A physiologically based pharmacokinetic model linking plasma protein binding interactions with drug disposition

Research in Veterinary Science 86 (2009) 293–301

Contents lists available at ScienceDirect

Research in Veterinary Science journal homepage: www.elsevier.com/locate/rvsc

A physiologically based pharmacokinetic model linking plasma protein binding interactions with drug disposition J.L. Buur *, R.E. Baynes, G.W. Smith, J.E. Riviere Food Animal Residue Avoidance Databank, Center for Chemical Toxicology Research and Pharmacokinetics, North Carolina State University, College of Veterinary Medicine, 4700 Hillsborough St. Raleigh, NC 27606, USA

a r t i c l e

i n f o

Article history: Accepted 8 July 2008

Keywords: Physiologically based pharmacokinetic model (PBPK) Drug–drug interaction Plasma protein binding Sulfamethazine Flunixin meglumine

a b s t r a c t Combination drug therapy increases the chance for an adverse drug reactions due to drug–drug interactions. Altered disposition for sulfamethazine (SMZ) when concurrently administered with flunixin meglumine (FLU) in swine could lead to increased tissue residues. There is a need for a pharmacokinetic modeling technique that can predict the consequences of possible drug interactions. A physiologically based pharmacokinetic model was developed that links plasma protein binding interactions to drug disposition for SMZ and FLU in swine. The model predicted a sustained decrease in total drug and a temporary increase in free drug concentration. An in vivo study confirmed the presence of a drug interaction. Neither the model nor the in vivo study revealed clinically significant changes that alter tissue disposition. This novel linkage approach has use in the prediction of the clinical impact of plasma protein binding interactions. Ultimately it could be used in the design of dosing regimens and in the protection of the food supply through prediction and minimization of tissue residues. Ó 2008 Elsevier Ltd. All rights reserved.

1. Introduction Physiologically based pharmacokinetic (PBPK) models predict drug disposition based on physiological mechanisms. These models use a series of mass balance equations to link together selected tissue and blood compartments. These models include physiological parameters (e.g. blood flow, tissue volume), physiochemical parameters (e.g. tissue:blood partition coefficients) as well as those obtained from in vitro studies (e.g. Michaelis–Menten enzyme kinetics and protein binding) (Reitz et al., 1988; Grass and Sinko, 2002; Teeguarden et al., 2005). PBPK models allow for the analysis of data from multiple study designs as well as predictions for a wide variety of dosing regimens, routes of administration, species and interindividual variability (Riviere, 1999; Young et al., 2001; Clewell et al., 2004, Gentry et al., 2003). The flexibility of PBPK models lends itself to the testing of biological hypotheses, prediction of tissue dosimetry, and the refinement of pharmacokinetic mechanisms. Currently, PBPK models are used in toxicology to predict internal dose metrics, in human medicine to calculate individual dose regimens for drugs of low therapeutic indices such as cancer chemotherapeutic agents, and in veterinary medicine to estimate meat withdrawal times for

* Corresponding author. Address: Western University of Health Sciences, College of Veterinary Medicine, 309 East Second Street, Pomona, CA 91766, USA. Tel.: +1 909 706 3518; fax: +1 909 469 5635. E-mail address: [email protected] (J.L. Buur). 0034-5288/$ - see front matter Ó 2008 Elsevier Ltd. All rights reserved. doi:10.1016/j.rvsc.2008.07.003

drugs after extralabel use (Kawai et al., 1994; Tsukamoto et al., 2001; Craigmill, 2003; Gentry et al., 2003; Buur et al., 2005, 2006a,b). PBPK models have also been used to predict drug interactions due to enzyme inhibition and to explore mechanisms behind specific pharmacokinetic phenomenon such as time to reach equilibrium in protected spaces or transdermal absorption of chemicals (Simmons, 1996; Kanamitsu et al., 2000; Liu et al., 2005; van der Merwe et al., 2006). Since combination drug therapy is quickly becoming the standard of practice in both human and veterinary medicine, there is an increased likelihood of adverse drug reactions occurring due to drug–drug interactions (Saltvedt et al., 2005). The safety and efficacy of chemotherapeutics is determined by the concentration of free drug in the system. The interplay between a drug’s affinity for plasma proteins, defined by the dissociation constant Kd, and the maximum plasma protein binding capacity, defined by Bmax, contribute to the amount of free drug available in the system. Ultimately, free drug concentration is determined by a variety of physiological mechanisms including systemic clearance as well as the binding properties defined above (Wilkinson, 2001). Alterations in either Kd, Bmax, or both could cause an increase or decrease in free drug concentration. Many theoretic arguments have been presented to show that the alteration of free drug concentration within an open system, such as a patient, would be transient due to compensatory mechanisms in free drug clearance and thus plasma protein binding interactions would have no clinical effect (Benet and Hoener, 2002; Toutain and Bousquet-Melou, 2002). However,

294

J.L. Buur et al. / Research in Veterinary Science 86 (2009) 293–301

there have been no attempts to model or experimentally validate these interactions. We hypothesized that PBPK models could be used as a tool to elucidate the underlying mechanism of and to evaluate the clinical consequences of plasma protein binding interactions. To our knowledge, there are no published models that link plasma protein binding interactions to drug disposition using a PBPK modeling approach. Sulfamethazine (SMZ), a sulfonamide antibiotic, is commonly used in swine medicine. It is labeled for use in the prevention and treatment of cervical abscesses, colibacillosis, swine dysentery, and bacterial pneumonia. Flunixin meglumine (FLU), a nonsteroidal anti-inflammatory agent, has recently been approved for use in swine for the control of pyrexia associated with respiratory disease (Anon., 2006). Alterations in the pharmacokinetics of SMZ due to FLU have been reported in horses and attributed to protein binding interactions (el-Banna, 1999). Similar interactions in swine could theoretically alter tissue disposition and lead to increased tissue residues in the meat supply. The purpose of this study was to develop a theoretical PBPK model that links plasma protein interactions to the disposition of an individual drug. Given the likelihood of concurrent administration of SMZ and FLU in swine, these compounds were used as model drugs. Additionally, in vivo studies were conducted to assess the accuracy of the model prediction.

Q quick

Quick Plasma Renal Clearance SMZ Cf

Q slow

Slow Q

tot

Protein

Q liver

Liver FLU

Deacetylation

Acetylation

2. Materials and methods Q

2.1. SMZ PBPK model development

liver

Liver Met

A flow-limited PBPK model was developed for predicting SMZ concentrations in swine. The model consisted of tissue compartments including quickly perfused tissues, slowly perfused tissues, liver, and blood. Additional compartments for the N-acetyl metabolite were included and consisted of the liver, blood, and remaining body. In total, the final model had seven compartments (Fig. 1), and intravenous inputs for SMZ and FLU. Physiological parameters of organ volumes and blood flow rates were taken from the literature where available (Tranquilli et al., 1982; Lundeen et al., 1983; Pond, 2001; Buur et al., 2005). Slowly perfused tissue included muscle, gastrointestinal tract, bone, and fat. Quickly perfused tissue included kidney and other tissues. Where appropriate, parameters were calculated as the difference between unity and the remaining tissues. The density of plasma was assumed to be 1 g ml1. Hepatic blood flow was modeled as the combination of hepatic arterial and portal circulations. Renal clearance is mainly due to filtration and was modeled as a first order rate constant from the quickly perfused tissue block. Enterohepatic recycling of SMZ was considered insubstantial and not included in the model. The unique acetylation–deacetylation pathway of SMZ in swine was incorporated into the model as described previously (Buur et al., 2005). Final parameter values can be found in Table 1. Differential equations were used to describe the rate of change in mass in each compartment (Table 2). Model simulations were solved using ACSLxtreme, Version 2.3.0.12 (Aegis Technologies Group Inc., Huntsville, AL, USA). Protein binding was assumed to be linear in nature and free and bound drug concentrations were determined by the use of the following equations:

CT ¼

A V plasma

ð1Þ

CF ¼

Kd  C T Bmax þ Kd

ð2Þ

C met

Plasma Q tot Q body

Body Met

Clearance Met

Fig. 1. Schematic diagram of the PBPK model for sulfamethazine in swine. Cf, free drug concentration; Cmet, concentration of metabolite; Q, tissue blood flow; Qtot, cardiac output; SMZ, sulfamethazine; FLU, flunixin meglumine; Straight arrows, blood flow; Curved arrows, protein binding equilibrium.

CB ¼ CT  CF

ð3Þ

CT, CF, and CB are the total, free, and bound concentration of drug in the plasma compartment, respectively; A is the amount of drug; Vplasma is the volume of the plasma compartment; Kd is the dissociation constant for the drug; and Bmax is the maximum binding occupancy for the drug. Mass transferred to tissue blocks was limited to free drug concentration. Values for Kd and Bmax were taken from in vitro data derived in our laboratory. Values for parameters not available in the literature were estimated using the parameter estimation module included in the simulation software. Model parameters were adjusted to ‘‘best fit” by use of a maximum-likelihood estimation algorithm. Limits were set to ensure biologically plausible values. The optimization data set was created using data collected from four published studies (Nouws et al., 1986; Nouws et al., 1989; Sweeney et al., 1993; Yuan et al., 1997) and consisted of mean total plasma concentrations calculated from 12, 6, 7, and 3 samples/data point

J.L. Buur et al. / Research in Veterinary Science 86 (2009) 293–301 Table 1 Final values for parameters used within the sulfamethazine PBPK model and sulfamethazine–flunixin protein binding interaction PBPK model Parameter

Units

Value

Reference

Cardiac output

h1 kg1

12

Hematocrit Blood flow Quickly perfused

% 33 % Cardiac output 0.51

Tranquilli et al. (1982) and Lundeen et al. (1983) Tranquilli et al. (1982)

Slowly perfused

0.25

Liver

0.24

Body (metabolite) Blood

0.76 1

Organ volume Quickly perfused Slowly perfused Liver Body (metabolite) Blood

Fraction bodyweight 0.86 0.07 0.02 0.92 0.06

Tissue: blood partition coefficient Quickly perfused Slowly perfused Liver (parent) Liver (metabolite) Body (metabolite) Clearance Hepatic ml min1 kg1 Renal (parent) ml min1 kg1 Body (metabolite) ml min1 kg1 Deacetylation Rate h1 Binding parameters SMZ Kd SMZ Bmax FLU Kd FLU Bmax Protein binding (metabolite)

lg ml1 lg ml1 lg ml1 lg ml1 %

Pond (2001) Pond (2001) Pond (2001) 1-Liver-plasma Pond (2001) Estimated Estimated Estimated Buur et al. (2005) Buur et al. (2005)

0.62 0.35 2.56 3.66

Buur et al. (2005) Estimated Buur et al. (2005) Buur et al. (2005)

59.17 244.4 147.3 1833.5 0.57

In vitro study in vitro study in vitro study in vitro study Buur et al. (2005)

for each of the four studies, respectively. Estimation and optimization was performed for the tissue:blood partition coefficients (quickly perfused, slowly perfused, liver) and renal clearance. Validation of the SMZ model was performed by comparison of simulations to an external data set. The validation data set comprised values for total SMZ plasma concentration from four individual pigs reported in a previous study (Buur et al., 2006a,b). Simulated predicted values were plotted against observed values and a regression line was plotted against correlated values. Residual values were also evaluated for spread and distribution. 2.2. Protein binding interaction linkage Flunixin was incorporated into the SMZ model described above as a direct input within the blood compartment. Bound and free FLU drug concentrations were modeled using Eqs. (2) and (3). Interactions between SMZ and FLU were modeled assuming competitive inhibition using the following interaction equation.

  C FB Kd ¼ KdA  1 þ KdB

compared for both free and total SMZ drug concentrations against concentrations observed in a small pilot in vivo study described below. 2.3. In vitro protein binding saturation studies

Tranquilli et al. (1982) and Lundeen et al. (1983) Tranquilli et al. (1982) and Lundeen et al. (1983) Tranquilli et al. (1982) and Lundeen et al. (1983) 1-Liver

3.20 2.01 0.07 0.079 1.297

295

ð4Þ

Subscripts A and B refer to SMZ and FLU, respectively. The calculated Kd was then incorporated into the SMZ binding calculations within the SMZ model. Simulations that incorporated protein binding interactions were derived by directly incorporating observed total FLU concentrations into the model and recording the resulting free and total SMZ concentrations. Simulations were

For all protein binding studies, fresh plasma was obtained from multiple swine (ranging from 3 to 5 per experiment) using jugular venipuncture. Whole blood was spun at 1500 rev min1 for 10 min and plasma was harvested. Plasma was stored at room temperature and used within 3 h of the initial blood sample. For SMZ binding saturation studies, fresh plasma was diluted 1:10 (vol/vol plasma: phosphate buffered saline pH 7.4). Plasma samples were spiked with SMZ (in methanol vehicle) to final concentrations of 0, 0.5, 1, 5, 10, 20, 30, 60, 100, and 150 lg ml1. Samples were incubated and allowed to reach equilibrium in a 37 °C water bath for 60 min. One milliliter of sample was transferred to Centrifree YM-30 (Millipore, Bedford, MA) ultrafiltration cartridge (30,000 molecular weight cut off) and spun in a fixed rotor centrifuge at 2000g for 20 min. Ultrafiltrate was injected directly onto the HPLC. Each concentration was repeated in triplicate. Preliminary competitive binding studies determined that nonspecific binding was insignificant (data not shown). Mean data from multiple experiments were fit to a one site binding hyperbola, using Prism v. 4.03 for Windows (GraphPad Software, Inc., San Diego CA, USA). For FLU binding saturation studies, both total and non-specific binding was quantified. Fresh plasma was diluted 1:250 (vol/vol) with phosphate buffered saline (pH 7.4) to account for ligand depletion. Samples were spiked to FLU final concentrations of 0, 0.3, 1, 3, 30, 60, 100, 300, 600, 1000 lg ml1 and were incubated for 60 min in order to reach equilibrium in a 37 °C water bath with or without 170 lg ml1 (dissociation constant in human plasma) phenylbutazone. Phenylbutazone experiments were used to determine non-specific binding of FLU. One half milliliter was transferred to a Centrifree YM-30Ò (Millipore, Bedford, MA) ultrafiltration cartridge (30,000 molecular weight cut off) and spun in a fixed rotor centrifuge at 2000g for 30 min. Ultrafiltrate was discarded and clean receptacles were placed on the ultracentrifugation devices. One milliliter of methanol was added and the devises were vortexed for 30 s. Samples were then respun at 2000g for 30 min and the ultrafiltrate injected onto the UPLC. Specific binding was calculated by subtracting non-specific binding from total binding. Data was fit to a one site binding hyperbola as described for the SMZ studies. 2.4. Quantification methods 2.4.1. HPLC The HPLC system consisted of a Waters Alliance 2695 Separation Module with a 996 photodiode array detector (Milford, MA). A Zorbax SB-C8 column (4.6 by 150 mm, 5 um; Agilent, Wilmington, DE) was used. All injection volumes were 10 ll and flow rates were 1 ml min1. Autosampler and column temperatures were maintained at 25 °C and 30 °C, respectively. The optimum detector wavelength was 267 nm. Mobile-phase conditions were acetonitrile:ammonium acetate buffer (pH 4.5, 0.1 M) (17:83; vol/vol). Run times were 9 min. SMZ plasma samples were prepared using the technique as previously described (Buur et al., 2006a,b). Briefly, 1 ml of plasma was acidified with 20 ll O-phosphoric acid and then added to an Oasis MCX 3cc 60 mg sorbent weight (Waters, Milford, MA) cartridge that had been conditioned with 1 ml of methanol followed by 1 ml of water. Samples were washed with 1 ml of 0.1 N HCl and 1 ml of methanol and then dried under vacuum for 30 s. Elution

296

J.L. Buur et al. / Research in Veterinary Science 86 (2009) 293–301

Table 2 Differential equations used to describe the rate of change of sulfamethazine in each tissue compartment Tissue compartment equation Quickly perfused

V quick 

  dC quick C quick  Q quick  C a  Clrenal ¼ Ca  dt Pquick

V slow 

  dC slow C slow  Q slow ¼ Ca  dt Pslow

V liver 

  dC liver C liver ¼ Ca   Q liver  C liver  Clhepatic þ C livermet  Cldeacetylation dt Pliver

V liver 

  dC livermet C livermet ¼ Ca   Q liver  C livermet  Cldeacetylation þ C liver  Clhepatic dt Plivermet

V body 

  dC body C body ¼ Ca   Q body  C body  Clbody dt Pbody

V blood 

dC T ¼ dt

V blood 

    C body dC met C livermet  Q livermet þ  Q body  C met  Q tot ¼ dt Plivermet P body

Slowly perfused

liver

Liver (metabolite)

Body (metabolite)

Plasma

Plasma (metabolite)

      C slow C fast C liver  Q slow þ  Q fast þ  Q liver þ IV Dose  C F  Q tot Pslow Pfast Pliver

V: volume; C: concentration SMZ; Q: blood flow; P: tissue:blood partition coefficient; Cl: clearance; IVDOSE: amount injected; Qtot: cardiac output. Subscripts refer to quick (quickly perfused), slow (slowly perfused), liver, blood (plasma), and body tissue compartments. Subscripts renal, hepatic, and deacetylation refer to methods of clearance. Subscripts a, T, F, and met refer to concentration of drug within the arterial circulation, total sulfamethazine, free sulfamethazine and N-4 metabolite concentrations, respectively.

was performed with 1 ml of ammonium hydroxide–methanol (5:95; vol/vol) and dried under vacuum for 30 s. Elution volumes were evaporated to dryness in a Turbo Vap LV evaporator (Zymark; Hopkington, MA) for 15 min at 50 °C and 15 mm Hg psi reagentgrade nitrogen gas. Residue was reconstituted in 0.5 ml ammonium acetate buffer (pH 4.5, 0.1 M), vortexed, and then injected onto the HPLC system. Concentrations were determined by comparison of peak areas to an external standard curve created by spiked plasma samples put through the same clean up process. Coefficients of variation for both interday and intraday were below 15% and recoveries ranged were between 90% and 105%. The standard curve was linear from 0.1 to 10 lg ml1. Samples were diluted until concentrations fell within the curve limits. Quality control samples run on the day of analysis had recoveries greater than 90%. Limit of detection was 0.05 lg ml1 and limit of quantification was 0.1 lg ml1. Ultrafiltrate from SMZ protein binding studies was injected directly onto the HPLC system. Concentrations were compared to an external standard curve prepared from SMZ standards in mobile phase. The standard curve was linear from 0.05 to 10 lg ml1. Limit of detection and limit of quantification were 0.01 and 0.05 lg ml1, respectively. 2.4.2. UPLC–MS The Acquity UPLC–MS system (Waters, Milford, MA) consisted of a BEH C18 column (1.7 um, 2.1 by 50 mm, Waters, Milford, MA) and filter disc. The mobile phase was acetonitrile: 1% acetic acid in water (50:50; vol/vol). The EMD 1000 was a single quadrupole mass spectrometer run in ESI+ mode. The ion used for quantification was 297. Column and sample temperatures were

maintained at 30 °C and 24 °C, respectively. Run times were 1.5 min. FLU plasma samples were prepared using the technique as previously described (Buur et al., 2006a,b). Briefly, 0.5 ml of plasma was acidified using 20 ll of O-phosphoric acid. Samples were added to an Oasis MCX 6cc 150 mg sorbent weight (Waters, Milford, MA) cartridge that had been conditioned with 3 ml of methanol followed by 3 ml of water. Samples were washed with 3 ml of 0.1 N HCl and 3 ml of methanol: 0.1 N HCl (90:10; vol/vol) and then dried under vacuum for 30 s. Elution was achieved using 3 ml of methanol:ammonium hydroxide (5:95; vol/vol). Samples were again dried under vacuum for 30 s. Elution volumes were evaporated to dryness in a Turbo Vap LV evaporator for 20 min at 50 °C and 15 mm Hg psi reagent grade nitrogen gas. Residue was reconstituted in 250 ll mobile phase and filtered through a 22 um nylon syringe filter and then injected onto the UPLC system. Concentrations were determined by comparison of peak areas to an external standard curve created by spiked plasma samples put through the same clean up process. Coefficients of variation for both interday and intraday were below 15% and recoveries were within 15% of true concentration. The standard curve was linear from 0.01 to 10 lg ml1. All samples were diluted until the results fell with the curve limits. Quality control samples run on the day of analysis had recoveries greater than 90%. Limit of detection was 0.01 lg ml1 and limit of quantification was 0.02 lg ml1. 2.5. In vivo interaction study Double lumen catheters were placed in both the left and right jugular veins of three female Yorkshire cross pigs weighing be-

297

J.L. Buur et al. / Research in Veterinary Science 86 (2009) 293–301

3. Results 3.1. In vitro protein binding characteristics for SMZ and FLU Final binding curves for SMZ and FLU in fresh plasma are presented in Figs. 2 and 3, respectively. Plasma was diluted in order to eliminate the effects of ligand depletion. All curves fit a one site hyperbola. Final results for Kd and Bmax are presented in Table 3. 3.2. In vivo model validation of SMZ model When compared to the external data set, the SMZ PBPK model with SMZ alone had a plasma concentration correlation (R2) of 0.85. Results of the simulation plotted against observed concentration as well as residual analysis and validation regression are presented in Fig. 4. Residual analysis revealed a generalized underprediction for total drug concentration. 3.3. Protein binding interaction PBPK model Comparisons between model predictions and in vivo observations for individual pigs are presented in Fig. 5 (total drug) and Fig. 6 (free drug). Model simulations predicted an interaction that is described by a 59% decrease in total drug concentration followed by a rapid return to steady state concentrations and a 1% increase in free drug concentration that also returns to steady state concentrations. Residual analysis for total and free drug SMZ concentrations are presented in Figs. 7 and 8, respectively. Given the steady state nature of both the free and total SMZ concentrations, there was an inadequate spread of SMZ concentrations to calculate a correlation coefficient. The in vivo study confirmed the presence of a drug interaction between SMZ and FLU in pigs. The interaction was characterized

Bound Sulfamethazine (ppm)

25

20

15

10

5

0 0

50

100

150

Free Sulfamethazine (ppm) Fig. 2. Protein binding saturation curve for sulfamethazine in fresh porcine plasma. Data points represent the mean and standard deviations from five experiments with each experiment having three replicates and plasma pooled from a minimum of three pigs.

14 12

Bound Flunixin Meglumine (ppm)

tween 15.15 and 17.5 Kg. Catheterization was performed under anesthesia induced with a combination of ketamine, tiletamine, and xylazine. Catheters were placed using the Seldinger technique and placement was verified by successful blood draws (Taber and Bergamini, 1997). All procedures were approved the North Carolina State University Institutional Animal Care and Use Committee. At the time of catheterization, each pig was given an IV dose of 27 mg kg1 SMZ and then were started on a constant rate infusion (CRI) of SMZ at a rate of 1.5 mg kg1 h1. All CRI’s were maintained by the Homepump Eclipse C-Series pump (100 ml volume, 0.5 ml h1 rate, I-Flow Co., Lake Forest, CA) attached to the backs of the pigs. All blood samples were 10 ml in volume and taken from the contralateral jugular catheter. The in vivo study consisted of three pigs prepped as described above. For 2 pigs, blood samples were obtained at 0, 18, 25, 42, and 49 h after start of the SMZ CRI to confirm steady state. At 68 h after the start of the SMZ CRI, a 2.0 mg kg1 bolus dose of FLU was given through the second lumen of the SMZ CRI catheter. For the 3rd pig, blood samples were obtained at 22 and 28 h post SMZ CRI to confirm steady state and the bolus FLU dose (2.0 mg kg1) was given 45 h post SMZ CRI. Blood samples for all 3 pigs were continued at 5, 10, 15, 30, and 45 min, and 1, 2, 4, 8, 12, 24, 36, 48, 60, 72, 84, and 96 h post FLU dose. All samples were kept at 37 °C in a hot water bath until spun at 1500 rev min1 for 10 min and the plasma was harvested within 30 min of sampling. One milliliter of plasma was immediately transferred to a Centrifree YM-30 (Millipore, Bedford, MA) ultrafiltration cartridge (30,000 molecular weight cut off) and spun at 2000g for 30 min. The ultrafiltrate and remaining plasma sample were frozen at 80 °C until analyzed. Analysis of total and free SMZ as well as total FLU concentrations occurred within 1 week of sampling.

10 8 6 4 2 0 0

200

400

600

800

1000

Free Flunixin Meglumine (ppm) Fig. 3. Representative protein binding saturation curve for flunixin in fresh porcine plasma. Data points are comprised of means and standard deviations from five replicates and plasma was pooled from three pigs. Error between replicates was less than 5% and so error bars do not show on this scale.

Table 3 Final protein binding kinetic parameters for Sulfamethazine (n = 5 studies pooled from three pigs) and flunixin (n = 5 studies pooled from five pigs) in fresh porcine plasma Drug

Kd (lMol)

Bmax (lMol)

Sulfamethazine Flunixin meglumine

212.6 ± 39 299.7 ± 117

878.1 ± 75 3730.7 ± 514

by a 25% decrease in total drug concentration and an increase in free drug concentration (1.5%). Time to reach steady state after bolus injection of FLU was on average 3.4 h for total drug concentration. Free drug concentration was highly variable throughout the course of the study for each individual pig. Average parameters are presented in Table 4. Comparison between the predicted simulations and the observed data from the in vivo study revealed that in both cases the interaction could be described by a decrease in total drug and an increase in free drug concentrations, thus creating a temporary

298

J.L. Buur et al. / Research in Veterinary Science 86 (2009) 293–301 1000

Sulfamethazine Concentration (ppm)

A

Sulfamethazine (ppm)

10

1

0.1

100

10

1 0

20

40

60

80

0.01 0

20

40

60

80

120

140

160

100

Time (h)

B Time (h) 0 0

20

40

60

80

100

-0.5 -1 -1.5 -2

Sulfamethazine Concentration (ppm)

1000

0.5

Residual

100

Time (h)

-2.5

100

10

1 0

20

40

60

80

-3

100

120

140

160

180

120

140

160

180

Time (h)

-3.5

Sulfamethazine Concentration (ppm)

1000

Observed Sulfamethazine Concentration (ppm)

-4

C 7 6 5 4 3

100

10

2 1 0

1

20

40

60

80

100

Time (h) 0

0

0.5

1

1.5

2

2.5

Predicted Sulfamethazine Concentration (ppm) Fig. 4. Results of validation of sulfamethazine PBPK model in swine for total drug. (A) Model simulation (line) compared to external data set (squares). (B) Residual plot. (C) Comparison of predicted and observed data points. The solid line represents the line of best fit.

change in the fraction unbound. There were significant differences between model simulation and observed data for all comparisons except the total drug concentrations at steady state and the percent change in concentration of free drug during the time of interaction. In addition the model simulation did not predict as rapid of a recovery back to steady state conditions for total drug concentra-

Fig. 5. Comparison between observed (squares with dashed line) and predicted (solid line) values of total drug concentration for individual pigs using the sulfamethazine–flunixin protein binding interaction PBPK model.

tions as seen in the in vivo study. These comparisons are presented in Table 4. 4. Discussion We have developed a PBPK model that links plasma protein interactions to drug disposition that is consistent with known pharmacokinetic mechanisms (Rowland and Tozer, 1995) Our model predicted a temporary increase in the free drug concentra-

299

J.L. Buur et al. / Research in Veterinary Science 86 (2009) 293–301 100 80 60 40

Residual

Sulfamethazine Concentration (ppm)

100

10

20 0 20

40

60

80

100

120

140

160

180

-20 -40 -60 -80 -100

1 0

20

40

60

80

100

120

140

160

Time (h)

-120

Time (h)

Fig. 7. Residual plot for fit of the results of the sulfamethazine–flunixin protein binding PBPK model for total sulfamethazine concentration.

Sulfamethazine Concentration (ppm)

100

10

1 0

20

40

60

80

100

120

140

160

180

120

140

160

180

Time (h)

Sulfamethazine Concentration (ppm)

100

10

1 0

20

40

60

80

100

Time (h) Fig. 6. Comparison between observed (squares) and predicted (solid line) values of free drug concentration for individual pigs using the sulfamethazine–flunixin protein binding interaction PBPK model.

tion due to displacement from plasma proteins. The increase in free drug concentration would allow more drug to be cleared and thus contribute to a concurrent decrease in total drug concentration. Since there was continual input of drug from the CRI, the decrease in total drug concentration was transient in nature. These model predictions parallel the theoretical models presented by Benet and Hoener (2002) and Toutain and Bousquet-Melou (2002). Additionally, these results parallel in vitro binding studies performed with warfarin and phenylbutazone (Harder and Thurmann, 1996). The model simulations also parallel the observed interaction seen in the in vivo study.

Comparison between model simulations and observed data revealed a similar trend in the interaction effect but a difference in both magnitude and duration of the effect. Most likely this is due to interindividual variability. This is further substantiated by the large variability seen in the in vivo study. Accuracy of the model prediction could be increased by direct quantification of Kd, Bmax from in vitro studies for each individual pig. Anesthesia alters blood flow and thus distribution of drugs. However, the effects of anesthesia would have dissipated long before the challenge bolus dose of FLU. This is also supported by the fact that our steady state concentrations were accurately predicted by the model. The model assumes mean values for Kd and Bmax for both SMZ and FLU. Flunixin binding saturation studies showed variability between experiments. This is most likely due to variability within the in vitro system as well as interindividual variability. The numbers of replicates and number of pigs used to pool plasma were increased in order to reduce the variability. However, due to solubility problems, concentrations of FLU greater than 1000 lg/ml could not be achieved. Terminal portions of the protein binding saturation curve tended to have larger variation which could be explained by the precipitation of FLU within the in vitro system. While this affected both Kd and Bmax, the differences between the individual experiments were not statistically significant (data not shown). Also, each experiment had very good fits at early time points that correspond to concentrations found in vivo. It is possible that alterations in these binding parameters for each individual pig could decrease the accuracy of the model. The high percentage of protein binding of FLU could have contributed to ligand depletion and thus altered the in vitro results for both Kd and Bmax. To minimize these effects, we diluted plasma such that only 10% of FLU was bound at lower concentrations. However, this could contribute to the variability seen within the FLU binding studies. Additionally, preliminary studies showed there was a significant difference between the Kd and Bmax parameter values derived from fresh plasma and those derived from plasma that had been frozen for up to one week (data not shown). This illustrates the dependency upon good parameter values inherent within a PBPK modeling approach. There are minor differences between the calculated total drug concentration (based on Kd and Bmax) and those predicted by the model and observed in the in vivo study (data not shown). These variations most likely reflect the variability within the binding studies. Another contributing factor could interindividual variability. It is also important to take into account that these comparisons were made between means and not calculated for each individual animal. Also, binding studies and calculations based on the in vitro

300

J.L. Buur et al. / Research in Veterinary Science 86 (2009) 293–301 20 15

Residual

10 5 0 20

40

60

80

100

120

140

160

180

-5 -10 -15 -20 -25

Time (h)

Fig. 8. Residual plot for fit of the results of the sulfamethazine–flunixin protein binding PBPK model for free sulfamethazine concentration.

Table 4 Comparison of model simulation and in vivo data for sulfamethazine at the time of interaction Parameter

Units

Model prediction

Observed data

Total drug Steady state Interaction % change from steady state

lg ml1 lg ml1

92.0 ± 3.2 37.8 ± 14 59.2 ± 25

72.7 ± 11.8 55.3 ± 32.8 25.0 ± 38.9

17.7 ± 0.4 18.0 ± 0.6 1.1 ± 1.5

13.9 ± 3.4 14.1 ± 0.5 1.5 ± 18.8

19.3 ± 0.3 52.1 ± 19.1 2.7 ± 0.9 14.7 ± 8.4

19.2 ± 4.4 32.0 ± 21.9 1.9 ± 1.8 3.4 ± 0.9

Free drug Steady state Interaction % Change from steady state Fraction unbound Steady state Interaction Fold increase from steady state Time to reach steady state after FLU bolus

%

lg ml1 lg ml1 % % % h

Each number represents mean ± standard deviation for n = 3 pigs.

studies may not reflect the entire complex interaction between drug and protein seen in an in vivo environment. Protein binding interactions are considered to be important for drugs with low extraction ratios (ER < 0.3) (Wilkinson, 2001). For a low ER drug, there would be a decrease in total drug concentration while free drug concentration would remain unchanged due to increased clearance of free drug. Conversely, a high ER drug would see no change in total drug concentrations with a concurrent increase in free drug concentration (Rowland and Tozer, 1995). Sulfamethazine has been reported to have an extraction ratio of 0.58 in rabbits (Reeves et al., 1988). The extraction ratio has not been reported for pigs. As a moderate ER drug, you would expect a mild, but not clinically significant alteration in kinetics due to changes in protein binding. A single study in horses showed that there was a decrease in the mean residence time (MRT) and an increase in the elimination rate (Kel) for total SMZ when administered in the presence of FLU. Authors concluded that these changes were due to displacement from plasma proteins (el-Banna, 1999). This would reflect the increase in free SMZ concentration due to displacement from plasma proteins. The increased free SMZ allows a greater fraction of total drug to be cleared at any one time, thus resulting in a decrease in total drug concentration and concurrent decrease in MRT. These conclusions are consistent with our model and with the current thinking on the clinical effects of protein binding interactions (Toutain and Bousquet-Melou, 2002). Our model predicted

a temporary decrease in total drug concentration that would alter the calculation of MRT. The horse study used single bolus doses as the dosing regimen. Since our model and the in vivo studies used CRI’s, the binding effect of FLU was transient in nature. The PBPK model did not predict, nor was there any significant change in free drug concentrations in the in vivo experiments. Since clinical activity is based on free drug concentration, there is no evidence that this interaction would contribute to a clinical significant change in either safety or efficacy of SMZ. Additionally, since distribution into tissues is also determined by free drug concentration, there should be no significant alterations in tissue residues when both SMZ and FLU are given concurrently to pigs. Previous PBPK models for SMZ in pigs have provided accurate predictions of total drug concentrations within edible tissues for individual pigs as well as over pig populations (Buur et al., 2005, 2006a,b). Inclusion of linkage models such as the PBPK model presented here into such existing tissue residue models would allow for further refinement of tissue residue concentration and withdrawal time predictions during concurrent drug administration since drug disposition is determined by free drug concentration. The model assumed that protein binding interaction was due to competitive inhibition. It is possible that other forms of interaction, such as allosteric/cooperative binding or noncompetitive inhibition, could contribute to the interactions seen in vivo. These types of interactions would contribute to different interaction curves. The exploration of different linkage mechanisms is worthy of future study. Additionally, further research needs to be conducted on drugs that have altered binding properties and lower extraction ratios. 5. Conclusion We have developed a PBPK model that predicts changes in both total and free drug disposition due to plasma protein binding interactions. This novel use of PBPK models represents the first time a PBPK model has been used to elucidate the underlying mechanisms of plasma protein binding interactions and to link that interaction to drug disposition. Additionally, no clinically relevant interaction occurred or was predicted by our model for the concurrent use of SMZ and FLU in pigs. Beyond the use of exploring mechanisms, this type of modeling approach is also useful in elucidating the clinical implications of drug interactions. Linkage models, such as this one, can be included into previously validated PBPK models and used to explore the effects of chemical mixtures on drug disposition. Given the frequency of concurrent drug use in both human and veterinary medicine, PBPK models such as this one will be important for accurate derivation of drug dosing regimens and meat withdrawal times. And they will protect human health not only from toxicity or therapeutic failure, but also by protecting the food supply from increased tissue residues. Acknowledgments Funding for this project was supported in part by the Food Animal Residue Avoidance Databank (FARAD) and USDA-Cooperative State Research, Education and Extension Service Grant No. 200245051-01362. References Anon., 2006. Compendium of Veterinary Products. Adrian J. Bayley, Port Huron, Michigan. Benet, L.Z., Hoener, B.A., 2002. Changes in plasma protein binding have little clinical relevance. Clin. Pharmacol. Ther. 71 (3), 115–121. Buur, J., Baynes, R., Smith, G., Riviere, J., 2006a. Use of probabilistic modeling within a physiologically based pharmacokinetic model to predict sulfamethazine

J.L. Buur et al. / Research in Veterinary Science 86 (2009) 293–301 residue withdrawal times in edible tissues in swine. Antimicrob. Agents Chemother. 50 (7), 2344–2351. Buur, J.L., Baynes, R.E., Craigmill, A.L., Riviere, J.E., 2005. Development of a physiologic-based pharmacokinetic model for estimating sulfamethazine concentrations in swine and application to prediction of violative residues in edible tissues. Am. J. Vet. Res. 66 (10), 1686–1693. Buur, J.L., Baynes, R.E., Smith, G., Riviere, J.E., 2006b. Pharmacokinetics of flunixin meglumine in swine after intravenous dosing. J. Vet. Pharmacol. Ther. 29 (5), 437–440. Clewell, H.J., Gentry, P.R., Covington, T.R., Sarangapani, R., Teeguarden, J.G., 2004. Evaluation of the potential impact of age- and gender-specific pharmacokinetic differences on tissue dosimetry. Toxicol. Sci. 79 (2), 381–393. Craigmill, A.L., 2003. A physiologically based pharmacokinetic model for oxytetracycline residues in sheep. J. Vet. Pharmacol. Ther. 26 (1), 55–63. el-Banna, H.A., 1999. Pharmacokinetic interactions between flunixin and sulfadimidine in horses. Dtsch. Tierarztl. Wochenschr. 106 (9), 400–403. Gentry, P.R., Covington, T.R., Clewell III, H.J., Anderson, M.E., 2003. Application of a physiologically based pharmacokinetic model for reference dose and reference concentration estimation for acetone. J Toxicol. Environ. Health Part A 66 (23), 2209–2225. Grass, G.M., Sinko, P.J., 2002. Physiologically-based pharmacokinetic simulation modelling. Adv. Drug Delivery Rev. 54 (3), 433–451. Harder, S., Thurmann, P., 1996. Clinically important drug interactions with anticoagulants. An update. Clin. Pharmacokinet. 30 (6), 416–444. Kanamitsu, S., Ito, K., Green, C.E., Tyson, C.A., Shimada, N., Sugiyama, Y., 2000. Prediction of in vivo interaction between triazolam and erythromycin based on in vitro studies using human liver microsomes and recombinant human CYP3A4. Pharm. Res. 17 (4), 419–426. Kawai, R., Lemaire, M., Steimer, J.L., Bruelisauer, A., Niederberger, W., Rowland, M., 1994. Physiologically based pharmacokinetic study on a cyclosporin derivative, SDZ IMM 125. J. Pharmacokinet. Biopharm. 22 (5), 327–365. Liu, X., Smith, B.J., Chen, C., Callegari, E., Becker, S.L., Chen, X., Cianfrogna, J., Doran, A.C., Doran, S.D., Gibbs, J.P., Hosea, N., Liu, J., Nelson, F.R., Szewc, M.A., Van Deusen, J., 2005. Use of a physiologically based pharmacokinetic model to study the time to reach brain equilibrium: an experimental analysis of the role of blood-brain barrier permeability, plasma protein binding, and brain tissue binding. J. Pharmacol. Exp. Ther. 313 (3), 1254–1262. Lundeen, G., Manohar, M., Parks, C., 1983. Systemic distribution of blood flow in swine while awake and during 1.0 and 1.5 MAC isoflurane anesthesia with or without 50% nitrous oxide. Anesth. Analg. 62 (5), 499–512. Nouws, J.F., Mevius, D., Vree, T.B., Degen, M., 1989. Pharmacokinetics and renal clearance of sulfadimidine, sulfamerazine and sulfadiazine and their N4-acetyl and hydroxy metabolites in pigs. Vet. Q 11 (2), 78–86. Nouws, J.F., Vree, T.B., Baakman, M., Driessens, F., Vellenga, L., Mevius, D.J., 1986. Pharmacokinetics, renal clearance, tissue distribution, and residue aspects of sulfadimidine and its N4-acetyl metabolite in pigs. Vet. Q 8 (2), 123–135. Pond, W.G., 2001. Biology of the Domestic Pig. Cornell University Press, Ithaca, NY. Reeves, P.T., Minchin, R.F., Ilett, K.F., 1988. Induction of sulfamethazine acetylation by hydrocortisone in the rabbit. Drug Metab. Dispos. 16 (1), 110–115.

301

Reitz, R.H., Mendrala, A.L., Park, C.N., Andersen, M.E., Guengerich, F.P., 1988. Incorporation of in vitro enzyme data into the physiologically-based pharmacokinetic (PB–PK) model for methylene chloride: implications for risk assessment. Toxicol. Lett. 43 (1–3), 97–116. Riviere, J.E., 1999. Comparative Pharmacokinetics: Principles, Techniques and Applications. Blackwell Publishing, Inc., Ames, Iowa. Rowland, M., Tozer, T.N., 1995. Clinical Pharmacokinetics: Concepts and Applications. Lippincott Williams and Wilkins, Baltimore, MD. Saltvedt, I., Spigset, O., Ruths, S., Fayers, P., Kaasa, S., Sletvold, O., 2005. Patterns of drug prescription in a geriatric evaluation and management unit as compared with the general medical wards: a randomised study. Eur. J. Clin. Pharmacol. 61 (12), 921–928. Simmons, J.E., 1996. Application of physiologically based pharmacokinetic modelling to combination toxicology. Food Chem. Toxicol. 34 (11–12), 1067– 1073. Sweeney, R.W., Bardalaye, P.C., Smith, C.M., Soma, L.R., Uboh, C.E., 1993. Pharmacokinetic model for predicting sulfamethazine disposition in pigs. Am. J. Vet. Res. 54 (5), 750–754. Taber, S.W., Bergamini, T.M., 1997. Long-term venous access: indications and choice of site and catheter. Semin. Vasc. Surg. 10 (3), 130–134. Teeguarden, J.G., Waechter Jr., J.M., Clewell III, H.J., Covington, T.R., Barton, H.A., 2005. Evaluation of oral and intravenous route pharmacokinetics, plasma protein binding, and uterine tissue dose metrics of bisphenol A: a physiologically based pharmacokinetic approach. Toxicol. Sci. 85 (2), 823– 838. Toutain, P.L., Bousquet-Melou, A., 2002. Free drug fraction vs free drug concentration: a matter of frequent confusion. J. Vet. Pharmacol. Ther. 25 (6), 460–463. Tranquilli, W.J., Parks, C.M., Thurmon, J.C., Benson, G.J., Koritz, G.D., Manohar, M., Theodorakis, M.C., 1982. Organ blood flow and distribution of cardiac output in nonanesthetized swine. Am. J. Vet. Res. 43 (5), 895–897. Tsukamoto, Y., Kato, Y., Ura, M., Horii, I., Ishikawa, T., Ishitsuka, H., Sugiyama, Y., 2001. Investigation of 5-FU disposition after oral administration of capecitabine, a triple-prodrug of 5-FU, using a physiologically based pharmacokinetic model in a human cancer xenograft model: comparison of the simulated 5-FU exposures in the tumour tissue between human and xenograft model. Biopharm. Drug Dispos. 22 (1), 1–14. van der Merwe, D., Brooks, J.D., Gehring, R., Baynes, R.E., Monteiro-Riviere, N.A., Riviere, J.E., 2006. A physiologically based pharmacokinetic model of organophosphate dermal absorption. Toxicol. Sci. 89 (1), 188–204. Wilkinson, G.R., 2001. The dynamics of drug absorption, distribution, and elimination. In: Limbird, H.A. (Ed.), The Dynamics of Drug Absorption, Distribution, and Elimination, 10th ed. McGraw-Hill, New York, pp. 3–27 Young, J.F., Wosilait, W.D., Luecke, R.H., 2001. Analysis of methylmercury disposition in humans utilizing a PBPK model and animal pharmacokinetic data. J. Toxicol. Environ. Health A 63 (1), 19–52. Yuan, Z.H., Miao, X.Q., Yin, Y.H., 1997. Pharmacokinetics of ampicillin and sulfadimidine in pigs infected experimentally with Streptococcus suum. J. Vet. Pharmacol. Ther. 20 (4), 318–322.