ORIGINAL ARTICLE
Population Pharmacokinetics of Omeprazole in Critically Ill Pediatric Patients Maria Jose Solana, MD,* Helena Colom, PhD,† Jesús López-Herce, PhD, MD,* Javier Urbano, MD,* Rafael González, MD,* Jorge López, MD,* Cecilia Manzanares, PhD,‡ and Angel Carrillo, PhD, MD*
Background: To develop a population pharmacokinetic model for intravenous omeprazole in critically ill children.
Methods: One hundred eighty-six omeprazole concentration-time data from 40 critically ill children were analyzed using the nonlinear mixed-effects approach with the nonlinear mixed-effects modeling software, version 7.2 software. Patients were randomized into 2 groups and received intravenous omeprazole at a dose of 0.5 or 1 mg/kg twice daily. Blood samples were drawn at 0.5, 2, 6, 12, 24, and 48 hours after the first infusion. Results: The pharmacokinetic profile was best described by a 2-compartment model with a first-order elimination process. Betweenpatient variability could only be associated with plasma clearance (CL). The typical values for plasma CL were 24.9 L$h21$70 kg21 (10.08%), with a distributional clearance of 53.9 L$h21$70 kg21 (11.00%) and central and peripheral compartment distribution volumes of 4.23 L/70 kg (19.62%) and 674 L/70 kg (0.89%), respectively. Allometric size models seemed to predict changes adequately in all the pharmacokinetic parameters. High values of between-patient variability of CL [75.50% (2.60%)] and residual variability [130.0% (5.26%)] were still found in the final model. Model-based simulations suggested that the most suitable dose was 1 mg/kg because this yielded similar exposure (defined by the area under the concentration-time curve) to that obtained in adults after a 20-mg dose of omeprazole intravenously.
Conclusions: An allometric size model allows changes to be predicted in all the pharmacokinetic parameters, making dose adjustment by body weight important to achieve the most effective omeprazole exposure. This is the first step toward a population pharmacokinetic study, including more data to develop a predictable model to be used during therapeutic drug monitoring.
Received for publication March 10, 2013; accepted November 11, 2013. From the *Paediatric Intensive Care Department, Hospital General Universitario Gregorio Marañón, Madrid, Spain; †Biopharmacy and Pharmacokinetics Department, School of Pharmacy, University of Barcelona, Spain; and ‡Pharmacy Service, Hospital General Universitario Gregorio Marañón, Madrid, Spain. Supported by a grant from the Spanish Health Institute Carlos III (grants N. EC07/90670 and RD08/0072: Maternal, Child Health and Development Network) within the framework of the VI National I + D + i Research Programme (2008–2011). The authors declare no conflicts of interest. Correspondence: Jesús López-Herce, PhD, MD, Servicio de Cuidados Intensivos Pediátricos, Hospital General Universitario Gregorio Marañón, Dr Castelo, 47 28009 Madrid, Spain (e-mail: [email protected]). Copyright © 2013 by Lippincott Williams & Wilkins
Ther Drug Monit Volume 36, Number 4, August 2014
Key Words: omeprazole, population pharmacokinetics, nonlinear mixed-effects models, critically ill children, therapeutic drug monitoring (Ther Drug Monit 2014;36:519–527)
INTRODUCTION Gastrointestinal bleeding in critically ill patients is usually secondary to acute lesions of the gastric mucosa and is associated with increased morbidity and mortality.1–4 Acid gastric pH is one of the factors that influence the development of acute lesions of the gastric mucosa.5,6 The proton pump inhibitors (PPIs) are drugs of choice for the prophylaxis and treatment of digestive tract hemorrhage in critically ill patients.7 There are 5 types of PPI; their pharmacokinetics vary because of differences in their molecular structure.7 Omeprazole was the first PPI and is still the most widely used. There are very few studies that have analyzed the pharmacokinetics of PPI in children8–12 and in critically ill patients.13 Critically ill patients have certain pharmacokinetic characteristics that affect the absorption, metabolism, bioavailability, and excretion of drugs. These pharmacokinetic alterations are the result of organ dysfunction (particularly of the liver and kidney), the release of acute-phase reactants, therapeutic interventions and interactions between different drugs. Hypovolaemic states, heart failure, and the alphaagonists decrease hepatic blood flow, producing a fall in drug clearance.14 In contrast, vasodilators favor clearance.15 Inflammatory cytokines and stress hormones can inhibit cytochrome P450, altering the metabolism of omeprazole. Conjugation is also affected, though to a lesser extent.15,16 In addition, drug interactions with drugs frequently prescribed to critically ill patients can also affect PPI metabolism. The objective of this study was to investigate the pharmacokinetics of intravenous omeprazole in critically ill children and to analyze the influence of demographic factors (age, weight) and of different doses on the pharmacokinetic behavior of the drug.
MATERIAL AND METHODS A prospective, randomized, open-label clinical trial (Eudra-CT no: OM1//2007-006102-19) was performed after approval of the protocol by the hospital ethics committee. Patients admitted to the pediatric intensive care unit whose
519
Solana et al
Ther Drug Monit Volume 36, Number 4, August 2014
parents signed the informed consent form were included in the study. Exclusion criteria included treatment with H2receptor antagonists or PPIs at the time of admission, any contraindication for the insertion of a nasogastric tube and gastric hemorrhage before study entry. Patients were randomly assigned to 1 of the 2 therapeutic groups, each of 20 patients: group A (omeprazole 0.5 mg/kg per intravenously) and a group B (omeprazole 1 mg/kg per intravenously). The omeprazole infusion was prepared by diluting a 40-mg vial of omeprazole in 100 mL of normal saline and administered by slow intravenous infusion over 20 minutes via a continuous infusion pump every 12 hours, with a maximum of 40 mg per dose.
mixed-effects modeling (NONMEM) software, version 7.2 (ICON Development Solutions, Ellicott City, MD) (Bauer Maryland 2011). Graphical diagnostics were assessed using Xpose version 4.2.1 (http://xpose.sourceforge.net) implemented into R version 2.11.1 (http://www.r-project.com)17 and Perl speaks-NONMEM (PsN) version 3.5.3 Tool-kit (http://psn.sourceforge.net).18 Noncompartmental analysis was performed using WinNonlin version 5.3. Pooled concentration-time data were simultaneously analyzed after log transformation. The modeling process consisted of 3 steps: (1) base pharmacokinetic model development, (2) covariate selection, and (3) model evaluation.
Measurement of the Plasma Concentration of Omeprazole
One, 2, and 3-compartment models with first-order and nonlinear elimination processes from the central compartment were tested to characterize the changes in omeprazole plasma concentrations over time. The models were parameterized in terms of clearances (CL) and apparent volumes of distribution (V) and VM (maximal elimination rate) and KM (concentration of the drug at which the elimination is half maximal) in the case of Michaelis–Menten kinetics. A mixture model to identify subpopulations between extensive and poor metabolizers was also tested. Between-patient variability (BPV) was evaluated for each PPK parameter and was modeled exponentially, assuming a lognormal distribution. Residual variability was modeled using an additive error model on a logarithmic scale. The difference in the minimum objective function value (MOFV) was used to distinguish between nested models because this difference presents an approximate x2 distribution. Significance was taken as a P value ,0.005, which corresponded to a difference in the MOFV of 7.879 for 1 degree of freedom. For nonhierarchical models, the most parsimonious model with the lowest objective function according to the Akaike information criterion was chosen.19 Some data were reported as concentrations below the limit of quantification. Below the limit of quantification values were included during the analyses and treated as censored information using the M3 method.20 The Laplacian numerical estimation method was used for parameter estimation.
The plasma concentrations of omeprazole were measured at 30 minutes, 2 hours, and 6 hours after the first dose and before the doses at 12, 24, and 48 hours. The samples were stored in lithium heparin tubes until processing. After centrifugation at 3100 rpm for 10 minutes at 48C, 2 aliquots of 0.75 mL of plasma were extracted and stored in round-bottomed plastic tubes at 2808C until their analysis using rapid resolution liquid chromatography tandem mass spectrometry.
Assay of Omeprazol The assay was performed according to the Food and Drug Administration Guidance for Industry Bioanalytical Method Validation, Biopharmaceutics Coording Committee in the Center for Drug Evaluation and Research in cooperation with the Center for Veterinary Medicine at the Food and Drug Administration (May 2001) and previous publications.17 To each 0.2-mL volume of plasma sample, 50 mL of the internal standard (lansoprazol purchased from Sigma Aldrich, purity .99%, respectively) (1 mg/mL) was added and mixed. Working solutions were obtained by dilution of stock solutions using calibrated pipettes and volumetric flasks. Then, 2 mL of ethyl acetate was added. Extraction was performed by shaking for 30 seconds and then by centrifugation at 3000 rpm for 10 minutes. The organic layer was separated and evaporate to dryness under a steady stream of nitrogen N2 (458C) and the extract was reconstituted with 100 mL of water/methanol (1:9) (v/v). Chromatographic analyses were performed using a highperformance liquid chromatography binary pump (Agilent 1200 Chemstation, Agilent Technology Co, Ltd) and a massspectrometer (API 4000, Applied Byosystems). After sample treatment, omeprazol in human plasma and the internal standard lansoprazol was chromatographed on a Luna C18 100A (4.6 mm · 50 mm, 5 microm, Phenomenex) column with a mobile phase of 10 mM ammonium acetate, pH 4.0, and acetonitrile. Detection was performed on a triple quadrupole tandem mass spectrometry by multiple reaction monitoring in positive mode. The limit of detection was 1 mg/L and the assay exhibited a linear range between 1 and 500 mg/L.
Population Pharmacokinetic Analysis The population pharmacokinetic (PPK) model development and simulations were performed with the nonlinear
520
Base Population Pharmacokinetic Model
Covariate Model Allometric weight scaling was added to all clearance and volume fixed effects a priori and standardized to a body weight of 70 kg21,22 according to the following relationships: Pj = Pstd$(WGT/70)PER, where Pj is the parameter of the jth individual and Pstd is the typical value of the PPK parameter for an adult weighing 70 kg. As previously reported,23 the power parameter was 0.75 for clearances and 1 for distribution volumes.24 The effects of age and gender were then tested on the PPK parameters to identify possible differences not explained by the allometric models. The effect of age was investigated using a sigmoid hyperbolic or Hill model as follows: Maturation process (MF) = AGEHill/TM50Hill + AGEHill, where TM50 describes the maturation half time, while the Hill coefficient relates the slope of this maturation profile.23 Categorical covariates such as gender were tested in their respective parameters as indicated by the relationship: TVPj = u1 for Z = 0 and TVPj = u1. u2 for Z = 1, where 2013 Lippincott Williams & Wilkins
Ther Drug Monit Volume 36, Number 4, August 2014
Z values represent each level of the categorical covariate. Specifically, in the case of gender, TVPj is the typical value of the jth pharmacokinetic parameter for males and u2 is the fractional change in u1 by females. Multiplicative equations were used to describe the combined effect of multiple covariates on the same parameter. Covariates were first tested univariately in the model and then by cumulative forward inclusion/backward elimination procedures. Significance levels of 5% (reduction .3.841 units in the MOFV) and 0.1% (increase .10.8 units in the MOFV) were employed during the forward inclusion and backward elimination steps. The decrease in the MOFV (22 · log likelihood), parameter precision expressed as the relative standard error (RSE%), reductions in BPV associated with a specific pharmacokinetic parameter, model completion status (eg, successful convergence or termination), and visual inspection of goodness-of-fit plots with Xpose were also taken into account for model selection. A decrease of at least 10% in interpatient variability associated with a specific pharmacokinetic parameter was considered clinically relevant for the inclusion of that specific covariate.
Evaluation of Predictive Ability The extent of Bayesian shrinkage was also evaluated for each parameter in the final population pharmacokinetic model.24 The ability of the parameter estimates to describe the observed data was further evaluated by exploring the results from a visual predictive check (VPC), posterior predictive check (PPC),25 and normalized prediction distribution errors (NPDE).26 For the VPC, 1000 individual profiles were simulated and the 5th, 50th, and 95th percentiles were calculated for each sampling time point and represented with the raw data. If the model adequately described the data, the lines corresponding to the median and to the 5th and 95th percentiles of the observed data would mainly fall in the area covering the 90% confidence interval. A prediction-corrected VPC was also performed.27 By using prediction-corrected VPC, both the observations and the model predictions were normalized for the typical model predictions in each bin of independent variables. The 95% confidence intervals for the median, the 5th and 95th percentiles of the predicted data were then calculated and plotted together with the median and the 5th and 95th percentiles of the observed data. If the model described the data adequately, the lines corresponding to the median and to the 5th and 95th percentiles of the observed data would fall in the respective 95% confidence intervals of the predicted data. For the PPC, we investigated how well the model predicted the values of the area under the concentration-time curve (AUC) of omeprazole from 0 to 12 hours by simulating 1000 data sets as the original data set. The AUC0–12 h values were then calculated and compared with those calculated from the observed data by noncompartmental analysis using the WinNonlin package. The AUC was selected for evaluation because this parameter was the most relevant surrogate marker of efficacy of omeprazole28 and was expected to vary with concentration and with relevant covariates. For the NPDE, 1000 individual profiles, as for the original data set, 2013 Lippincott Williams & Wilkins
Population Pharmacokinetics of Omeprazole
were simulated from the final model. Then, the differences between each observation and simulation were calculated and then normalized by using the inverse of the cumulative density function. If the model fitted the data adequately, NPDEs should result in a normal distribution with a mean of 0 and variance of 2.
Model-Based Simulations
Once the final model was achieved, the simulated values of the AUC0–12 h after doses of 0.5, 1, 1.5, and 2 mg/kg were investigated. A new data set containing 40 patients with the same demographics as the patients of the target population was created. Using the final model, 1000 simulations were run in which subjects received doses of 0.5, 1, 1.5, and 2 mg/kg. The resulting simulated concentration-time data for each dose were used to calculate median AUC0–12 h values. Pediatric median AUC0–12 h values after the 4 doses were divided by the previously reported AUC0–N values after the intravenous administration of 20 mg of omeprazole to adults29 to give a ratio. Values closest to 1 indicated the most similar exposure and thus identified the recommended dose. Once the best recommended dose was selected, the percentages of patients achieving exposures within the recommended median AUC0–12 h 620% interval and those outside this interval (below 220% or above +20%) were estimated for 3 different body weight cutoff points: 3–6 kg, .6–14 kg, and .14–25 kg.
RESULTS Patients Forty patients (25 boys and 15 girls) were included in the study. The median weight was 6.2 kg (range: 3.5–24.4 kg) and the median age was 7 months (range: 1–84 months). Most of the patients (70%) were less than 12 months old and weighed ,10 kg. Twenty patients were treated with a dose of 0.5 mg/kg per intravenously and another 20 with a dose of 1 mg/kg. For the 48-hour duration of the study, none of the patients received drugs that induced or inhibited the enzyme CYP450.
Omeprazole Plasma Concentrations The omeprazole plasma concentrations after 0.5 and 1 mg/kg are shown in Figure 1. In this figure, concentrations are plotted versus time elapsed from the corresponding last given dose in each case (postdosing time). The final data set used for the pharmacokinetic analysis consisted of 186 omeprazole concentrations; 53 of the 186 samples (28.5%) were below the limit of quantification, mostly corresponding to trough concentration values.
Population Pharmacokinetic Model The pharmacokinetic profile of omeprazole was best described by a 2-compartment open linear model with firstorder elimination. The 3-compartment model did not improve the fit, nor the Michaelis–Menten elimination kinetics. Between-patient variability could only be associated to CL. After investigation of different relationships with body weight, all PPK parameters were allometrically scaled to body
521
Solana et al
Ther Drug Monit Volume 36, Number 4, August 2014
FIGURE 1. Omeprazole plasma concentrations (mg/L) versus time (postdosing time, hour) observed after the intravenous administration of omeprazole at doses of 1 and 0.5 mg/kg to critically ill children. Circles: observed omeprazole concentrations. Solid line: the smooth line shows the general trend of the data and the sampling points adequately defined the overall pharmacokinetic profile of omeprazole.
weight in the base model. Body weight was raised to the power of ¾ for clearance parameters (CL and CLD) and to 1 for distribution volumes (VC and VP). The allometric relationship was found to be superior to the linear relationship in the case of clearance. Evaluation of the accuracy of the 0.75 scaling factor in the allometric model was performed by estimating the allometric factor on body weight. The parameter was estimated to be about 11% higher than 0.75. Among the other covariates tested, gender and age were not found to have any statistically significant influence on CL. Although scatter plots of individual CL values from the base model suggested some kind of hyperbolic relationship with age, no statistically significant age-dependent effect on CL could be detected. The mixture model to identify subpopulations (extensive vs. poor metabolizers) was not statistically significant. The pharmacokinetic parameters estimated from the final model are listed in Table 1. Mean values with relative standard error (RSE%) were as follows: population CL 24.9 L21$h21$70 kg21 (10.08%), VC 53.9 L/70 kg (11.00%), CLD 4.23 L21$h21$70 kg21 (19.62%), and VP 674 L/70 kg (0.89%); these values corresponded to 4.04 L/h, 4.77 L, 0.69 L/h, and 59.70 L, respectively, for the typical patient in this series with a body weight of 6.2 kg. High BPV values associated with CL (75.50%; RSE, 2.60%) and residual variability (130.0%; RSE, 5.26%) were still found in the final model. All parameters were estimated with adequate precision. The estimate of h-shrinkage for CL was 12.56% while that of e-shrinkage was 16.21%. According to these values, shrinkage was not present, so that it did not lead to misinterpretation of goodnessof-fit plots based on empirical-based estimates, individual predictions, and individual weighted residuals. Figure 2 shows the observed versus population predicted and individual predicted concentration plots. Although a random distribution of data around the line exists, these plots suggest a trend to underprediction of low concentration values. Figure 3 shows CL values estimated from the final model, by body weight, plotted against different weight strata. Our results show higher CL values for smaller children than for larger children.
522
Evaluation of Predictive Ability The VPCs (Fig. 4) or prediction-corrected VPCs (Fig. 5A) show that the model simulates data in a similar manner to the observed data, except for trough concentrations corresponding to the dosing intervals of 0–12, 12–24, and 36– 48 hours, where a trend to underprediction was observed. Figure 5B shows that the proportion of observations below the limit of quantification is also adequately predicted by the model, with the exception of those corresponding to trough concentrations (at 12 hours after the last dose given) with a trend to underprediction. These results are in line with those of the density of predictive model discrepancies (NPDE). In effect, Figure 6 shows discrepancies between simulated and observed data from a sampling of n = 1000 simulations of the final model, displayed as a histogram and dispersion graph.
TABLE 1. Omeprazole Population Pharmacokinetic Parameter Estimates for the Final Model Parameter Pharmacokinetic parameters CL VC CLD VP Between-patient variability BPVCL Residual variability Proportional
Units
Final Model Parameter Estimate (RSE)*
(10.1)$(WGT/70)0.75 (11.0)$(WGT/70) (19.6)$(WGT/70)0.75 (0.89)$(WGT/70)
L/h L L/h L
24.9 53.9 4.23 674
%
76.0 (2.6)
%
130.0 (5.3)
BPV and residual variability are given as the coefficient of variation (%). *All final parameter estimates are shown with the relative standard error (RSE) indicated by italic numbers in parentheses. CL, clearance; VC and VP, volumes of distribution for central and peripheral compartments; CLD, intercompartmental clearance between central and peripheral compartments; BPV, between-patient variability; RSE, relative standard error; WGT, body weight in kilogram; 70, standard body weight in adults in kilogram.
2013 Lippincott Williams & Wilkins
Ther Drug Monit Volume 36, Number 4, August 2014
Population Pharmacokinetics of Omeprazole
Model Simulations
FIGURE 2. Goodness-of-fit plots corresponding to the final population pharmacokinetic model. Left panel: observed versus population predicted concentrations. Right panel: observed versus individual predicted concentrations. Dashed line: identity line. Solid line: smoothed line showing the general trend of the data.
The NDPE deviated from a theoretical normal distribution. Deviations were clear at the extremes, particularly for low concentration values, in agreement with results of goodnessof-fit plots. Discrepancy errors for predicted concentrations had a homogeneous distribution around NDPE = 0, but their magnitude was higher at lower concentrations. Regarding the results of the PPC (Fig. 7), the median of the simulated AUC0–12 h values was within the 5th to 95th percentile range of observed AUC0–12 h values.
FIGURE 3. NPDE obtained from 1000 simulations of the final model. Left upper panel: Q-Q plot of NPDE. Right upper panel: histogram of NPDE. Left lower panel: dispersion graphic of NPDE versus time. Right lower panel: dispersion graphic of NPDE versus predicted concentrations. Dashed lines: 95% prediction interval for a normal distribution. Solid lines: the 5th, 50th, and 95th 615 percentiles of the observed NPDE values. Shaded areas: 90% confidence intervals for the median, the 5th and 95th 616 percentiles of the NPDEs from simulated data. The 5th, 50th, and 95th 617 percentiles of the observed NPDEs fell within the corresponding 90% prediction intervals of the simulated NPDEs. 2013 Lippincott Williams & Wilkins
The results of the simulated AUC0–12 h after intravenous administration of 0.5, 1, 1.5, and 2 mg/kg of omeprazole and the ratio of the median AUC0–12 h values to the AUC0–N after an intravenous 20-mg dose in adults are detailed in Table 2. According to these results, the closest ratio to 1 was provided by the 1-mg/kg dose. We used AUC0–N (1440 mg/L$h) after the single intravenous dose of 20 mg in adults instead of AUC0–12 h because this latter value has not been reported in the literature. Nevertheless, results should not be affected by this because concentrations from 8 hours onward after the intravenous 20-mg dose in adults were below the limit of quantification (3 mg/L), meaning that no high values of extrapolated areas are to be expected, and the AUC0–N values will be close to those of AUC0–12 h. Table 3 summarizes the percentages of patients achieving exposures within the recommended median AUC0–12 h (1620 mg/L$h) 620% interval (1300–1980 mg/L$h) and those outside this interval, below 220% (,1300 mg/L$h), or above +20% (.1980 mg/L$h) estimated for 3 different body weight ranges (3–6 kg, .6–14 kg, and .14–25 kg). In all cases, around 20% of patients were within the 620% interval and .60% of patients showed AUC0–12 h values over 1300 mg/L$h.
DISCUSSION The PPIs are imidazole compounds that differ from one another in their substitution group. The PPIs reach the gastric parietal cells where they are transformed into their active form and bind selectively and irreversibly to the proton exchange ATPase (H/K ATPase) and inhibit acid secretion for up to 36 hours.30 PPIs are metabolized in the liver by cytochrome P450. The principal enzymes involved in its metabolism are CYP2C19 and CYP3A4. Genetic polymorphisms of CYP2C19 and CYP3A4, and the presence of drugs that act on cytochrome P450 directly affect the pharmacokinetics of omeprazole.30,31 Few data are available on the pharmacokinetics of intravenous omeprazole despite the fact that this is the route of choice in critically ill children.9,11,32 Jacqz-Aigrain, in a study on 13 children of different ages, observed that the pharmacokinetic parameters were similar to those of adults and that there was interindividual variability secondary to variations in liver and kidney function and to concomitant treatment.9 Our study was the first to determine the PPKs of intravenous omeprazole in critically ill children using the NONMEM software.33 PPKs not only estimate the mean values of the PPK parameters but also enable investigators to estimate the magnitude of the interindividual variability, the random or residual variability associated with the experimental concentrations, and the magnitude of variability between events associated with the pharmacokinetic parameters. In addition, the magnitude of the influence of specific covariables (demographic, pathophysiological, or therapeutic) on the value of the pharmacokinetic parameters can be identified and evaluated. Although the pharmacokinetics of omeprazole has already been described in children,8–11 no population
523
Solana et al
Ther Drug Monit Volume 36, Number 4, August 2014
FIGURE 4. Population clearance estimates by body weight (L/kg$h) in different weight strata.
pharmacokinetic models have been reported that enable optimization of the dose regimen in infants and children. Age-specific dosing information is not provided on the approved product labeling, and a population pharmacokinetic modeling approach can be a suitable tool to address this. The pharmacokinetics of omeprazole administered intravenously to children was best described by a 2-compartment model with first-order elimination. A nonlinear Michaelis–Menten elimination process did not improve the fit. The study population included those with body weights between 3.5 and 24.4 kg and ages between 1 month and 7 years; as was to be expected, the 2 covariates showed strong correlations. Investigation of
FIGURE 5. Superimposed values of the observed (filled circles) and simulated plasma concentrations (ln values of concentrations, expressed as mg/L) versus time (postdosing time, hour) profiles after the intravenous administration of 0.5 and 1 mg/kg omeprazole to critically ill children with a body weight of 6.2 kg. Mean and 95% confidence intervals obtained from 1000 simulations of omeprazole plasma concentration-time profiles. Solid line (mean predictions, 50th percentile). Dashed lines (5.0th and 95.0th percentiles).
524
the effects of weight and age on the pharmacokinetics of omeprazole in our population was one of our primary aims because these covariates will define the influence of growth and maturity. Because PPK parameters, such as clearances and distribution volumes, are usually a function of body size (with weight exponents for the power function of 0.75 and 1, respectively), the effect of this covariate may mask the effect of other covariates.34 Therefore, body weight was tested as the primary covariate on all PPK parameters followed by the effects of age on CL. Due to limited information on relevant covariates to account for variability in omeprazole pharmacokinetics, such as genetic polymorphism of CYP2C19 or the concomitant administration of inhibitors of CYP2C19 activity, we did not seek to optimize the covariate model other than through the inclusion of standard allometric models and the influence of age. Omeprazole elimination occurs primarily by metabolism and it is known that the basal metabolic rate scales with (body weight)0.75 across species in the same way as liver size. For this reason, all PPK parameters were allometrically scaled, and the scaled model adequately described the data over a 7-fold weight range (from 3.5 to 24.4 kg). However, even with the wide range of age values of the study population (from 1 month to 7 years), no age-related changes in clearance were detected. Omeprazole is primarily metabolized by cytochromes P450 2C19 and to a lesser extent by CYP3A, both polymorphically expressed. The CYP3A4 and CYP2C families appear during the first week of life, so that maturation of the metabolic pathways of omeprazole should be achieved rapidly; this could explain why CL values were not affected by age.35 However, there is marked interindividual variability both in children and in adults due in part to CYP2C19 polymorphism.36 Klotz37 reported CL values of 24–37.2 L/h, very close to the population value estimated for an adult of 70 kg in the current study. This supports both the validity of the allometric scaling model developed and also the feasibility of using the laplacian numerical estimation method for handling concentrations below of the limit of quantification (BLOQ). In that respect, it should be noted that, when the first-order conditional estimation method was applied during the first steps of the modeling process with no inclusion of BLOQs, population clearance values that deviated more from those of Klotz were obtained (15.1 L/h for an adult of 70 kg).37 The population estimates of the final PPK parameters (Fig. 2) indicated that plasma clearance was higher in smaller 2013 Lippincott Williams & Wilkins
Ther Drug Monit Volume 36, Number 4, August 2014
FIGURE 6. Prediction-corrected VPC. A, Open circles represent the observed data. The solid gray line is the median of the observed data and the dashed gray lines are the 5.0th and 95.0th percentiles of the observed data. Gray shaded areas represent the 95% confidence intervals for the simulated median and the simulated 5.0th and 95.0th percentiles. B, VPC for the fraction of data censored because they were below the limit of quantification. The solid gray line is the observed fraction of censored observations and the gray shaded area is the 95% confidence interval for the simulated fraction of censored observations. Time given in hours elapsed after the last administered dose (postdosing time).
children than in larger children or adults, falling from 0.75 L/kg$h in children with a body weight of 3.5 kg to 0.46 L/kg$h in those with a body weight of 24.0 kg and 0.36 L/kg$h in individuals with body weight of 70 kg. Therefore, the lower the weight, the higher the dose (in mg/kg) required to produce the same target exposure (AUC). This was especially relevant in children weighing ,20 kg, who required nearly twice the dose of a 70-kg adult. Consequently, the clearance in small children is faster than predicted from data derived from adults and in larger children is slower than predicted on the basis of tissue mass alone. However, similar CL values were observed when we compared our results with those of previous studies in children. Faure11 found an omeprazole CL of 0.53 L/kg$h (range: 0.37–1) in 9 noncritically ill children who received intravenous omeprazole at a dose between 0.5 and 1 mg/kg. Jacqz-Aigrain, in 13 patients found a CL of 0.23 L/kg$h, a VD of 0.45 L/kg and an elimination halflife of 0.86 hours; these values showed marked interindividual variability.9 Andersson38 and Marier39 found that pharmacokinetics of oral omeprazole are similar in children and adults. Pettersen, in critically ill children between 1 month and 5 years, found that the median values for CL, volume of distribution at steady state, and elimination half-life of pantoprazole were 0.14 L/kg$h, 0.22 L/kg, and 2.0 hours, respectively.40 Pantoprazole clearance increased with weight and age. 2013 Lippincott Williams & Wilkins
Population Pharmacokinetics of Omeprazole
FIGURE 7. Distribution of 1000 simulated AUC0–12 h (mg/L$h) values of omeprazole superimposed on the corresponding median values and the 5th and 95th percentiles of observed AUC0–12 h values (dashed lines) calculated from raw data by the noncompartmental approach. The median of simulated AUC0–12 h values (solid line) was within the 5th and 95th percentiles of the observed AUC0–12 h values (dashed lines) in all cases.
VPCs showed that our model adequately mirrored the central tendency and variability of the whole time course of plasma concentrations. The PPC plots, using the AUC as a surrogate marker of efficacy, as proposed by Junghard et al28 for esomeprazole also revealed that the model performed well enough to predict actual mean AUC0–12 h values estimated by the noncompartmental analysis at different doses. As an alternative option to optimize the daily dose of omeprazole, various dose regimens were explored via simulation based on our population PPK model. Our results
TABLE 2. Ratios of the Median Simulated AUC0–12 h in the Studied Population Versus Median AUC0–N in Adults Receiving 20 mg of Intravenous Omeprazole Dose (mg/kg) 0.5 1.0 1.5 2.0
AUC0–12
h
(mg/L$h)
Ratio
770 1620 2430 3240
The target dose would yield a simulated AUC0–12 (1440 mg/L$h) ratio of 1.
0.53 1.13 1.68 2.25 h/target
adult AUC0–N
525
Ther Drug Monit Volume 36, Number 4, August 2014
Solana et al
TABLE 3. Percentages of Patients Achieving Exposures Within the Recommended Median AUC0–12 h (1620 mg/L$h) 620% Interval (1300–1980 mg/L$h) and Those Outside this Interval, Below 220% (,1300 mg/L$h), or Above +20% (.1980 mg/ L$h) Estimated for 3 Different Body Weight Ranges (3–6 kg, .6–14 kg, and .14–25 kg) Body Weight (kg) 3–6 6–14 14–25
AUC0–12
h
, 1300
4420 3860 2780
AUC0–12 h = 1300–1980 2090 2090 2100
AUC0–12
h
. 1980
3490 4050 5110
showed an intravenous dose of 1 mg/kg produced similar exposure, expressed as the AUC value, to a 20-mg dose in adults, suggesting that this could be the recommended dose in children. Nevertheless, further pharmacokinetic studies are necessary to confirm this. Our study has certain limitations including the fact that only a limited number of covariates were recorded and could be tested. This could account for the high BPV of plasma clearance that persisted in the final model. The reasons for this may be unrecorded factors such as genetic polymorphism of CYP2C19 or the possible co-administration of inhibitors of omeprazole metabolism.40 On the other hand, the assay was not sensitive enough (even at 1 ng/mL), which resulted in nearly 30% of the samples giving a result BLOQ. This is the likely reason for the apparent underprediction of trough concentrations, which is related to the use of the Beal M3 method that distributes the BLOQ concentrations from 2infinity to the limit of quantification and negative concentrations are possible in this method. This fact suggests that the model is not under predicting at the trough times because many of the observed trough concentrations that are BLOQ cannot be plotted. Another limitation is the relatively small number of samples obtained beyond 12 hours after the start of the treatment. In summary, the pharmacokinetics of intravenous omeprazole in our study were described using a 2-compartment model with an intravenous administration according to a zeroorder kinetics and linear elimination. Our population pharmacokinetic model demonstrates the relevancy of allometric power models to quantify the effect of size on disposition parameters, enables the changes in CL and V during development to be explained, and allows dose regimens to be established that result in drug exposure equivalent to that observed in adults, for whom omeprazole efficacy has been established. Furthermore, this is the first step toward pharmacokinetic–pharmacodynamic modeling to optimize the dose regimen in this population. REFERENCES 1. Lin PC, Chang CH, Hsu PI, et al The efficacy and safety of proton pump inhibitors vs histamine-2 receptor antagonists for stress ulcer bleeding prophylaxis among critical care patients: a meta-analysis. Crit Care Med. 2010;38:1197–1205. 2. Stollman N, Metz DC. Pathophysiology and prophylaxis of stress ulcer in intensive care unit patients. J Crit Care. 2005;20:35–45. 3. Reveiz L, Guerrero-Lozano R, Camacho A, et al Stress ulcer, gastritis, and gastrointestinal bleeding prophylaxis in critically ill pediatric patients: a systematic review. Pediatr Crit Care Med. 2010;11:124–132.
526
4. Cook DJ, Griffith LE, Walter SD, et al; Canadian Critical Care Trials Group. The attributable mortality and length of intensive care unit stay of clinically important gastrointestinal bleeding in critically ill patients. Crit Care. 2001;5:368–375. 5. Steinberg KP. Stress-related mucosal disease in the critically ill patient: risk factors and strategies to prevent stress-related bleeding in the intensive care unit. Crit Care Med. 2002;30(suppl):S362–S364. 6. Metz DC. Preventing the gastrointestinal consequences of stress-related mucosal disease. Curr Med Res Opin. 2005;21:11–18. 7. Leontiadis GI, Sreedharan A, Dorward S, et al. Systematic reviews of the clinical effectiveness and cost-effectiveness of proton pump inhibitors in acute upper gastrointestinal bleeding. Health Technol Assess. 2007;11: iii–iv, 1–164. 8. Solana MJ, López-Herce J. Pharmacokinetics of intravenous omeprazole in critically ill paediatric patients. Eur J Clin Pharmacol. 2010;66:323–330. 9. Jacqz-Aigrain E, Bellaich M, Faure C, et al. Pharmacokinetics of intravenous omeprazole in children. Eur J Clin Pharmacol. 1994;47:181–185. 10. Litalien C, Théorêt Y, Faure C. Pharmacokinetics of proton pump inhibitors in children. Clin Pharmacokinet. 2005;44:441–466. 11. Faure C, Michaud L, Shaghaghi EK, et al. Intravenous omeprazole in children: pharmacokinetics and effect on 24-hour intragastric pH. J Pediatr Gastroenterol Nutr. 2001;33:144–148. 12. Faure C, Michaud L, Shaghaghi EK, et al. Lansoprazole in children: pharmacokinetics and efficacy in reflux oesophagitis. Aliment Pharmacol Ther. 2001;15:1397–1402. 13. Phillips JO, Olsen KM, Rebuck JA, et al. A randomized, pharmacokinetic and pharmacodynamic, cross-over study of duodenal or jejunal administration compared to nasogastric administration of omeprazole suspension in patients at risk for stress ulcers. Am J Gastroenterol. 2001;96:367–372. 14. Meier-Hellmann A, Reinhart K, Bredle DL, et al. Epinephrine impairs splanchnic perfusion in septic shock. Crit Care Med. 1997;25:399–404. 15. Boucher BA, Wood GC, Swanson JM. Pharmacokinetic changes in critical illness. Crit Care Clin. 2006;22:255–271. 16. McKindley DS, Hanes SD, Boucher BA. Hepatic drug metabolism in critical illness. Pharmacotheraphy. 1998;18:759–778. 17. Jonsson EN, Karlsson MO. Xpose–an S-PLUS based population pharmacokinetic/pharmacodynamic model building aid for NONMEM. Comput Methods Programs Biomed. 1999;58:51–64. 18. Lindbom L, Pihlgren P, Jonsson EN. PsN-Toolkit–a collection of computer intensive statistical methods for non-linear mixed effect modeling using NONMEM. Comput Methods Programs Biomed. 2005;79:241–257. 19. Yamaoka T, Nakagawa T, Uno T. Application of Akaike’s information criterion (AIC) in the evaluation of linear pharmacokinetics equations. J Pharmacokinet Biopharm. 1978;6:165–175. 20. Beal SL. Ways to fit a PK model with some data below the quantification limit. J Pharmacokinet Pharmacodyn. 2001;28:481–504. 21. Meibohm B, Läer S, Panetta JC, et al. Population pharmacokinetic studies in pediatrics: issues in design and analysis. AAPS J. 2005;7:E475–E487. 22. Anderson BJ, Holford NH. Mechanism-based concepts of size and maturity in pharmacokinetics. Annu Rev Pharmacol Toxicol. 2008;48:303–332. 23. Anderson BJ, Holford NH. Mechanistic basis of using body size and maturation to predict clearance in humans. Drug Metab Pharmacokinet. 2009;24:25–36. 24. Savic RM, Karlsson MO. Importance of shrinkage in empirical bayes estimates for diagnostics: problems and solutions. AAPS J. 2009;11:558–569. 25. Yano Y, Beal S, Sheiner LB. Pharmacokinetic-pharmacodynamic models were evaluated using the posterior predictive check. J Pharmacokinet Biopharm. 2001;28:171–192. 26. Mentré F, Escolano Sylvie. Prediction discrepancies for the evaluation of non-linear mixed effects models. J Pharmacokinet Pharmacodyn. 2006; 33:345–367. 27. Bergstrand M, Karlsson MO. Handling data below the limit of quantification in mixed effect models. AAPS J. 2009;11:371–380. 28. Junghard O, Hassan-Alin M, Hasselgren G. The effect of the area under the plasma concentration vs time curve and the maximum plasma concentration of esomeprazole on intragastric pH. Eur J Clin Pharmacol. 2002;58:453–458. 29. Uno T, Niioka T, Hayakari M, et al. Absolute bioavailability and metabolism of omeprazole in relation to CYP2C19 genotypes following single intravenous and oral administrations. Eur J Clin Pharmacol. 2007;63: 143–149.
2013 Lippincott Williams & Wilkins
Ther Drug Monit Volume 36, Number 4, August 2014
30. Shi S, Klotz U. Proton pump inhibitors: an update of their clinical use and pharmacokinetics. Eur J Clin Pharmacol. 2008;64:935–951. 31. Vanderhoff BT, Tahboub RM. Proton pump inhibitors, an update. Clin Pharmacol. 2002;66:273–280. 32. Kaufman SS, Lyden ER, Brown CR, et al. Omeprazole therapy in pediatric patients after liver and intestinal transplantation. J Pediatr Gastroenterol Nutr. 2002;34:194–198. 33. Sakurai Y, Hirayama M, Hashimoto M, et al. Population pharmacokinetics and proton pump inhibitory effects of intravenous lansoprazole in healthy Japanese males. Biol Pharm Bull. 2007;30:2238–2243. 34. Peters HP. Physiological Correlates of Size. Beck E, Birks HJB, Conner EF, eds. Cambridge, MA: Cambridge university press; 1983:48–53. Chapter 4. 35. Kearns GL, Abdel-Rahman SM, Alander SW, et al. Developmental pharmacology-drug disposition action and therapy in infants and children N Engl J Med. 2003;349:1157–1167.
2013 Lippincott Williams & Wilkins
Population Pharmacokinetics of Omeprazole
36. Zimmermann AE, Walters K, Katona BG, et al. A review of omeprazol use in the treatment of acid-related disorders in children. Clin Ther. 2001; 23:660–679. 37. Klotz U. Pharmacokinetic considerations in the eradication of Helicobacter pylori. Clin Pharmacokinet. 2000;38:243–270. 38. Andersson T, Hassall E, Lundborg P, et al. Pharmacokinetics of orally administered omeprazole in children. International Pediatric Omeprazole Pharmacokinetic Group. Am J Gastroenterol. 2000;95: 3101–3106. 39. Marier JF, Dubuc MC, Drouin E, et al. Pharmacokinetics of omeprazole in healthy adults and in children with gastroesophageal reflux disease. Ther Drug Monit. 2004;26:3–8. 40. Pettersen G, Mouksassi M-S, Théorêt Y, et al. Population pharmacokinetics of intravenous pantoprazole in paediatric intensive care patients. Br J Clin Pharmacol. 2009;67:216–227.
527
Population Pharmacokinetics of Omeprazole in Critically Ill Pediatric Patients Maria Jose Solana, MD,* Helena Colom, PhD,† Jesús López-Herce, PhD, MD,* Javier Urbano, MD,* Rafael González, MD,* Jorge López, MD,* Cecilia Manzanares, PhD,‡ and Angel Carrillo, PhD, MD*
Background: To develop a population pharmacokinetic model for intravenous omeprazole in critically ill children.
Methods: One hundred eighty-six omeprazole concentration-time data from 40 critically ill children were analyzed using the nonlinear mixed-effects approach with the nonlinear mixed-effects modeling software, version 7.2 software. Patients were randomized into 2 groups and received intravenous omeprazole at a dose of 0.5 or 1 mg/kg twice daily. Blood samples were drawn at 0.5, 2, 6, 12, 24, and 48 hours after the first infusion. Results: The pharmacokinetic profile was best described by a 2-compartment model with a first-order elimination process. Betweenpatient variability could only be associated with plasma clearance (CL). The typical values for plasma CL were 24.9 L$h21$70 kg21 (10.08%), with a distributional clearance of 53.9 L$h21$70 kg21 (11.00%) and central and peripheral compartment distribution volumes of 4.23 L/70 kg (19.62%) and 674 L/70 kg (0.89%), respectively. Allometric size models seemed to predict changes adequately in all the pharmacokinetic parameters. High values of between-patient variability of CL [75.50% (2.60%)] and residual variability [130.0% (5.26%)] were still found in the final model. Model-based simulations suggested that the most suitable dose was 1 mg/kg because this yielded similar exposure (defined by the area under the concentration-time curve) to that obtained in adults after a 20-mg dose of omeprazole intravenously.
Conclusions: An allometric size model allows changes to be predicted in all the pharmacokinetic parameters, making dose adjustment by body weight important to achieve the most effective omeprazole exposure. This is the first step toward a population pharmacokinetic study, including more data to develop a predictable model to be used during therapeutic drug monitoring.
Received for publication March 10, 2013; accepted November 11, 2013. From the *Paediatric Intensive Care Department, Hospital General Universitario Gregorio Marañón, Madrid, Spain; †Biopharmacy and Pharmacokinetics Department, School of Pharmacy, University of Barcelona, Spain; and ‡Pharmacy Service, Hospital General Universitario Gregorio Marañón, Madrid, Spain. Supported by a grant from the Spanish Health Institute Carlos III (grants N. EC07/90670 and RD08/0072: Maternal, Child Health and Development Network) within the framework of the VI National I + D + i Research Programme (2008–2011). The authors declare no conflicts of interest. Correspondence: Jesús López-Herce, PhD, MD, Servicio de Cuidados Intensivos Pediátricos, Hospital General Universitario Gregorio Marañón, Dr Castelo, 47 28009 Madrid, Spain (e-mail: [email protected]). Copyright © 2013 by Lippincott Williams & Wilkins
Ther Drug Monit Volume 36, Number 4, August 2014
Key Words: omeprazole, population pharmacokinetics, nonlinear mixed-effects models, critically ill children, therapeutic drug monitoring (Ther Drug Monit 2014;36:519–527)
INTRODUCTION Gastrointestinal bleeding in critically ill patients is usually secondary to acute lesions of the gastric mucosa and is associated with increased morbidity and mortality.1–4 Acid gastric pH is one of the factors that influence the development of acute lesions of the gastric mucosa.5,6 The proton pump inhibitors (PPIs) are drugs of choice for the prophylaxis and treatment of digestive tract hemorrhage in critically ill patients.7 There are 5 types of PPI; their pharmacokinetics vary because of differences in their molecular structure.7 Omeprazole was the first PPI and is still the most widely used. There are very few studies that have analyzed the pharmacokinetics of PPI in children8–12 and in critically ill patients.13 Critically ill patients have certain pharmacokinetic characteristics that affect the absorption, metabolism, bioavailability, and excretion of drugs. These pharmacokinetic alterations are the result of organ dysfunction (particularly of the liver and kidney), the release of acute-phase reactants, therapeutic interventions and interactions between different drugs. Hypovolaemic states, heart failure, and the alphaagonists decrease hepatic blood flow, producing a fall in drug clearance.14 In contrast, vasodilators favor clearance.15 Inflammatory cytokines and stress hormones can inhibit cytochrome P450, altering the metabolism of omeprazole. Conjugation is also affected, though to a lesser extent.15,16 In addition, drug interactions with drugs frequently prescribed to critically ill patients can also affect PPI metabolism. The objective of this study was to investigate the pharmacokinetics of intravenous omeprazole in critically ill children and to analyze the influence of demographic factors (age, weight) and of different doses on the pharmacokinetic behavior of the drug.
MATERIAL AND METHODS A prospective, randomized, open-label clinical trial (Eudra-CT no: OM1//2007-006102-19) was performed after approval of the protocol by the hospital ethics committee. Patients admitted to the pediatric intensive care unit whose
519
Solana et al
Ther Drug Monit Volume 36, Number 4, August 2014
parents signed the informed consent form were included in the study. Exclusion criteria included treatment with H2receptor antagonists or PPIs at the time of admission, any contraindication for the insertion of a nasogastric tube and gastric hemorrhage before study entry. Patients were randomly assigned to 1 of the 2 therapeutic groups, each of 20 patients: group A (omeprazole 0.5 mg/kg per intravenously) and a group B (omeprazole 1 mg/kg per intravenously). The omeprazole infusion was prepared by diluting a 40-mg vial of omeprazole in 100 mL of normal saline and administered by slow intravenous infusion over 20 minutes via a continuous infusion pump every 12 hours, with a maximum of 40 mg per dose.
mixed-effects modeling (NONMEM) software, version 7.2 (ICON Development Solutions, Ellicott City, MD) (Bauer Maryland 2011). Graphical diagnostics were assessed using Xpose version 4.2.1 (http://xpose.sourceforge.net) implemented into R version 2.11.1 (http://www.r-project.com)17 and Perl speaks-NONMEM (PsN) version 3.5.3 Tool-kit (http://psn.sourceforge.net).18 Noncompartmental analysis was performed using WinNonlin version 5.3. Pooled concentration-time data were simultaneously analyzed after log transformation. The modeling process consisted of 3 steps: (1) base pharmacokinetic model development, (2) covariate selection, and (3) model evaluation.
Measurement of the Plasma Concentration of Omeprazole
One, 2, and 3-compartment models with first-order and nonlinear elimination processes from the central compartment were tested to characterize the changes in omeprazole plasma concentrations over time. The models were parameterized in terms of clearances (CL) and apparent volumes of distribution (V) and VM (maximal elimination rate) and KM (concentration of the drug at which the elimination is half maximal) in the case of Michaelis–Menten kinetics. A mixture model to identify subpopulations between extensive and poor metabolizers was also tested. Between-patient variability (BPV) was evaluated for each PPK parameter and was modeled exponentially, assuming a lognormal distribution. Residual variability was modeled using an additive error model on a logarithmic scale. The difference in the minimum objective function value (MOFV) was used to distinguish between nested models because this difference presents an approximate x2 distribution. Significance was taken as a P value ,0.005, which corresponded to a difference in the MOFV of 7.879 for 1 degree of freedom. For nonhierarchical models, the most parsimonious model with the lowest objective function according to the Akaike information criterion was chosen.19 Some data were reported as concentrations below the limit of quantification. Below the limit of quantification values were included during the analyses and treated as censored information using the M3 method.20 The Laplacian numerical estimation method was used for parameter estimation.
The plasma concentrations of omeprazole were measured at 30 minutes, 2 hours, and 6 hours after the first dose and before the doses at 12, 24, and 48 hours. The samples were stored in lithium heparin tubes until processing. After centrifugation at 3100 rpm for 10 minutes at 48C, 2 aliquots of 0.75 mL of plasma were extracted and stored in round-bottomed plastic tubes at 2808C until their analysis using rapid resolution liquid chromatography tandem mass spectrometry.
Assay of Omeprazol The assay was performed according to the Food and Drug Administration Guidance for Industry Bioanalytical Method Validation, Biopharmaceutics Coording Committee in the Center for Drug Evaluation and Research in cooperation with the Center for Veterinary Medicine at the Food and Drug Administration (May 2001) and previous publications.17 To each 0.2-mL volume of plasma sample, 50 mL of the internal standard (lansoprazol purchased from Sigma Aldrich, purity .99%, respectively) (1 mg/mL) was added and mixed. Working solutions were obtained by dilution of stock solutions using calibrated pipettes and volumetric flasks. Then, 2 mL of ethyl acetate was added. Extraction was performed by shaking for 30 seconds and then by centrifugation at 3000 rpm for 10 minutes. The organic layer was separated and evaporate to dryness under a steady stream of nitrogen N2 (458C) and the extract was reconstituted with 100 mL of water/methanol (1:9) (v/v). Chromatographic analyses were performed using a highperformance liquid chromatography binary pump (Agilent 1200 Chemstation, Agilent Technology Co, Ltd) and a massspectrometer (API 4000, Applied Byosystems). After sample treatment, omeprazol in human plasma and the internal standard lansoprazol was chromatographed on a Luna C18 100A (4.6 mm · 50 mm, 5 microm, Phenomenex) column with a mobile phase of 10 mM ammonium acetate, pH 4.0, and acetonitrile. Detection was performed on a triple quadrupole tandem mass spectrometry by multiple reaction monitoring in positive mode. The limit of detection was 1 mg/L and the assay exhibited a linear range between 1 and 500 mg/L.
Population Pharmacokinetic Analysis The population pharmacokinetic (PPK) model development and simulations were performed with the nonlinear
520
Base Population Pharmacokinetic Model
Covariate Model Allometric weight scaling was added to all clearance and volume fixed effects a priori and standardized to a body weight of 70 kg21,22 according to the following relationships: Pj = Pstd$(WGT/70)PER, where Pj is the parameter of the jth individual and Pstd is the typical value of the PPK parameter for an adult weighing 70 kg. As previously reported,23 the power parameter was 0.75 for clearances and 1 for distribution volumes.24 The effects of age and gender were then tested on the PPK parameters to identify possible differences not explained by the allometric models. The effect of age was investigated using a sigmoid hyperbolic or Hill model as follows: Maturation process (MF) = AGEHill/TM50Hill + AGEHill, where TM50 describes the maturation half time, while the Hill coefficient relates the slope of this maturation profile.23 Categorical covariates such as gender were tested in their respective parameters as indicated by the relationship: TVPj = u1 for Z = 0 and TVPj = u1. u2 for Z = 1, where 2013 Lippincott Williams & Wilkins
Ther Drug Monit Volume 36, Number 4, August 2014
Z values represent each level of the categorical covariate. Specifically, in the case of gender, TVPj is the typical value of the jth pharmacokinetic parameter for males and u2 is the fractional change in u1 by females. Multiplicative equations were used to describe the combined effect of multiple covariates on the same parameter. Covariates were first tested univariately in the model and then by cumulative forward inclusion/backward elimination procedures. Significance levels of 5% (reduction .3.841 units in the MOFV) and 0.1% (increase .10.8 units in the MOFV) were employed during the forward inclusion and backward elimination steps. The decrease in the MOFV (22 · log likelihood), parameter precision expressed as the relative standard error (RSE%), reductions in BPV associated with a specific pharmacokinetic parameter, model completion status (eg, successful convergence or termination), and visual inspection of goodness-of-fit plots with Xpose were also taken into account for model selection. A decrease of at least 10% in interpatient variability associated with a specific pharmacokinetic parameter was considered clinically relevant for the inclusion of that specific covariate.
Evaluation of Predictive Ability The extent of Bayesian shrinkage was also evaluated for each parameter in the final population pharmacokinetic model.24 The ability of the parameter estimates to describe the observed data was further evaluated by exploring the results from a visual predictive check (VPC), posterior predictive check (PPC),25 and normalized prediction distribution errors (NPDE).26 For the VPC, 1000 individual profiles were simulated and the 5th, 50th, and 95th percentiles were calculated for each sampling time point and represented with the raw data. If the model adequately described the data, the lines corresponding to the median and to the 5th and 95th percentiles of the observed data would mainly fall in the area covering the 90% confidence interval. A prediction-corrected VPC was also performed.27 By using prediction-corrected VPC, both the observations and the model predictions were normalized for the typical model predictions in each bin of independent variables. The 95% confidence intervals for the median, the 5th and 95th percentiles of the predicted data were then calculated and plotted together with the median and the 5th and 95th percentiles of the observed data. If the model described the data adequately, the lines corresponding to the median and to the 5th and 95th percentiles of the observed data would fall in the respective 95% confidence intervals of the predicted data. For the PPC, we investigated how well the model predicted the values of the area under the concentration-time curve (AUC) of omeprazole from 0 to 12 hours by simulating 1000 data sets as the original data set. The AUC0–12 h values were then calculated and compared with those calculated from the observed data by noncompartmental analysis using the WinNonlin package. The AUC was selected for evaluation because this parameter was the most relevant surrogate marker of efficacy of omeprazole28 and was expected to vary with concentration and with relevant covariates. For the NPDE, 1000 individual profiles, as for the original data set, 2013 Lippincott Williams & Wilkins
Population Pharmacokinetics of Omeprazole
were simulated from the final model. Then, the differences between each observation and simulation were calculated and then normalized by using the inverse of the cumulative density function. If the model fitted the data adequately, NPDEs should result in a normal distribution with a mean of 0 and variance of 2.
Model-Based Simulations
Once the final model was achieved, the simulated values of the AUC0–12 h after doses of 0.5, 1, 1.5, and 2 mg/kg were investigated. A new data set containing 40 patients with the same demographics as the patients of the target population was created. Using the final model, 1000 simulations were run in which subjects received doses of 0.5, 1, 1.5, and 2 mg/kg. The resulting simulated concentration-time data for each dose were used to calculate median AUC0–12 h values. Pediatric median AUC0–12 h values after the 4 doses were divided by the previously reported AUC0–N values after the intravenous administration of 20 mg of omeprazole to adults29 to give a ratio. Values closest to 1 indicated the most similar exposure and thus identified the recommended dose. Once the best recommended dose was selected, the percentages of patients achieving exposures within the recommended median AUC0–12 h 620% interval and those outside this interval (below 220% or above +20%) were estimated for 3 different body weight cutoff points: 3–6 kg, .6–14 kg, and .14–25 kg.
RESULTS Patients Forty patients (25 boys and 15 girls) were included in the study. The median weight was 6.2 kg (range: 3.5–24.4 kg) and the median age was 7 months (range: 1–84 months). Most of the patients (70%) were less than 12 months old and weighed ,10 kg. Twenty patients were treated with a dose of 0.5 mg/kg per intravenously and another 20 with a dose of 1 mg/kg. For the 48-hour duration of the study, none of the patients received drugs that induced or inhibited the enzyme CYP450.
Omeprazole Plasma Concentrations The omeprazole plasma concentrations after 0.5 and 1 mg/kg are shown in Figure 1. In this figure, concentrations are plotted versus time elapsed from the corresponding last given dose in each case (postdosing time). The final data set used for the pharmacokinetic analysis consisted of 186 omeprazole concentrations; 53 of the 186 samples (28.5%) were below the limit of quantification, mostly corresponding to trough concentration values.
Population Pharmacokinetic Model The pharmacokinetic profile of omeprazole was best described by a 2-compartment open linear model with firstorder elimination. The 3-compartment model did not improve the fit, nor the Michaelis–Menten elimination kinetics. Between-patient variability could only be associated to CL. After investigation of different relationships with body weight, all PPK parameters were allometrically scaled to body
521
Solana et al
Ther Drug Monit Volume 36, Number 4, August 2014
FIGURE 1. Omeprazole plasma concentrations (mg/L) versus time (postdosing time, hour) observed after the intravenous administration of omeprazole at doses of 1 and 0.5 mg/kg to critically ill children. Circles: observed omeprazole concentrations. Solid line: the smooth line shows the general trend of the data and the sampling points adequately defined the overall pharmacokinetic profile of omeprazole.
weight in the base model. Body weight was raised to the power of ¾ for clearance parameters (CL and CLD) and to 1 for distribution volumes (VC and VP). The allometric relationship was found to be superior to the linear relationship in the case of clearance. Evaluation of the accuracy of the 0.75 scaling factor in the allometric model was performed by estimating the allometric factor on body weight. The parameter was estimated to be about 11% higher than 0.75. Among the other covariates tested, gender and age were not found to have any statistically significant influence on CL. Although scatter plots of individual CL values from the base model suggested some kind of hyperbolic relationship with age, no statistically significant age-dependent effect on CL could be detected. The mixture model to identify subpopulations (extensive vs. poor metabolizers) was not statistically significant. The pharmacokinetic parameters estimated from the final model are listed in Table 1. Mean values with relative standard error (RSE%) were as follows: population CL 24.9 L21$h21$70 kg21 (10.08%), VC 53.9 L/70 kg (11.00%), CLD 4.23 L21$h21$70 kg21 (19.62%), and VP 674 L/70 kg (0.89%); these values corresponded to 4.04 L/h, 4.77 L, 0.69 L/h, and 59.70 L, respectively, for the typical patient in this series with a body weight of 6.2 kg. High BPV values associated with CL (75.50%; RSE, 2.60%) and residual variability (130.0%; RSE, 5.26%) were still found in the final model. All parameters were estimated with adequate precision. The estimate of h-shrinkage for CL was 12.56% while that of e-shrinkage was 16.21%. According to these values, shrinkage was not present, so that it did not lead to misinterpretation of goodnessof-fit plots based on empirical-based estimates, individual predictions, and individual weighted residuals. Figure 2 shows the observed versus population predicted and individual predicted concentration plots. Although a random distribution of data around the line exists, these plots suggest a trend to underprediction of low concentration values. Figure 3 shows CL values estimated from the final model, by body weight, plotted against different weight strata. Our results show higher CL values for smaller children than for larger children.
522
Evaluation of Predictive Ability The VPCs (Fig. 4) or prediction-corrected VPCs (Fig. 5A) show that the model simulates data in a similar manner to the observed data, except for trough concentrations corresponding to the dosing intervals of 0–12, 12–24, and 36– 48 hours, where a trend to underprediction was observed. Figure 5B shows that the proportion of observations below the limit of quantification is also adequately predicted by the model, with the exception of those corresponding to trough concentrations (at 12 hours after the last dose given) with a trend to underprediction. These results are in line with those of the density of predictive model discrepancies (NPDE). In effect, Figure 6 shows discrepancies between simulated and observed data from a sampling of n = 1000 simulations of the final model, displayed as a histogram and dispersion graph.
TABLE 1. Omeprazole Population Pharmacokinetic Parameter Estimates for the Final Model Parameter Pharmacokinetic parameters CL VC CLD VP Between-patient variability BPVCL Residual variability Proportional
Units
Final Model Parameter Estimate (RSE)*
(10.1)$(WGT/70)0.75 (11.0)$(WGT/70) (19.6)$(WGT/70)0.75 (0.89)$(WGT/70)
L/h L L/h L
24.9 53.9 4.23 674
%
76.0 (2.6)
%
130.0 (5.3)
BPV and residual variability are given as the coefficient of variation (%). *All final parameter estimates are shown with the relative standard error (RSE) indicated by italic numbers in parentheses. CL, clearance; VC and VP, volumes of distribution for central and peripheral compartments; CLD, intercompartmental clearance between central and peripheral compartments; BPV, between-patient variability; RSE, relative standard error; WGT, body weight in kilogram; 70, standard body weight in adults in kilogram.
2013 Lippincott Williams & Wilkins
Ther Drug Monit Volume 36, Number 4, August 2014
Population Pharmacokinetics of Omeprazole
Model Simulations
FIGURE 2. Goodness-of-fit plots corresponding to the final population pharmacokinetic model. Left panel: observed versus population predicted concentrations. Right panel: observed versus individual predicted concentrations. Dashed line: identity line. Solid line: smoothed line showing the general trend of the data.
The NDPE deviated from a theoretical normal distribution. Deviations were clear at the extremes, particularly for low concentration values, in agreement with results of goodnessof-fit plots. Discrepancy errors for predicted concentrations had a homogeneous distribution around NDPE = 0, but their magnitude was higher at lower concentrations. Regarding the results of the PPC (Fig. 7), the median of the simulated AUC0–12 h values was within the 5th to 95th percentile range of observed AUC0–12 h values.
FIGURE 3. NPDE obtained from 1000 simulations of the final model. Left upper panel: Q-Q plot of NPDE. Right upper panel: histogram of NPDE. Left lower panel: dispersion graphic of NPDE versus time. Right lower panel: dispersion graphic of NPDE versus predicted concentrations. Dashed lines: 95% prediction interval for a normal distribution. Solid lines: the 5th, 50th, and 95th 615 percentiles of the observed NPDE values. Shaded areas: 90% confidence intervals for the median, the 5th and 95th 616 percentiles of the NPDEs from simulated data. The 5th, 50th, and 95th 617 percentiles of the observed NPDEs fell within the corresponding 90% prediction intervals of the simulated NPDEs. 2013 Lippincott Williams & Wilkins
The results of the simulated AUC0–12 h after intravenous administration of 0.5, 1, 1.5, and 2 mg/kg of omeprazole and the ratio of the median AUC0–12 h values to the AUC0–N after an intravenous 20-mg dose in adults are detailed in Table 2. According to these results, the closest ratio to 1 was provided by the 1-mg/kg dose. We used AUC0–N (1440 mg/L$h) after the single intravenous dose of 20 mg in adults instead of AUC0–12 h because this latter value has not been reported in the literature. Nevertheless, results should not be affected by this because concentrations from 8 hours onward after the intravenous 20-mg dose in adults were below the limit of quantification (3 mg/L), meaning that no high values of extrapolated areas are to be expected, and the AUC0–N values will be close to those of AUC0–12 h. Table 3 summarizes the percentages of patients achieving exposures within the recommended median AUC0–12 h (1620 mg/L$h) 620% interval (1300–1980 mg/L$h) and those outside this interval, below 220% (,1300 mg/L$h), or above +20% (.1980 mg/L$h) estimated for 3 different body weight ranges (3–6 kg, .6–14 kg, and .14–25 kg). In all cases, around 20% of patients were within the 620% interval and .60% of patients showed AUC0–12 h values over 1300 mg/L$h.
DISCUSSION The PPIs are imidazole compounds that differ from one another in their substitution group. The PPIs reach the gastric parietal cells where they are transformed into their active form and bind selectively and irreversibly to the proton exchange ATPase (H/K ATPase) and inhibit acid secretion for up to 36 hours.30 PPIs are metabolized in the liver by cytochrome P450. The principal enzymes involved in its metabolism are CYP2C19 and CYP3A4. Genetic polymorphisms of CYP2C19 and CYP3A4, and the presence of drugs that act on cytochrome P450 directly affect the pharmacokinetics of omeprazole.30,31 Few data are available on the pharmacokinetics of intravenous omeprazole despite the fact that this is the route of choice in critically ill children.9,11,32 Jacqz-Aigrain, in a study on 13 children of different ages, observed that the pharmacokinetic parameters were similar to those of adults and that there was interindividual variability secondary to variations in liver and kidney function and to concomitant treatment.9 Our study was the first to determine the PPKs of intravenous omeprazole in critically ill children using the NONMEM software.33 PPKs not only estimate the mean values of the PPK parameters but also enable investigators to estimate the magnitude of the interindividual variability, the random or residual variability associated with the experimental concentrations, and the magnitude of variability between events associated with the pharmacokinetic parameters. In addition, the magnitude of the influence of specific covariables (demographic, pathophysiological, or therapeutic) on the value of the pharmacokinetic parameters can be identified and evaluated. Although the pharmacokinetics of omeprazole has already been described in children,8–11 no population
523
Solana et al
Ther Drug Monit Volume 36, Number 4, August 2014
FIGURE 4. Population clearance estimates by body weight (L/kg$h) in different weight strata.
pharmacokinetic models have been reported that enable optimization of the dose regimen in infants and children. Age-specific dosing information is not provided on the approved product labeling, and a population pharmacokinetic modeling approach can be a suitable tool to address this. The pharmacokinetics of omeprazole administered intravenously to children was best described by a 2-compartment model with first-order elimination. A nonlinear Michaelis–Menten elimination process did not improve the fit. The study population included those with body weights between 3.5 and 24.4 kg and ages between 1 month and 7 years; as was to be expected, the 2 covariates showed strong correlations. Investigation of
FIGURE 5. Superimposed values of the observed (filled circles) and simulated plasma concentrations (ln values of concentrations, expressed as mg/L) versus time (postdosing time, hour) profiles after the intravenous administration of 0.5 and 1 mg/kg omeprazole to critically ill children with a body weight of 6.2 kg. Mean and 95% confidence intervals obtained from 1000 simulations of omeprazole plasma concentration-time profiles. Solid line (mean predictions, 50th percentile). Dashed lines (5.0th and 95.0th percentiles).
524
the effects of weight and age on the pharmacokinetics of omeprazole in our population was one of our primary aims because these covariates will define the influence of growth and maturity. Because PPK parameters, such as clearances and distribution volumes, are usually a function of body size (with weight exponents for the power function of 0.75 and 1, respectively), the effect of this covariate may mask the effect of other covariates.34 Therefore, body weight was tested as the primary covariate on all PPK parameters followed by the effects of age on CL. Due to limited information on relevant covariates to account for variability in omeprazole pharmacokinetics, such as genetic polymorphism of CYP2C19 or the concomitant administration of inhibitors of CYP2C19 activity, we did not seek to optimize the covariate model other than through the inclusion of standard allometric models and the influence of age. Omeprazole elimination occurs primarily by metabolism and it is known that the basal metabolic rate scales with (body weight)0.75 across species in the same way as liver size. For this reason, all PPK parameters were allometrically scaled, and the scaled model adequately described the data over a 7-fold weight range (from 3.5 to 24.4 kg). However, even with the wide range of age values of the study population (from 1 month to 7 years), no age-related changes in clearance were detected. Omeprazole is primarily metabolized by cytochromes P450 2C19 and to a lesser extent by CYP3A, both polymorphically expressed. The CYP3A4 and CYP2C families appear during the first week of life, so that maturation of the metabolic pathways of omeprazole should be achieved rapidly; this could explain why CL values were not affected by age.35 However, there is marked interindividual variability both in children and in adults due in part to CYP2C19 polymorphism.36 Klotz37 reported CL values of 24–37.2 L/h, very close to the population value estimated for an adult of 70 kg in the current study. This supports both the validity of the allometric scaling model developed and also the feasibility of using the laplacian numerical estimation method for handling concentrations below of the limit of quantification (BLOQ). In that respect, it should be noted that, when the first-order conditional estimation method was applied during the first steps of the modeling process with no inclusion of BLOQs, population clearance values that deviated more from those of Klotz were obtained (15.1 L/h for an adult of 70 kg).37 The population estimates of the final PPK parameters (Fig. 2) indicated that plasma clearance was higher in smaller 2013 Lippincott Williams & Wilkins
Ther Drug Monit Volume 36, Number 4, August 2014
FIGURE 6. Prediction-corrected VPC. A, Open circles represent the observed data. The solid gray line is the median of the observed data and the dashed gray lines are the 5.0th and 95.0th percentiles of the observed data. Gray shaded areas represent the 95% confidence intervals for the simulated median and the simulated 5.0th and 95.0th percentiles. B, VPC for the fraction of data censored because they were below the limit of quantification. The solid gray line is the observed fraction of censored observations and the gray shaded area is the 95% confidence interval for the simulated fraction of censored observations. Time given in hours elapsed after the last administered dose (postdosing time).
children than in larger children or adults, falling from 0.75 L/kg$h in children with a body weight of 3.5 kg to 0.46 L/kg$h in those with a body weight of 24.0 kg and 0.36 L/kg$h in individuals with body weight of 70 kg. Therefore, the lower the weight, the higher the dose (in mg/kg) required to produce the same target exposure (AUC). This was especially relevant in children weighing ,20 kg, who required nearly twice the dose of a 70-kg adult. Consequently, the clearance in small children is faster than predicted from data derived from adults and in larger children is slower than predicted on the basis of tissue mass alone. However, similar CL values were observed when we compared our results with those of previous studies in children. Faure11 found an omeprazole CL of 0.53 L/kg$h (range: 0.37–1) in 9 noncritically ill children who received intravenous omeprazole at a dose between 0.5 and 1 mg/kg. Jacqz-Aigrain, in 13 patients found a CL of 0.23 L/kg$h, a VD of 0.45 L/kg and an elimination halflife of 0.86 hours; these values showed marked interindividual variability.9 Andersson38 and Marier39 found that pharmacokinetics of oral omeprazole are similar in children and adults. Pettersen, in critically ill children between 1 month and 5 years, found that the median values for CL, volume of distribution at steady state, and elimination half-life of pantoprazole were 0.14 L/kg$h, 0.22 L/kg, and 2.0 hours, respectively.40 Pantoprazole clearance increased with weight and age. 2013 Lippincott Williams & Wilkins
Population Pharmacokinetics of Omeprazole
FIGURE 7. Distribution of 1000 simulated AUC0–12 h (mg/L$h) values of omeprazole superimposed on the corresponding median values and the 5th and 95th percentiles of observed AUC0–12 h values (dashed lines) calculated from raw data by the noncompartmental approach. The median of simulated AUC0–12 h values (solid line) was within the 5th and 95th percentiles of the observed AUC0–12 h values (dashed lines) in all cases.
VPCs showed that our model adequately mirrored the central tendency and variability of the whole time course of plasma concentrations. The PPC plots, using the AUC as a surrogate marker of efficacy, as proposed by Junghard et al28 for esomeprazole also revealed that the model performed well enough to predict actual mean AUC0–12 h values estimated by the noncompartmental analysis at different doses. As an alternative option to optimize the daily dose of omeprazole, various dose regimens were explored via simulation based on our population PPK model. Our results
TABLE 2. Ratios of the Median Simulated AUC0–12 h in the Studied Population Versus Median AUC0–N in Adults Receiving 20 mg of Intravenous Omeprazole Dose (mg/kg) 0.5 1.0 1.5 2.0
AUC0–12
h
(mg/L$h)
Ratio
770 1620 2430 3240
The target dose would yield a simulated AUC0–12 (1440 mg/L$h) ratio of 1.
0.53 1.13 1.68 2.25 h/target
adult AUC0–N
525
Ther Drug Monit Volume 36, Number 4, August 2014
Solana et al
TABLE 3. Percentages of Patients Achieving Exposures Within the Recommended Median AUC0–12 h (1620 mg/L$h) 620% Interval (1300–1980 mg/L$h) and Those Outside this Interval, Below 220% (,1300 mg/L$h), or Above +20% (.1980 mg/ L$h) Estimated for 3 Different Body Weight Ranges (3–6 kg, .6–14 kg, and .14–25 kg) Body Weight (kg) 3–6 6–14 14–25
AUC0–12
h
, 1300
4420 3860 2780
AUC0–12 h = 1300–1980 2090 2090 2100
AUC0–12
h
. 1980
3490 4050 5110
showed an intravenous dose of 1 mg/kg produced similar exposure, expressed as the AUC value, to a 20-mg dose in adults, suggesting that this could be the recommended dose in children. Nevertheless, further pharmacokinetic studies are necessary to confirm this. Our study has certain limitations including the fact that only a limited number of covariates were recorded and could be tested. This could account for the high BPV of plasma clearance that persisted in the final model. The reasons for this may be unrecorded factors such as genetic polymorphism of CYP2C19 or the possible co-administration of inhibitors of omeprazole metabolism.40 On the other hand, the assay was not sensitive enough (even at 1 ng/mL), which resulted in nearly 30% of the samples giving a result BLOQ. This is the likely reason for the apparent underprediction of trough concentrations, which is related to the use of the Beal M3 method that distributes the BLOQ concentrations from 2infinity to the limit of quantification and negative concentrations are possible in this method. This fact suggests that the model is not under predicting at the trough times because many of the observed trough concentrations that are BLOQ cannot be plotted. Another limitation is the relatively small number of samples obtained beyond 12 hours after the start of the treatment. In summary, the pharmacokinetics of intravenous omeprazole in our study were described using a 2-compartment model with an intravenous administration according to a zeroorder kinetics and linear elimination. Our population pharmacokinetic model demonstrates the relevancy of allometric power models to quantify the effect of size on disposition parameters, enables the changes in CL and V during development to be explained, and allows dose regimens to be established that result in drug exposure equivalent to that observed in adults, for whom omeprazole efficacy has been established. Furthermore, this is the first step toward pharmacokinetic–pharmacodynamic modeling to optimize the dose regimen in this population. REFERENCES 1. Lin PC, Chang CH, Hsu PI, et al The efficacy and safety of proton pump inhibitors vs histamine-2 receptor antagonists for stress ulcer bleeding prophylaxis among critical care patients: a meta-analysis. Crit Care Med. 2010;38:1197–1205. 2. Stollman N, Metz DC. Pathophysiology and prophylaxis of stress ulcer in intensive care unit patients. J Crit Care. 2005;20:35–45. 3. Reveiz L, Guerrero-Lozano R, Camacho A, et al Stress ulcer, gastritis, and gastrointestinal bleeding prophylaxis in critically ill pediatric patients: a systematic review. Pediatr Crit Care Med. 2010;11:124–132.
526
4. Cook DJ, Griffith LE, Walter SD, et al; Canadian Critical Care Trials Group. The attributable mortality and length of intensive care unit stay of clinically important gastrointestinal bleeding in critically ill patients. Crit Care. 2001;5:368–375. 5. Steinberg KP. Stress-related mucosal disease in the critically ill patient: risk factors and strategies to prevent stress-related bleeding in the intensive care unit. Crit Care Med. 2002;30(suppl):S362–S364. 6. Metz DC. Preventing the gastrointestinal consequences of stress-related mucosal disease. Curr Med Res Opin. 2005;21:11–18. 7. Leontiadis GI, Sreedharan A, Dorward S, et al. Systematic reviews of the clinical effectiveness and cost-effectiveness of proton pump inhibitors in acute upper gastrointestinal bleeding. Health Technol Assess. 2007;11: iii–iv, 1–164. 8. Solana MJ, López-Herce J. Pharmacokinetics of intravenous omeprazole in critically ill paediatric patients. Eur J Clin Pharmacol. 2010;66:323–330. 9. Jacqz-Aigrain E, Bellaich M, Faure C, et al. Pharmacokinetics of intravenous omeprazole in children. Eur J Clin Pharmacol. 1994;47:181–185. 10. Litalien C, Théorêt Y, Faure C. Pharmacokinetics of proton pump inhibitors in children. Clin Pharmacokinet. 2005;44:441–466. 11. Faure C, Michaud L, Shaghaghi EK, et al. Intravenous omeprazole in children: pharmacokinetics and effect on 24-hour intragastric pH. J Pediatr Gastroenterol Nutr. 2001;33:144–148. 12. Faure C, Michaud L, Shaghaghi EK, et al. Lansoprazole in children: pharmacokinetics and efficacy in reflux oesophagitis. Aliment Pharmacol Ther. 2001;15:1397–1402. 13. Phillips JO, Olsen KM, Rebuck JA, et al. A randomized, pharmacokinetic and pharmacodynamic, cross-over study of duodenal or jejunal administration compared to nasogastric administration of omeprazole suspension in patients at risk for stress ulcers. Am J Gastroenterol. 2001;96:367–372. 14. Meier-Hellmann A, Reinhart K, Bredle DL, et al. Epinephrine impairs splanchnic perfusion in septic shock. Crit Care Med. 1997;25:399–404. 15. Boucher BA, Wood GC, Swanson JM. Pharmacokinetic changes in critical illness. Crit Care Clin. 2006;22:255–271. 16. McKindley DS, Hanes SD, Boucher BA. Hepatic drug metabolism in critical illness. Pharmacotheraphy. 1998;18:759–778. 17. Jonsson EN, Karlsson MO. Xpose–an S-PLUS based population pharmacokinetic/pharmacodynamic model building aid for NONMEM. Comput Methods Programs Biomed. 1999;58:51–64. 18. Lindbom L, Pihlgren P, Jonsson EN. PsN-Toolkit–a collection of computer intensive statistical methods for non-linear mixed effect modeling using NONMEM. Comput Methods Programs Biomed. 2005;79:241–257. 19. Yamaoka T, Nakagawa T, Uno T. Application of Akaike’s information criterion (AIC) in the evaluation of linear pharmacokinetics equations. J Pharmacokinet Biopharm. 1978;6:165–175. 20. Beal SL. Ways to fit a PK model with some data below the quantification limit. J Pharmacokinet Pharmacodyn. 2001;28:481–504. 21. Meibohm B, Läer S, Panetta JC, et al. Population pharmacokinetic studies in pediatrics: issues in design and analysis. AAPS J. 2005;7:E475–E487. 22. Anderson BJ, Holford NH. Mechanism-based concepts of size and maturity in pharmacokinetics. Annu Rev Pharmacol Toxicol. 2008;48:303–332. 23. Anderson BJ, Holford NH. Mechanistic basis of using body size and maturation to predict clearance in humans. Drug Metab Pharmacokinet. 2009;24:25–36. 24. Savic RM, Karlsson MO. Importance of shrinkage in empirical bayes estimates for diagnostics: problems and solutions. AAPS J. 2009;11:558–569. 25. Yano Y, Beal S, Sheiner LB. Pharmacokinetic-pharmacodynamic models were evaluated using the posterior predictive check. J Pharmacokinet Biopharm. 2001;28:171–192. 26. Mentré F, Escolano Sylvie. Prediction discrepancies for the evaluation of non-linear mixed effects models. J Pharmacokinet Pharmacodyn. 2006; 33:345–367. 27. Bergstrand M, Karlsson MO. Handling data below the limit of quantification in mixed effect models. AAPS J. 2009;11:371–380. 28. Junghard O, Hassan-Alin M, Hasselgren G. The effect of the area under the plasma concentration vs time curve and the maximum plasma concentration of esomeprazole on intragastric pH. Eur J Clin Pharmacol. 2002;58:453–458. 29. Uno T, Niioka T, Hayakari M, et al. Absolute bioavailability and metabolism of omeprazole in relation to CYP2C19 genotypes following single intravenous and oral administrations. Eur J Clin Pharmacol. 2007;63: 143–149.
2013 Lippincott Williams & Wilkins
Ther Drug Monit Volume 36, Number 4, August 2014
30. Shi S, Klotz U. Proton pump inhibitors: an update of their clinical use and pharmacokinetics. Eur J Clin Pharmacol. 2008;64:935–951. 31. Vanderhoff BT, Tahboub RM. Proton pump inhibitors, an update. Clin Pharmacol. 2002;66:273–280. 32. Kaufman SS, Lyden ER, Brown CR, et al. Omeprazole therapy in pediatric patients after liver and intestinal transplantation. J Pediatr Gastroenterol Nutr. 2002;34:194–198. 33. Sakurai Y, Hirayama M, Hashimoto M, et al. Population pharmacokinetics and proton pump inhibitory effects of intravenous lansoprazole in healthy Japanese males. Biol Pharm Bull. 2007;30:2238–2243. 34. Peters HP. Physiological Correlates of Size. Beck E, Birks HJB, Conner EF, eds. Cambridge, MA: Cambridge university press; 1983:48–53. Chapter 4. 35. Kearns GL, Abdel-Rahman SM, Alander SW, et al. Developmental pharmacology-drug disposition action and therapy in infants and children N Engl J Med. 2003;349:1157–1167.
2013 Lippincott Williams & Wilkins
Population Pharmacokinetics of Omeprazole
36. Zimmermann AE, Walters K, Katona BG, et al. A review of omeprazol use in the treatment of acid-related disorders in children. Clin Ther. 2001; 23:660–679. 37. Klotz U. Pharmacokinetic considerations in the eradication of Helicobacter pylori. Clin Pharmacokinet. 2000;38:243–270. 38. Andersson T, Hassall E, Lundborg P, et al. Pharmacokinetics of orally administered omeprazole in children. International Pediatric Omeprazole Pharmacokinetic Group. Am J Gastroenterol. 2000;95: 3101–3106. 39. Marier JF, Dubuc MC, Drouin E, et al. Pharmacokinetics of omeprazole in healthy adults and in children with gastroesophageal reflux disease. Ther Drug Monit. 2004;26:3–8. 40. Pettersen G, Mouksassi M-S, Théorêt Y, et al. Population pharmacokinetics of intravenous pantoprazole in paediatric intensive care patients. Br J Clin Pharmacol. 2009;67:216–227.
527
