Page 1

CLINICAL PHARMACOKINETICS

ELECTRONIC SUPPLEMENTARY MATERIAL 1

PharmacokineticDrug Interaction Studies with Enzalutamide

Jacqueline A. Gibbons*· Michiel de Vries · Walter Krauwinkel · Yoshiaki Ohtsu · Jan Noukens· JanStefan van der Walt · Roelof Mol · Joyce Mordenti · Taoufik Ouatas

Correspondence to: J. A. Gibbons [
Medivation, Inc., San Francisco, CA, USA

POPULATION PHARMACOKINETIC MODELING AND SIMULATION TO ASSESS THE EFFECT OF GEMFIBROZIL ON ENZALUTAMIDE AND ITS MAJOR METABOLITES

Pharmacokinetic analyses were performed based on population pharmacokinetic modeling and simulation as part of the drug interaction study with cytochrome P450 (CYP) 2C8 and CYP3A4 inhibitors (NCT01913379). The methods were similar to those described previously for drug interaction studies with long half-life drugs [1]. The purpose of the analyses was to estimate the full extent of the drug interaction between gemfibrozil (strong CYP2C8 inhibitor) and enzalutamide.

Pharmacokinetic models were developed using combined plasma concentration-time data for enzalutamide, Ndesmethyl enzalutamide, and the carboxylic acid metabolite from the three treatment arms in the drug interaction study (NCT01913379) and from a 14C mass balance and biotransformation study in six healthy male subjects, which also provided urine concentration-time data (NCT01911715 [2]).

Two pharmacokinetic models were developed (Fig. S1):

  1. Model 1, which describes the pharmacokinetics of enzalutamide.
  2. Model 2, which describes the combined pharmacokinetics of enzalutamide, Ndesmethyl enzalutamide, and the carboxylic acid metabolite.

In Model 1, the effects of gemfibrozil and itraconazole were described by assuming an inhibitory effect on the apparent clearance of enzalutamide. In Model 2, the effects of gemfibrozil and itraconazole were described by assuming an inhibitory effect on the clearance of enzalutamide to N-desmethyl enzalutamide, enzalutamide to carboxylic acid metabolite, enzalutamide to unmeasured metabolites, and N-desmethyl enzalutamide to carboxylic acid metabolite.

Model 1 was a two-compartmental model with first-order absorption with lag time and linear clearance from the central compartment (Table S1). A separate absorption rate constant was implemented for the itraconazole arm (arm 3). The itraconazole effect on the bioavailable fraction (F) was implemented for stability reasons; F was set to 1 for arm 1. Body weight was used as an allometric covariate on both clearance and volume of peripheral compartment (V3). Inhibition for each of the two inhibitors (gemfibrozil and itraconazole) was implemented using a multiplicative factor for enzalutamide clearance. At the end of the inhibitory period, clearance was switched back to the uninhibited clearance. As the point in time was not estimable for arm 3 (itraconazole), the same end of inhibition time was estimated for arm 2 and arm 3. Interindividual variability (IIV) was added to the following parameters: absorption lag time, absorption rate constant, volume of central compartment, volume of peripheral compartment, clearance, intercompartmental clearance, and inhibition factor. For each parameter, variability was implemented in a logarithmic-normal manner. The model has a combined proportional and additive residual error structure. That Model1 adequately described the data is demonstrated bythe goodness-of-fit diagnostic plots (Fig. S2). The adequacy of the model is further supported by visual predictive checks (VPCs), which showed that the model characterized the time-course of the effect of the concomitant inhibitors during the absorption phase (Fig.S3 top panel), as well as after the concomitant inhibitors were stopped(Fig. S3, bottom panel). All structural parameters could be estimated with high precision (relative standard error [RSE] <21%), except for the allometric coefficients for which RSE was moderate (30%–37%). IIV parameters were estimated with moderate to high precision (RSE <50%), except for the IIV on the absorption rate constant, which was 89%. Precision of the residual variability term was acceptable (13% for the proportional component).

The uninhibited enzalutamide apparent clearance was estimated at 0.505 L/h. During gemfibrozil coadministration, apparent clearance was changed by a factor 0.235 and relative bioavailability was changed by a factor 0.934, reducing the apparent clearance to 0.111 L/h. During itraconazole coadministration, apparent clearance was changed by a factor 0.799, reducing the apparent clearance to 0.403 L/h. The end of inhibition was estimated as a model event time (MTIME in NONMEM software) [3] and defines the time at which the effect of the inhibitor on enzalutamide clearance was switched off. The estimate of 477 h indicatedthat inhibition continued for 57 h after last gemfibrozil dose and 69h after last itraconazole dose. The extended inhibitory effect of gemfibrozil may be explained by the action of gemfibrozil as a mechanism-based inhibitor, which requires synthesis of new CYP2C8 enzyme in order for CYP2C8 activity to recover. Itraconazole is a reversible inhibitor that has a half-life of about 20 h.

Model 1 was used as the base structure of Model 2, including the covariate structure. Model 2 was subsequently expanded by including the structure for the metabolites and the various inhibitory effects. The metabolic pathways for enzalutamide were based on in vitrodata. Although enzalutamide is known to be metabolized to a precursor that undergoes spontaneous and rapid degradation to Ndesmethyl enzalutamide, the pharmacokinetic model assumed direct formation of N-desmethyl enzalutamide from enzalutamide. Consistent with in vitro data, Ndesmethyl enzalutamide was assumed to be metabolized to the carboxylic acid metabolite and other metabolites. Based on data from a human mass balance study, the majority of enzalutamide (62.7%) is excreted in urine as the carboxylic acid metabolite, while excretion of Ndesmethyl enzalutamide in feces and urine is negligible [2]. Therefore, the apparent clearance of enzalutamide (CLP) was separated into three metabolic pathways:

  1. Clearance to N-desmethyl enzalutamide (CLPM2);
  2. Clearance to carboxylic acid metabolite (CLPM1);
  3. Clearance to unmeasured metabolites (CLPOTH).

Thus,

/ (1)

The fractions of enzalutamide eliminated via formation of N-desmethyl enzalutamide (FM2), the carboxylic acid metabolite (FM1), and unmeasured metabolites (FMOTH) were expressed as the ratio of the respective clearances compared with CLP:

/ (2)
/ (3)
/ (4)

The carboxylic acid metabolite forms by clearance of N-desmethyl enzalutamide (CLM2) and enzalutamide (CLPM1) and is excreted to urine (CLM1U) or eliminated via other routes (CLM1OTH).

To improve the ability to identify the effects of gemfibrozil and itraconazole on the clearance parameters, the following assumptions were made:

  • Clearance of N-desmethyl enzalutamide to unmeasured metabolites was considered negligible and all N-desmethyl enzalutamide was metabolized to the carboxylic acid metabolite;
  • The amount of enzalutamide excreted in urine as the carboxylic acid metabolite was fixed to 62.7% of the administered dose [2];
  • The observed urine volumes were used to predict the carboxylic acid metabolite concentrations in urine;
  • As the molecular masses of N-desmethyl enzalutamide and the carboxylic acid metabolite are about 97% of enzalutamide, differences in the molecular weight of enzalutamide, N-desmethyl enzalutamide, and the carboxylic acid metabolite were disregarded;
  • All clearances were scaled using a bodyweight of 70 kg as a reference with a fixed allometric coefficient of 0.75.

That Model 2 adequately describes the data is evidenced by diagnostic plots (Fig. S4 and Fig. S5), which show good agreement between the observed versus individual predicted plasma concentrations of enzalutamide, Ndesmethyl enzalutamide, and the carboxylic acid metabolite.

Parameters for Model 2 are summarized in Table S2. Enzalutamide pharmacokinetics wereadequately described by Model 2, but residual error favored Model 1. For this reason, Model 1 was used to simulate enzalutamide, and Model 2 was used for the metabolite simulations. The adequacy of Model 2 for characterizing the time-course of the concomitant inhibitors on N-desmethyl enzalutamide and the carboxylic acid metabolite(and especially in arm 2) is demonstrated by the prediction-corrected VPCs (Fig.S6).

Model 1 and Model 2 were used to assess the effects of gemfibrozil on the subjects participating in the study. This involved simulated concentration-time profiles for enzalutamide alone and enzalutamide combined with gemfibrozil for each of the 41 individual subjects in the study. When modeling concomitant gemfibrozil, it was assumed that gemfibrozil remained at steady-state during the entire pharmacokinetic sampling period for enzalutamide and metabolites. Model 1 was used to simulate enzalutamide concentration-time data, Model 2 was used for N-desmethyl enzalutamide and the carboxylic acid metabolite data, and the results of Models 1 and 2 were combined for the sum of enzalutamide plus Ndesmethyl enzalutamide. The simulated concentration-time data were analyzed by noncompartmental (NCA) methods in WinNonlin® (Pharsight Corp., Palo Alto, CA, USA).

Based on the geometric mean ratio (GMR) for enzalutamide combined with gemfibrozil (test) to enzalutamide alone (reference), gemfibrozil caused the area under the plasma concentration-time curve from time zero to infinity for enzalutamide to increase 4.26-fold (90% confidence interval [CI], 3.59–5.05), N-desmethyl enzalutamide to decrease by 25% (GMR, 0.75; 90% CI, 0.64–0.87), the carboxylic acid metabolite to increase by 2.70-fold (90% CI, 2.24–3.26), and the sum of enzalutamide plus N-desmethyl enzalutamide to increase by 2.17-fold (90% CI, 1.91–2.47).

Simulated concentration-time profiles were additionally generated at the population level, which accounted for IIV as defined by Models 1 and 2. Demographics data were used for the 41 subjects in the study and the data were simulated 77 times, thereby generating simulations for at least 1,000 subjects per treatment. The mean ratios (enzalutamide combined with itraconazole or gemfibrozil [test]/enzalutamide alone [reference]) were similar to the GMRs based on individual simulations for each of the 41 subjects (Table S3).

To further assess the performance of the modeling and simulation for quantifying drug interaction effects, the GMRs of the simulations for the itraconazole arm (arm 3) were compared with the GMRs from the previous NCA analysis based on observed data(Table S3). The results with the simulated and observed data were essentially the same for all molecules except for the carboxylic acid metabolite, for which the GMR was 18% higher for the simulated data than for the observed data. As the carboxylic acid metabolite does not contribute to clinical effects of enzalutamide, this discrepancy is considered acceptable.

Table S1Model 1 parameter estimates

Parameter (unit) [abbreviationa] / NONMEM 7.2 estimates
Typical
values / IIV,
% / 95% confidence
intervalsb
Absorption lag time, h / 0.417 / 24.5 / 0.391–0.446
Absorption rate constant [ka],h-1 / 0.669 / 8.8 / 0.603–0.758
Clearance [CL],L/h / 0.505 / 21.6 / 0.471–0.540
Intercompartmental clearance [Q], L/h / 15.5 / 14.0 / 14.3–16.8
Central volume of distribution [V2],L / 3.52 / 124.0 / 2.29–5.15
Peripheral volume of distribution [V3], L / 56.5 / 21.0 / 52.4–60.9
Effect of gemfibrozil on enzalutamide
Inhibition factor for clearance / 0.235 / 13.5 / 0.209–0.260
Relative bioavailability [F] / 0.934 / NA / 0.855–1.02
Effect of itraconazole on enzalutamide
Inhibition factor for clearance / 0.799 / 13.5 / 0.753–0.845
Relative bioavailability [F] / 1.00 / NA / 0.905–1.12
Factor on absorption rate constant / 1.17 / NA / 1.01–1.33
End of inhibition, hc / 477 / NA / 473–483
Allometric coefficients for weight
Peripheral volume of distribution / 1.47 / NA / 0.667–2.39
Clearance / 1.32 / NA / 0.413–2.24
Proportional residual variability / 0.00585 / NA / 0.005–0.007
Additive residual variability, µg/mL / 9.87e-6 / NA / 1.0e-9–2.64e-5

IIV interindividual variability, NA not applicable

aAbbreviations for parameters used in model schematic (Fig. S1a)

bThe 95% confidence intervals for the typical values were obtained from nonparametric bootstrapping

cThe prolonged inhibitory after discontinuation effect of gemfibrozil and itraconazole dosing was estimated as time since the dose of enzalutamide

Table S2Model 2 parameter estimates

Parameter [abbreviationa] / NONMEM 7.2 estimates
Typical
values / IIV,
% / 95% confidence intervalsb
Enzalutamide
Absorption lag time, h / 0.391 / 38.7 / 0.284–0.422
Absorption rate constant [ka], h-1 / 0.634 / 8.8 / 0.586–0.742
Clearance to other metabolites [CLPOTH], L/h / 0.116 / 42.2 / 0.0767–0.175
Clearance to N-desmethyl enzalutamide[CLPM2], L/h / 0.332 / 24.8 / 0.294–0.391
Clearance to carboxylic acid metabolite[CLPM1],L/h / 0.0242 / 25.1 / 0.0199–0.0313
Intercompartmental clearance [Q], L/h / 14.5 / 15.1 / 13.8–16.3
Central volume of distribution [V2],L / 3.80 / 118.9 / 2.68–4.91
Peripheral volume of distribution [V3], L / 57.9 / 20.6 / 54.3–62.8
Carboxylic acid metabolite
Clearance [CLM1],L/h / 3.35 / 30.0 / 2.88–4.01
Volume of distribution [V6],L / 1.33 / NA / 1.16–1.84
N-desmethyl enzalutamide
Clearance to carboxylic acid metabolite[CLM2],L/h / 0.235 / 17.3 / 0.202–0.274
Volume of distribution [V4],L / 53.0 / NA / 45.9–61.8
Inhibitory factors for gemfibrozil
Enzalutamide clearance to other metabolites / 0.480 / NA / 0.368–0.725
Enzalutamide clearance to carboxylic acid metabolite / 0.802 / NA / 0.673–0.970
Enzalutamide clearance to N-desmethyl enzalutamide / 0.132 / NA / 0.117–0.148
N-desmethyl enzalutamide clearance to carboxylic acid metabolite / 0.773 / NA / 0.637–0.864
Inhibitory factors for itraconazole
Enzalutamide clearance to other metabolites / 1.18 / NA / 0.858–1.37
Enzalutamide clearance to carboxylic acid metabolite / 0.729 / NA / 0.592–0.900
Enzalutamide clearance to N-desmethyl enzalutamide / 0.737 / NA / 0.690–0.889
Enzalutamide absorption rate constant / 0.136 / NA / -0.0389–0.275
N-desmethyl enzalutamide clearance to carboxylic acid metabolite / 0.889 / NA / 0.831–0.939
End of inhibition effect on enzalutamide, hc / 469 / NA / 466–475
End of inhibition effect on N-desmethyl enzalutamide, hc / 474 / NA / 472–499
Residual variability
Proportional enzalutamide / 0.0809 / NA / 0.0712–0.0899
Proportional carboxylic acid metabolite / 0.252 / NA / 0.234–0.265
Proportional N-desmethyl enzalutamide / 0.0751 / NA / 0.0636–0.0848
Additive enzalutamide,µg/mL / 0.00496 / NA / 0.0027–0.0082
Additive carboxylic acid metabolite,µg/mL / 0.00562 / NA / 0.0039–0.0073
Additive N-desmethyl enzalutamide,µg/mL / 0.0185 / NA / 0.016–0.0206

IIV interindividual variability,NA not applicable

aAbbreviations used for parameters in model schematic (Fig. S1b)

bThe 95% confidence intervals for the typical values are presented and were obtained from the nonparametric bootstrap evaluation

cThe prolonged inhibitory after discontinuation effect of gemfibrozil and itraconazole dosing was estimated as time since the dose of enzalutamide

Table S3Comparison of methods for estimating drug interactions with strong CYP2C8 and CYP3A4 inhibitors in healthy male subjects in study NCT01911715

Molecule / AUC geometric mean ratio for test/reference
Strong CYP2C8 inhibitor (gemfibrozil) / Strong CYP3A4 inhibitor (itraconazole)
Individual subjectsa / Simulated populationb / Noncompartmental analysisc / Modeling and simulationd
Enzalutamide / 4.26 / 4.04 / 1.41 / 1.45
N-desmethyl enzalutamide / 0.75 / 0.67 / 1.21 / 0.99
Carboxylic acid metabolite / 2.70 / 2.82 / 1.06 / 1.25
AUCarea underthe plasma concentration-time curve extrapolated to infinity, Cmax maximum plasma concentration, CYP cytochrome P450
a / Simulated concentration-time profiles for enzalutamide alone and enzalutamide combined with gemfibrozil were generated for each of the 41 individual subjects in the study
b / Simulated concentration-time profiles were generated at the population level. Demographics data were used for the 41 subjects in the study and the data were simulated 77 times, thereby generating at least 1,000 subjects per treatment
c / Concentration-time profiles for enzalutamide alone (n=14 subjects) and enzalutamide combined with itraconazole(n=13 subjects) were described by noncompartmental methods for subjects in the study
d / Simulated concentration-time profiles for enzalutamide alone and enzalutamide combined with itraconazole were generated for each of the 41 individual subjects in the study

Fig. S1Schematic for the pharmacokinetic models for enzalutamide and its major human metabolites. (a) Model 1 was used to describe the pharmacokinetics of enzalutamide; (b)Model2 was used to describe the combined pharmacokinetics of enzalutamide, Ndesmethyl enzalutamide, and the carboxylic acid metabolite. The microrate constants were parameterized in terms of clearance and volume terms (see Table S1 for Model 1, and Table S2 for Model 2) to estimate the effects of gemfibrozil and itraconazole on the clearance terms (kout= CL/V2 [for Model 1], kout= CLPOTH/V2 [for Model 2], k23= Q/V2, k32= Q/V3, k24= CLPM2/V2, k26 = CLPM1/V4, k46 = CLM2/V4, and k60 = CLM1/V6).

CLM1 clearance of carboxylic acid metabolite,CLM2 clearance of N-desmethyl enzalutamide to carboxylic acid metabolite,CLPM1 clearance of enzalutamide to carboxylic acid metabolite,CLPM2 clearance of enzalutamide to N-desmethyl enzalutamide,ka absorption rate constant,k23 and k32 distribution rate constants for intercompartmental transfer of enzalutamide between the central and peripheral compartments,k24, k26, and k46 biotransformation rate constants for enzalutamide to N-desmethyl enzalutamide, enzalutamide to the carboxylic acid metabolite, and N-desmethyl enzalutamide to the carboxylic acid metabolite, respectively,k60 elimination rate constant for the carboxylic acid metabolite,kout biotransformation rate constant for enzalutamide to unmeasured metabolites,Q intercompartmental clearance,Vx volume of distribution associated with the relevant compartment

Fig. S2 Goodness-of-fit plots for Model 1, enzalutamide. (a) Observed versus population predictions of plasma concentrations (g/mL); (b) observed plasma concentrations (g/mL) versus individual predictions of plasma concentrations (g/mL); (c) individual weighted residuals (iWRES) vs individual predictions of plasma concentrations (g/mL); and (d) conditional weighted residuals versus the time after dose (hours).

Fig. S3Prediction-corrected visual predictive checks for Model 1, enzalutamide. The circles represent observed plasma concentration data, the solid lines represent the median of the observed plasma concentrations, the dashed lines represent the 5th and 95th percentiles of the observed plasma concentrations, the darkgrey shaded area represents the 95% confidence interval for the 50th percentile (median) of 1000 simulations, and the lower and upper light grey shaded areas represent the 95% confidence interval for the 5th and 95th percentiles from the 1000 simulations, respectively. (a) First 10 hours postdose; (b)10–1176h postdose.

Fig. S4 Goodness-of-fit plots for Model 2, all analytes (enzalutamide, N-desmethyl enzalutamide, carboxylic acid metabolite).(a) Observed versus population predictions of plasma concentrations (g/mL); (b) observed plasma concentrations (g/mL) versus individual predictions of plasma concentrations (g/mL); (c) individual weighted residuals (iWRES) vs individual predictions of plasma concentrations (g/mL); and (d) conditional weighted residuals versus the time after dose (hours).

Fig. S5 Plots of the observed versus individual predicted concentrations for model observed versus individual predictions of plasma concentrations of (a) enzalutamide; (b) N-desmethyl enzalutamide; and (c) carboxylic acid metabolite.

Fig. S6 Prediction-corrected visual predictive checks for Model 2, N-desmethyl enzalutamide and carboxylic acid metabolite. The circles represent observed plasma concentration data, the solid lines represent the median of the observed plasma concentrations, the dashed lines represent the 5th and 95th percentiles of the observed N-desmethyl enzalutamide plasma concentrations, the dark grey shaded area represents the 95% confidence interval for the 50th percentile (median) of 1000 simulations, and the lower and upper light grey shaded areas represent the 95% confidence interval for the 5th and 95th percentiles from the 1000 simulations, respectively. (a) N-desmethyl enzalutamide; (b) carboxylic acid metabolite.

References

  1. Lehr T, Staab A, Trommeshauser D, Schaefer HG, Kloft C. Semi-mechanistic population pharmacokinetic drug-drug interaction modelling of a long half-life substrate and itraconazole. Clin Pharmacokinet. 2010;49(1):53–66.
  2. Gibbons JA, Ouatas T, Krauwinkel W, Ohtsu Y, van der Walt JS, Beddo V, et al. Clinical pharmacokinetics of enzalutamide. Clin Pharmacokinet. 2015 [Epub ahead of print].
  3. Boeckmann AJ, Sheiner LB,Beal SL. NONMEM Users Guide - Part VIII. Ellicott City, MD, USA: ICON Development Solutions; 2010; p. 317. Available at ftp://nonmem.iconplc.com/Public/nonmem712/VIII.pdf.Accessed 15 April 2015.