1 min followed by a constant rate infusion of 6.3 mg/kg/h for 29 min, 4.5 mg/kg/h from 30 to 80 min, and 3.9 mg/kg/h from 80 to 120 min achieves and maintains median plasma concentration above 2 mg/mL for 2 h in children 1 year or older (weight 10 kg). 4. Discussion Little is known about the influence of cardiac diseases on ketamine disposition. We developed a population PK model for

Fig. 4. Quality of fit for pharmacokinetic model over the study time period. The points represent ratio of observed to population predicted (top panel) and individual predicted (bottom panel) ketamine concentrations on logarithmic scale. A line connects each subject's data.

M.H. Elkomy et al. / International Journal of Pharmaceutics 478 (2015) 223–231

229

Fig. 5. Visual predictive check of the final pharmacokinetic model on normal (left panel) and logarithmic scale (right panel). Dashed lines represent the 5th, 50th, and 95th percentile of observed concentrations. Shaded areas represent the 95% C.I. for the 5th, 50th, and 95th percentile of simulated concentrations. Points represent the observed concentrations.

ketamine administered to a group of 21 pediatric patients with heart disease undergoing procedures. Model evaluation using diagnostic plots (Figs. 3 and 4), bootstrap analysis (Table 2), and visual predictive check (Fig. 5) indicated that the model described the data well, was unbiased through the observations or time ranges, robust, stable, and able to predict the distribution of the observed data through stochastic simulations. Ketamine kinetics appears to be adequately described by a two-compartment model corresponding to plasma and tissue distribution. This is consistent with the finding of other investigators that ketamine disposition follows a bi- or tri-exponential decay depending upon duration and frequency of sampling (Brunette et al., 2011; Clements and Nimmo, 1981; Clements et al., 1982; Domino et al., 1984, 1982; Hartvig et al., 1993; Herd and Anderson, 2007; Herd et al., 2007; Hijazi et al., 2003; Malinovsky et al., 1996). Although a three-compartment model gave a slightly better fit to the data in this study, the risk of over parameterization justified its rejection. Clearance and steady-state volume of distribution in this study were in range of values reported previously in children or adults comprising healthy or diseased individuals (Table 4). The high steady-state volume in our study (3.4 L/kg) is consistent with a

high drug lipophilicity (Cohen and Trevor, 1974), and extensive distribution within the body (White et al., 1976). Our estimated clearance best matched the value reported by Malinovsky et al., (1996) (Table 4) from a study of 3 mg/kg IV ketamine dose in a group of 8 healthy children (ASA score of 1) undergoing minor urological surgery. Thus, it seems that ketamine clearance may not be influenced by cardiac disease. The pediatric patients in this study had a large variety of cardiac anomalies and thus it may be assumed that cardiac output and subsequent hepatic clearance would be affected. We did not have a direct measure of cardiac output and this would limit our ability to accurately determine a change in the pharmacokinetics of ketamine caused by a change in cardiac output. In this study, anesthesia was induced with sevoflurane. The main metabolic pathway for sevoflurane is defluoration via CYP 2E1 (Kharasch et al., 1995). Since CYP 3A4 is the primary enzyme involved in ketamine N-demethylation to its major metabolite, norketamine (Hijazi and Boulieu, 2002), it is most likely that sevoflurane does not affect ketamine clearance through competition for liver enzymatic system. Sevoflurane maintains hepatic arterial but decreases portal venous blood flow in dogs at 1.5 and 2.0 MAC (Frink et al., 1992). Because ketamine clearance rate is

Fig. 6. Simulated ketamine plasma concentration profiles in 1 year-old, 10 kg (A) and 14 year-old, 50 kg (B) children, receiving a loading dose of 2 mg/kg over 1 min followed by a constant rate infusion of 6.3 mg/kg/h for 29 min, 4.5 mg/kg/h from 30 to 80 min, and 3.9 mg/kg/h from 80 to 120 min. The lines represent the median of the simulations. The shaded areas represent the 90% prediction intervals of the concentrations.

230

M.H. Elkomy et al. / International Journal of Pharmaceutics 478 (2015) 223–231

perfusion-rate limited, it may be reduced through sevoflurane induced change in hepatic blood flow. However, the study design does not allow quantifying the magnitude of sevoflurane effect on ketamine disposition. CYP 3A4 enzymes reach maturity within the first year of life (Edginton et al., 2006). Inability to detect age-dependent changes in clearance after adjusting for body size may be attributed to the fact that 20 of 21 subjects investigated were older than 1 year (one subject was 6 months old). This observation is in accordance with other reports (Herd et al., 2007). Additionally, failure to observe an effect for patient physical condition, classified by ASA scores, is

Table 4 Comparison of ketamine clearance and steady-state distribution volume values reported in children and adults. Study

Population Dosage and CL administration (L/h/ kg)

Present study

Children

Grant et al. (1983) Hartvig et al. (1993) Malinovsky et al. (1996) Herd and Anderson (2007) Herd et al. (2008)

Children Children Children Children

Children, adults

Brunette et al. Children, (2011) adults

Clements and Nimmo (1981) Clements and Nimmo (1981) Domino et al. (1982) Grant et al. (1983) Domino et al. (1982) e Domino et al. (1982) e Geisslinger et al. (1993) f Geisslinger et al. (1993) f Hijazi et al. (2003)

Adults

2 mg/kg I.V. bolus 2 mg/kg I.V. bolus 1 or 2 mg/h/kg I.V. infusion 3 mg/kg I.V. bolus 1–1.5 mg/kg I. V. bolus

CLstda (L/ VSS h/70 kg) (L/ kg)

VSSstdb (L/70 kg)

1.30c

60.6

3.4c

1.02

53.2

1.9

133

0.96

40.2

3.7

259

1.32

66.3

2.8

196

–

1–3 mg/kg I.V. – bolus 0.5 or 6 mg/kg I.M 0.5 mg/kg oral 0.5–3 mg/kg I. – V. bolus 0.5 or 6 mg/kg I.M 0.5, 5 or 10 mg/ kg oral 0.1 mg/kg/h I. V. infusion 0.25 mg/kg I.V. 1.15 bolus

90

60

81

–

–

–

209d

141d

151d

130d

81.8

3.1

217

Adults

0.125 mg/kg I. V. bolus

0.98

69.7

2.1

147

Adults

2–2.2 mg/kg I. V. bolus 2 mg/kg I.V. bolus 2.2 mg/kg I.V. bolus 2.2 mg/kg I.V. bolus 2 mg/kg I.V. bolus

1.25

85.7

2.27

159

0.76

53.0

2.3

161

0.85

59.9

1.78

125

0.72

50.8

1.52

106

Adults Adults Adults Adults

1.15

Adults

2 mg/kg I.V. bolus

0.99

Adults

2 mg/kg I.V. bolus 2 mg/kg I.V. infusion

2.16

–

– 152

3.29

3.01

NR

–

–

understandable because only 19% of the subjects were relatively healthy (ASA <3), while 81% were severely ill (ASA 3) (Table 1). Similarly, only six out of twenty-one patients (28%) had oxygen saturation levels <90%, justifying the lack of hypoxia effect on PK parameters. A limitation of our study is that measures of hepatic function were not available for investigation as potential covariates on ketamine clearance. A simulation study was performed to rationalize an effective ketamine dosing regimen for 2 h. A limitation for this study is that prediction of an effective constant rate infusion regimen was based on a model developed using bolus data. Bolus and infusion data would have provided a richer set to characterize ketamine kinetics. In the simulation study, 2 mg/mL was chosen to be the ketamine anesthetic effect cut-off value. Ketamine plasma concentrations associated with anesthesia in children are highly variable. Grant et al. (1983) reported concentrations of 0.9–3.8 mg/mL at the time of consciousness recovery (expressed as purposeful movement in response to command) in a group of 9 children (1.5–3 years) using ketamine as the sole anesthetic. Herd et al. (2008) described that ketamine concentrations of 1–1.5 mg/mL were associated with arousal due to verbal or sustained painful stimulus in 95% of 43 children (1.5–14 years) receiving ketamine for short-term procedures in the emergency department. Hartvig et al. (1993) reported that post-operative cardiac children (n = 10, age: 1 week to 30 months) were aroused at concentration below 1–1.5 mg/mL and fully awake below 0.5 mg/mL. Our selection for a target therapeutic concentration of 2 mg/mL was to exceed the average of the upper limits for awakening and arousal endpoints from Grant et al. (1983), Herd et al. (2008), and Hartvig et al. (1993) weighted by the number of subjects in each study. Simulations based on our model (Fig. 6) indicates that ketamine plasma concentrations following administration of the same constant rate infusion per kg change with body weight (or age). Concentration at steady-state is determined by the ratio of the infusion rate to clearance. While infusion rate is scaled linearly with weight, the formalism of the covariate model used in this study assumes non-linear dependence of the clearance parameters on weight with power less than 1 (Eq. (4)), causing the infusion/ clearance ratio to increase non-linearly with weight. Therefore, higher infusion rates per kg are required to produce steady-state concentrations in younger (smaller size) children comparable to those in older (larger size) children. This is in line with the findings of previous studies that reported inverse correlation between age and ketamine dose per kg required for surgical anesthesia in children (Lockhart and Nelson, 1974); longer anesthesia duration in adults compared to children following the administration of 6 mg/ kg intramuscular ketamine (Nimmo and Clements, 1981); and inverse relationship between ketamine weight-adjusted clearance and age (Herd and Anderson, 2007). Due to immaturity of the enzymatic system responsible for ketamine metabolism in infants <1 year-old (Edginton et al., 2006), the influence of age on ketamine concentrations as predicted by our model may not be valid for this age group. 5. Conclusions

–

a If not reported in the original study, standardized clearance was calculated as: CLstd = CL (L/h/kg)
average weight (kg)
(70 kg/average weight (kg))0.75. b If not reported in the original study, standardized steady state volume of distribution was calculated as: VSSstd = VSS (L/kg)
70 kg. c Average of posthoc parameter estimates normalized by weight. d Calculated as: standardized central volume (VCstd) + standardized peripheral volume (VPstd). e Unmedicated and diazepam medicated, respectively. f S(+)- and R(—)-ketamine, respectively, from a racemic mixture.

In conclusion, we developed a population PK model for ketamine in pediatric patients with heart disease and applied the model to rationalize an effective 2-h anesthetic dosing for children 1 year of age (weight 10 kg). Ketamine clearance in children with pre-existing congenital heart disease is comparable to values reported in healthy subjects. Conflict of interest statement The authors declare no conflict of interest.

M.H. Elkomy et al. / International Journal of Pharmaceutics 478 (2015) 223–231

Acknowledgement This work was supported by an internal departmental grant to Chandra Ramamoorthy. References Anderson, B.J., Meakin, G.H., 2002. Scaling for size: some implications for paediatric anaesthesia dosing. Paediatr. Anaesth. 12, 205–219. Beal, S.L., Sheiner, L.B., Boeckmann, A.J., 1994. NONMEM User's Guides - Part V. NONMEM Project Group. University of California at San Francisco, San Francisco, CA, USA. Brunette, K.E., Anderson, B.J., Thomas, J., Wiesner, L., Herd, D.W., Schulein, S., 2011. Exploring the pharmacokinetics of oral ketamine in children undergoing burns procedures. Paediatr. Anaesth. 21, 653–662. Clements, J.A., Nimmo, W.S., 1981. Pharmacokinetics and analgesic effect of ketamine in man. Br. J. Anaesth. 53, 27–30. Clements, J.A., Nimmo, W.S., Grant, I.S., 1982. Bioavailability, pharmacokinetics, and analgesic activity of ketamine in humans. J. Pharm. Sci. 71, 539–542. Cohen, M.L., Trevor, A.J., 1974. On the cerebral accumulation of ketamine and the relationship between metabolism of the drug and its pharmacological effects. J. Pharmacol. Exp. Ther. 189, 351–358. Domino, E.F., Domino, S.E., Smith, R.E., Domino, L.E., Goulet, J.R., Domino, K.E., Zsigmond, E.K., 1984. Ketamine kinetics in unmedicated and diazepampremedicated subjects. Clin. Pharmacol. Ther. 36, 645–653. Domino, E.F., Zsigmond, E.K., Domino, L.E., Domino, K.E., Kothary, S.P., Domino, S.E., 1982. Plasma levels of ketamine and two of its metabolites in surgical patients using a gas chromatographic mass fragmentographic assay. Anesth. Analg. 61, 87–92. Edginton, A.N., Schmitt, W., Voith, B., Willmann, S., 2006. A mechanistic approach for the scaling of clearance in children. Clin. Pharmacokinet. 45, 683–704. Frink Jr., E.J., Morgan, S.E., Coetzee, A., Conzen, P.F., Brown Jr., B.R., 1992. The effects of sevoflurane, halothane, enflurane, and isoflurane on hepatic blood flow and oxygenation in chronically instrumented greyhound dogs. Anesthesiology 76, 85–90. Geisslinger, G., Hering, W., Thomann, P., Knoll, R., Kamp, H.D., Brune, K., 1993. Pharmacokinetics and pharmacodynamics of ketamine enantiomers in surgical patients using a stereoselective analytical method. Br. J. Anaesth. 70, 666–671. Grant, I.S., Nimmo, W.S., Mcnicol, L.R., Clements, J.A., 1983. Ketamine disposition in children and adults. Br. J. Anaesth. 55, 1107–1111. Greeley, W.J., Kern, F.H., Schulman, S.R., 2001. Congenital heart disease, In: Estafanous, F.G., Barash, P., Reves, J.G. (Eds.), Cardiac Anesthesia: Prinicples and Clinical Practice. Second ed. Lippincott Williams & Wilkins, Philadelphia, PA, pp. 585–614. Haas, D.A., Harper, D.G., 1992. Ketamine: a review of its pharmacologic properties and use in ambulatory anesthesia. Anesth. Prog. 39, 61–68. Hartvig, P., Larsson, E., Joachimsson, P.O., 1993. Postoperative analgesia and sedation following pediatric cardiac surgery using a constant infusion of ketamine. J. Cardiothorac. Vasc. Anesth. 7, 148–153.

231

Herd, D., Anderson, B.J., 2007. Ketamine disposition in children presenting for procedural sedation and analgesia in a children's emergency department. Paediatr. Anaesth. 17, 622–629. Herd, D.W., Anderson, B.J., Holford, N.H.G., 2007. Modeling the norketamine metabolite in children and the implications for analgesia. Paediatr. Anaesth. 17, 831–840. Herd, D.W., Anderson, B.J., Keene, N.A., Holford, N.H., 2008. Investigating the pharmacodynamics of ketamine in children. Paediatr. Anaesth. 18, 36–42. Hijazi, Y., Bodonian, C., Bolon, M., Salord, F., Boulieu, R., 2003. Pharmacokinetics and haemodynamics of ketamine in intensive care patients with brain or spinal cord injury. Br. J. Anaesth. 90, 155–160. Hijazi, Y., Boulieu, R., 2002. Contribution of CYP3A4, CYP2B6, and CYP2C9 isoforms to N-demethylation of ketamine in human liver microsomes. Drug Metab. Dispos. 30, 853–858. Holford, N.H.G., 1996. A size standard for pharmacokinetics. Clin. Pharmacokint. 30, 329–332. Karlsson, M.O., Holford, N.H., 2008. A tutorial on visual predictive checks (Abstr 1434). PAGE meeting. Kharasch, E.D., Armstrong, A.S., Gunn, K., Artru, A., Cox, K., Karol, M.D., 1995. Clinical sevoflurane metabolism and disposition. II. The role of cytochrome P450 2E1 in fluoride and hexafluoroisopropanol formation. Anesthesiology 82, 1379–1388. Lindbom, L., Pihlgren, P., Jonsson, E.N., 2005. PsN-toolkit—a collection of computer intensive statistical methods for non-linear mixed effect modeling using NONMEM. Comput. Methods Programs Biomed. 79, 241–257. Lockhart, C.H., Nelson, W.L., 1974. The relationship of ketamine requirement to age in pediatric patients. Anesthesiology 40, 507–508. Malinovsky, J.M., Servin, F., Cozian, A., Lepage, J.Y., Pinaud, M., 1996. Ketamine and norketamine plasma concentrations after i.v.: nasal and rectal administration in children. Br. J. Anaesth. 77, 203–207. Nimmo, W.S., Clements, J.A., 1981. Pharmacokinetics of ketamine. Br. J. Anaesth. 53, P186. Oklu, E., Bulutcu, F.S., Yalcin, Y., Ozbek, U., Cakali, E., Bayindir, O., 2003. Which anesthetic agent alters the hemodynamic status during pediatric catheterization? Comparison of propofol versus ketamine. J. Cardiothorac. Vasc. Anesth. 17, 686–690. Sheiner, L.B., Ludden, T.M., 1992. Population pharmacokinetics/dynamics. Annu. Rev. Pharmacol. Toxicol. 32, 185–209. Tobias, J.D., Martin, L.D., Wetzel, R.C., 1990. Ketamine by continuous infusion for sedation in the pediatric intensive-care unit. Crit. Care Med. 18, 819–821. Tugrul, M., Camci, E., Pembeci, K., Telci, L., Akpir, K., 2000. Ketamine infusion versus isoflurane for the maintenance of anesthesia in the prebypass period in children with tetralogy of Fallot. J. Cardiothorac. Vasc. Anesth. 14, 557–561. Ulke, Z.S., Kartal, U., Sungur, M.O., Camci, E., Tugrul, M., 2008. Comparison of sevoflurane and ketamine for anesthetic induction in children with congenital heart disease. Pediatr. Anesth. 18, 715–721. White, P.F., Marietta, M.P., Pudwill, C.R., Way, W.L., Trevor, A.J., 1976. Effects of halothane anesthesia on the biodisposition of ketamine in rats. J Pharmacol. Exp. Ther. 196, 545–555.