Population Pharmacokinetics and Pharmacodynamics of Sotalol in ...

Journal of Pharmacokinetics and Pharmacodynamics, Vol. 28, No. 6, December 2001 ( 2001)

Population Pharmacokinetics and Pharmacodynamics of Sotalol in Pediatric Patients with Supraventricular or Ventricular Tachyarrhythmia Jun Shi,1,4 Thomas M. Ludden,2 Armen P. Melikian,1 Marc R. Gastonguay2,5 and Peter H. Hinderling1,3 Received November 21, 2000—Final October 30, 2001 Aims: To deriûe useful pharmacokinetic (PK) and pharmacodynamic (PD) information for guiding the clinical use of sotalol in pediatric patients with supraûentricular (SVT) or ûentricular tachyarrhythmia (VT). Methods: Two studies were conducted in-patients with SVT or VT in the age range between birth and 12 years old. Both studies used an extemporaneously compounded formulation prepared from sotalol HCl tablets. In the PK study, following a single dose of 30 mg/m2 sotalol, extensiûe blood samples (n=10) were taken. The PK–PD study used a dose escalation design with doses of 10, 30, and 70 mg/m2, each administered three times at 8-hr interûals without a washout. Six ECG recordings for determination of QT and RR were obtained prior to the initial dose of sotalol. Four blood samples were collected six ECG’s were determined during the third interûal at each dose leûel. Plasma concentrations of sotalol (C) were assayed by LC/MS/MS. The data analysis used NONMEM to obtain the population PK and PD parameter estimates. The indiûidual PK and PD parameters were estimated with empirical Bayes methodology. Results: A total of 611 C from 58 patients, 477 QTc and 499 RR measurements from 23 and 22 patients, respectiûely, were aûailable for analysis. The PK of sotalol was best described by a linear two-compartment model. Oral clearance (CL/F) and ûolume of central compartment (Vc/F) were linearly correlated with body surface area (BSA), body weight or age. CL/F was also linearly correlated with creatinine clearance. The best predictor for both CL/F and Vc/F was BSA. The remaining intersubject coefficients of ûariation (CV’s) in CL/F, and Vc/F were 21.6% and 20.3%, respectiûely. The relationship of QTc to C was adequately described by a linear model. The intersubject CV ’s in slope (SL) and intercept (E0) were 56.2 and 4.7%, respectiûely. The relationship of RR to C was also adequately described by a linear model in which the baseline 1

Department of Clinical Pharmacology, Berlex Laboratories, Inc., 340 Changebridge Road, P.O. Box 1000, Montville, NJ 07039. GloboMax LLC, 7250 Parkway Dr., Hanover, MD 21076. 3 To whom all correspondence should be addressed. E-mail: [email protected] 4 Current address: Global Biopharmaceutics, Drug Metabolism and Pharmacokinetics, Aventis Pharmaceuticals, Route 202–206 North, P.O. Box 6800, Bridgewater NJ 08807-0800. 5 Current address: Pharmacokinetics Laboratory-MC 2205, University of Connecticut School of Pharmacy, UCHC, 263 Farmington Ave., Farmington CT 06030. 2

555 1567-567X兾01兾1200-0555$19.50兾0  2001 Plenum Publishing Corporation

556

Shi, Ludden, Melikian, Gastonguay, and Hinderling

RR and SL were related to age or BSA. The intersubject CV ’s for SL and E0 were 86.7 and 14.4%, respectiûely. Conclusions: BSA is the best predictor for the PK of sotalol. Both QTc and RR effects are linearly related to C. No coûariates are found for the QTc–C relation, while the RR–C relation shows age or BSA dependency. KEY WORDS: sotalol; population pharmacokinetics and pharmacodynamics; pediatrics.

INTRODUCTION d, l-Sotalol hydrochloride is an antiarrhythmic agent with Class III and non-specific β blocking properties. In adults sotalol is labeled for life threatening ventricular fibrillation and tachycardia (1) and for maintenance of sinus rhythm in patients with symptomatic atrial fibrillation or atrial flutter (2). In pediatric patients, sotalol has been used for similar indications (3– 6). The pharmacokinetics (PK) and pharmacodynamics (PD) of sotalol have been extensively studied in adults (7). Sotalol is nearly completely absorbed after oral administration. It has negligible plasma protein binding and is not metabolized. It is mainly eliminated unchanged by the kidneys. The PK of sotalol is linear. The respective pharmacological end-points for the Class III and blocking actions are the QTc and RR interval, recorded in the surface ECG (8). However, there is no information on the PK and PD of sotalol in children available. Hence, a basis for deriving a rational dose regimen in children is lacking. As sotalol is predominantly eliminated unchanged by the kidneys in adults, an important impact of maturation of renal function, body surface area (BSA) or body weight (BW) on the PK of sotalol in the pediatric population can be anticipated. In addition, the relationships between drug concentrations and QTc or RR interval prolongation in the pediatric patients are unknown. In order to derive useful PK and PD information for guiding the pediatric use of sotalol, a PK study and a PK–PD study were conducted in pediatric patients. A major challenge in PK and PK–PD studies in neonates and young infants arises from the strict limitation on the number of blood samples and the effect measurements that can be obtained. The population PK–PD approach has been developed to utilize sparse individual data, pool data from different studies with disparate dosing regimens and sampling schedules and allow the determination of PK and PD parameters as well as the influence of patient characteristics on these parameters (9–10). The aim of this investigation was to employ a parametric population approach to determine: (a) the impact of maturation of renal function and body size related changes as well as demographic characteristics on the PK and PD of sotalol in pediatric patients of various age groups with VT and SVT; (b) the population mean and individual PK and PD parameters for sotalol. The ability

PK and PD of Sotalol in Pediatric Patients with SVT or VT

557

to show individual concentration–effect–time plots is very useful in communicating a drug’s characteristics to scientists and clinicians not familiar with mixed-effect modeling. The results obtained in the PK and PK–PD studies using traditional data analysis approaches have already been published (11,12). METHODS Study Design and Patient Population Both studies were multi-center trials. An extemporaneously compounded formulation made by dissolving sotalol HCl tablets in syrup (5 mg兾mL) was used. In the PK study, each patient received a single dose of 30 mg兾m2 BSA. In the PK–PD study, each patient received a total of 9 oral doses with an upward titration without a washout period. Three dose levels, 30, 90, and 210 mg兾m2兾day, were studied. The daily doses were divided in three doses. Doses were given every 8 hr. In both studies, the patients were hospitalized during the course of the study from before the first sotalol dosing to at least a few hours after the last blood sampling. A total of 59 patients, from birth to 12 years old, were enrolled in the studies. Of these, 34 were in the PK study and 25 were in the PK–PD study. Fifty-four (54) were diagnosed to have SVT only, 3 VT only, and 2 SVT and VT. All had a normal renal function (¤80% of normal creatinine clearance for age). The studies were conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study protocols and patient consent procedures were approved by an Institutional Review Board (IRB) at each study center. Informed consent was obtained from all participants after the study was explained to the parents or guardians and the child, as appropriate. Study Procedures The protocols for the PK and PK–PD studies have been described elsewhere (11,12). Only essential details are provided here for clarity. In the PK study, blood samples were collected at 0.5, 1, 2, 3, 5, 8, 12, 16, 22, and 36 hr following the dose. In the PK–PD study, blood samples were taken at 0.5, 2, 4, and 8 hr following the third, sixth, and ninth dose. Each sample contained 0.4 mL blood. A more detailed description of the methods used is given in (11,12). In the PK–PD study, ECG’s were obtained for determination of QT and RR intervals at 0.5, 1.5, 2, 3, 4, and 8 hr after the third, sixth, and ninth doses of sotalol. Baseline QT and RR intervals were determined on six occasions before the first dose of sotalol, at approximately the same

558


times of the day as those recorded following sotalol dosing. Each of these baseline values was treated as an individual observation in the PD data sets. All ECG measurements were taken after a rest period of at least 10 min in the supine position using a chart speed of 50 mm兾sec and an amplitude of 10 mm兾mV. The ECG recordings were blinded to the reader as to the patient, the date and time of recording. A more detailed description of the procedures used is given in (12). Assay of Sotalol in Plasma Plasma concentrations of sotalol were determined using a validated, non-stereoselective, liquid chromatography (LC) method with tandem mass spectrometric (MS) detection (11,12). Database The sotalol plasma concentration data obtained from the PK and PK–PD studies were pooled for analyzing the population PK in the pediatric patients. For the patients participating in the PK–PD study, in addition to the drug concentration data, the QTc and RR interval data were also included in the database. The demographic information in the data base included age, BSA, BW, height (HT), gender, race, serum creatinine level (Ccr) and the estimated creatinine clearance (CLcr). CLcr was computed from K · height (cm)兾serum creatinine concentration (mg兾dL), where KG 0.45 for patientsF1 year and 0.55 for patients ¤1 year (13). BSA was computed from (14) BSA (m2)G

1

Ht(cm)BBW(kg) 3600

Mean QT and RR intervals were determined from 3–5 consecutive sinus beats at each recording time. The QTc interval was obtained from QTc G QT兾RR1兾2 (12,15). Data Analysis and Models The PK–PD data along with the demographic information were analyzed by NONMEM Version V, Level 1.1 to estimate the population mean parameters, inter-subject (η ), and residual (ε ) random effects. Primarily, the first-order (FO) method was used for PK and PD analyses. The final PK model was rerun by using the first-order conditional estimation (FOCE) method with η –ε interaction to refine parameter estimates (10). PK calculations were performed under MS–DOS on a Pentium II 400 mHz PC in conjunction with the Digital Visual Fortran compiler (Version 5.0). PD


559

modeling was performed on an Ultra-2 Sparc Sun Workstation equipped with the Sun WorkShop Fortran Compiler (Version 4.2). One and two compartment models with first order absorption and elimination were tested. The parametrization applied included the key parameters CL兾F, Vc兾F, and ka, where CL兾F is the oral total clearance, Vc兾F is the oral volume of the central compartment and ka is the absorption rate constant. For the two compartment model, the intercompartmental transfer rate constants between central and peripheral compartments, k12 and k21, were included. An absorption lag time was tested by incorporating one more structural parameter, tlag, in the model. Both linear and Emax models were explored for relating the QTc or RR intervals to the concentration of sotalol as follows: EGE0CSL · C

(1)

EGE0CEmax · C兾(EC50CC)

(2)

where E0 and C represent the average baseline effect and the drug concentration, respectively. In Eq. (1) SL is the slope of the linear relationship between effect, E, and drug concentration, C. In Eq. (2) Emax is the maximum drug effect and EC50 represents the concentration at which the effect is one-half the maximum. The population PK parameters (means and variances) were fixed to the values obtained from the final population PK fit throughout the PD model development procedure. The final PK–PD models were used to simultaneously reestimate all PK–PD parameters. The intersubject variability in PK or PD parameters was modeled in a manner consistent with the assumption that these parameters are log-normally distributed in the population as follows: Pi GP · exp(η Pi )

(3)

where Pi is the parameter for the ith individual and P is the population typical value for the parameter. The η Pi is a random variable that accounts for the difference between the individual value and the population typical value. The η Pi is assumed to be normally distributed with mean zero and the variance ω 2[N(0, ω 2)]. The collection of the variances (Ω) of each η of PK and PD parameters comprises the intersubject variability in the PK and PD, respectively. The nondiagonal elements of the variance-covariance matrix were assumed to be zero.

560


For residual variability in the PK and PD profiles, a series of error models was tested as follows: YGFCε ij

(4)

YGF · (1Cε ij )

(5)

YGF · (1Cε ij,1)Cε ij,2

(6)

where Y is the observed plasma concentration or effect, respectively, and F is the corresponding model predicted value. The ε ’s are random terms that are assumed to be N(0, σ 21 and N(0, σ 22), respectively. Depending on the nature of the covariate, a linear model for continuous variables or a step model for dichotomous variables was tested by adding one covariate at a time to the model as follows: PGθ 1Cθ 2 · (factor1Amean1)Cθ 3 · (factor2Amean2) · · ·

(7)

PGθ xCθ y · factor

(8)

where the θ’s are the parameters to be estimated. The factors in Eq. (7) are continuous variables such as BSA, BW or AGE, and the mean is the corresponding average value of the population. The factor in Eq. (8) is a discrete variable (0Gmale, 1Gfemale). In the comparisons between hierarchical models, the value of the objective function was used in χ 2 tests. A change in objective function (∆OFV) of H3.8 was initially taken as indicative of a significant change ( pF0.05) when comparing two models differing in one parameter. If a covariate was statistically significant, it was kept in the model and a new covariate was added and tested. If not, then the covariate was dropped from the model and the effect of another covariate was evaluated (step-up approach). The tentative final model was further tested by eliminating each covariate one at a time to evaluate the change in the objective function (Step-down approach). Because of multiple comparisons, the level of significance for putting the removed covariate back in the model was set at ∆OFVH10.8 ( pF0.001, dfG1). Only covariates that showed a significant contribution were conserved in the model. The model building process was also guided by various goodness-of-fit plots and changes in intersubject and residual variances that occurred as parameters were added or removed from the model or the model form was changed. With nonhierarchical models having equal or similar numbers of parameters a difference of 10 or more units was considered meaningful. The choice among models that exhibited similar overall goodness-of-fit characteristics was guided by parsimony and pragmatism. Therefore, model building in this context is not a purely statistical procedure.


561

Simulation The ability of the final population PK model to realistically describe the observed data was investigated using simulation (16). This model evaluation technique is based on the premise that a model derived from and fitted to a set of observed data should produce similar data when used in a simulation mode. Monte Carlo simulations using the final PK model, including final fixed and random effect parameters (intersubject and residual variances), were conducted. The NONMEM software was used to create a single large data set by simulating observed PK data for 100 replicates of the original PK data set. The resulting simulated observations were sorted by approximate target observation times (10 time-points for the single-dose data and 12 time-points for the multiple-dose data, 4 time-points per studied dosing interval) and 50th (median), 97.5th, and 2.5th quantiles of the simulated data were calculated for each time point. After the final PK model was defined, individual plasma concentration time profiles were simulated using individual post hoc Bayesian PK parameter estimates (CL兾F, Vc兾F, k12, k21, ka, and tlag) at a fixed dose level of 30 mg兾m2. The true steady-state maximum and minimum concentrations, Cmax,ss and Cmin,ss , fluctuation (GCmax,ss兾Cmin,ss), accumulation factors (GCmax,ss兾Cmax,1, where Cmax,1 is the maximum concentration after the first dose), the percentage of true steady-state reached during the third dose intervals of a multiple dose regimen using Cmin and the area under the drug concentration time curve during a dose interval at steady-state were calculated. RESULTS Table I summarizes the characteristics of the patients with analyzable PK and PD data by age category. All patients younger than 3 months had a normal birth weight and gestational age. The PK database consisted of 611 plasma sotalol concentration measurements. Of those 328 were collected from the PK study, 283 were collected from the PK–PD study. The average number of concentration observations per patient was 10.5. There were 499 observed RR intervals from 22 patients (22.7兾patient), and 477 observed QTc values from 23 patients (20.7兾patient) available for analysis. Pharmacokinetics In the initial analyses, the data were better described by the two compartment model than by the one compartment model (∆OFVG160, pF0.001). For residual variance, a combined additive and proportional error model was significantly better than the one with only the proportional

562

Shi, Ludden, Melikian, Gastonguay, and Hinderling Table I. Demographics of Patients with Analyzable Data by Age Category Age Group Neonates ⁄1 month

PK from both studies n 9 Sex, M兾F 5兾4 Race, C兾B兾H兾O 8兾1兾0兾0 BSA, m2 0.23J0.03 Body Weight, kg 3.6J0.8 Height, cm 52.0J2.8 Ccr, mg兾dL 0.5J0.2 CLcr,mL兾min兾1.73 m2 51.8J19.9

Infants >1 to ⁄24 months

Children >2 to