Original Article
A Comparison of the Pharmacokinetic Profile of an Ascending-Dose, ExtendedRegimen Combined Oral Contraceptive to Those of Other Extended Regimens
Reproductive Sciences 2014, Vol. 21(11) 1401-1410 ª The Author(s) 2014 Reprints and permission: sagepub.com/journalsPermissions.nav DOI: 10.1177/1933719114526472 rs.sagepub.com
Mona Darwish, PhD1, Mary Bond, MS, MBA2, Nancy Ricciotti, MSN3, Jennifer Hsieh, MS4, Jill Fiedler-Kelly, MS5, and Thaddeus Grasela, PharmD, PhD5
Abstract Quartette (levonorgestrel [LNG]/ethinyl estradiol [EE] and EE) is an ascending-dose, extended-regimen combined oral contraceptive (COC) that consists of a constant dose of LNG 150 mg on days 1 to 84 with EE 20 mg on days 1 to 42, 25 mg on days 43 to 63, 30 mg on days 64 to 84, and 10 mg of EE monotherapy on days 85 to 91. A population pharmacokinetic (PK) model for EE was developed using nonlinear mixed-effects modeling to characterize the PK profile of EE administered in Quartette and other extended-regimen LNG/EE COCs. Model-predicted plasma concentration–time profiles demonstrated a stepwise increase in systemic exposure to EE during the first 84 days of the cycle following each EE dose change. Lower concentrations of EE were noted during the final 7-day period of EE 10 mg. Gradual increases in EE seen with Quartette may decrease the incidence of unscheduled bleeding frequently observed during early cycles of extended-regimen COCs. Keywords combined oral contraceptive, ascending-dose, extended regimen, bleeding, ethinyl estradiol
Introduction Extended regimens of combined hormonal oral contraceptives (COCs) are an increasingly popular option for women seeking contraception because they reduce the frequency of scheduled withdrawal bleeding compared to traditional 28-day COCs.1,2 However, unscheduled bleeding and spotting with extendedregimen COCs is common in early COC cycles3-5 and may contribute to poor adherence and early treatment discontinuation.6 Two strategies for stabilizing the endometrium and reducing unscheduled bleeding include gradually increasing estrogen doses during the COC cycle and using a low-dose ethinyl estradiol (EE) during the traditional hormone-free interval.5,7-9 Quartette (levonorgestrel [LNG]/EE and EE) is a novel ascending-dose, extended-regimen COC that has been approved by the US Food and Drug Administration. The regimen consists of a continuous LNG dose combined with a gradually increasing EE dose for 84 days followed by a low dose of EE for 7 days. This design is intended to provide an increasing ratio of estrogen to progestin across the extended cycle, with the ratio increasing just prior to the times during the cycle when unscheduled bleeding most frequently occurs.9 The pharmacokinetics (PK) of LNG and EE are well studied, and while Quartette (Teva Women’s Health, Inc., Sellersville, Pennsylvania) uses LNG doses comparable to those in
other extended-regimen COCs, the PK of EE in the context of Quartette has not been assessed. Because LNG dosages (100 mg or 150 mg) and exposure are constant during the first 84 days of the extended regimens included in this report, this analysis focuses on the PK of EE. The objectives of this analysis, which used a population modeling approach, were to characterize the overall PK profile of EE after administration of Quartette, including its relative bioavailability, dose proportionality, and single- and multiple-dose EE PK. This analysis also explored the effects of selected covariates, such as weight, body mass index (BMI), age, race, and smoking status as
1
Clinical Pharmacology, SCI-MED BRIDGE, Malvern, PA, USA Phase 1 & Clinical Pharmacology, Teva Branded Pharmaceutical Products, R&D, Inc., Frazer, PA, USA 3 Teva Women’s Health, Teva Branded Pharmaceutical Products, R&D, Inc., Frazer, PA, USA 4 Global Biostatistics, Teva Branded Pharmaceutical Products, R&D, Inc., West Chester, PA, USA 5 Pharmacometric Services, Cognigen Corporation, Buffalo, NY, USA 2
Corresponding Author: Mary Bond, Phase I & Clinical Pharmacology, Teva Branded Pharmaceutical Products, R&D, Inc., 41 Moores Road, Frazer, PA 19355, USA. Email: [email protected]
1402
Reproductive Sciences 21(11)
Table 1. Description of Included Studies. Subjects With Usable Subjects EE-Concentration Data, na Enrolled, n
Dosing Regimen
Quartette
Days 1-42: LNG 150 mg/ Predose (0) and at 0.5, 1, 1.25, 1.5, 1.75, 2, 2.5, 3, 4, 6, 8, 10, 12, 16, 24, 36, 48, EE 20 mg 72, and 96 hours after dose Days 43-63: LNG 150 mg/ EE 25 mg Days 64-84: LNG 150 mg/ EE 30 mg Days 85-91: EE 10 mg Days 1-84: LNG 150 mg/ Predose (0) and at 0.5, 1.0, 1.25, 1.5, 1.75, 2, 2.5, 3, 4, 6, 8, 10, 12, 16, 24, EE 30 mg 36, 48, 72, and 96 hours after dose Days 85-91: placebo
18
17
Healthy, nontobaccousing women with normal menstrual cycle; 18- to 45-year-old; BMI 18-30 kg/m2
30
29
Days 1-84: LNG 150 mg/ Days 1, 21: predose (0) and at 0.5, 1, 1.33, 1.67, 2, 3, 4, 6, 8, 11, 15, and 24 EE 30 mg hours after dose Days 85-91: EE 10 mg Day 84: predose (0) and at 0.5, 1, 1.33, 1.67, 2, 3, 4, 6, 8, 11, 15, 24, 36, 48, 72, 96, 120, and 144 hours after dose Day 91: predose (0) and at 0.5, 1, 1.33, 1.67, 2, 3, 4, 6, 8, 11, 15, 24, 36, 48, 72, and 96 hours after dose Days 1-84: LNG 150 mg/ Predose (0) and at 0.33, 0.67, 1, 1.33, 1.67, 2, 2.5, 3, 4, 6, 8, 12, 16, 24, 36, EE 30 mg 48, and 72 hours after dose (EE and Days 85-91: EE 10 mg LNG) 96 and 120 hours after dose (LNG only)
30
30
Healthy women; 18- to 35-year-old; Within 15% of their ideal body weight Healthy women with normal menstrual cycle; 19- to 51-year-old; 119-191 lb
30
29
1-84: LNG 150 mg/ Predose (0) and at 0.33, 0.67, 1, 1.33, 1.67, 2, 2.5, 3, 4, 6, 8, 10, 12, 16, 24, 30 mg 36, 48, 72, and 96 hours after dose 85-91: EE 10 mg 1-84: LNG 100 mg/ Predose (0) and at 0.5, 1.0, 1.25, 1.5, 1.75, 2, 2.5, 3, 4, 6, 8, 10, 12, 16, 24, 20 mg 36, 48, 72, and 96 hours after dose 85-91: EE 10 mg
18
17
35
30
Seasonale
Seasonique Study 1b
Seasonique study 2
Seasonique study 3
Days EE Days LoSeasonique Days EE Days
PK Sampling
Subject Characteristics
Study
Healthy, nontobaccousing women with normal menstrual cycle; 19- to 51-year-old; BMI 18-30 kg/m2 Healthy women; 18- to 47-year-old; 53-87 kg Healthy women with regular menstrual cycle; 18- to 35-year-old; 48-75 kg
Abbreviations: EE, ethinyl estradiol; LNG, levonorgestrel; PK, pharmacokinetics; BMI, body mass index. a All studies assessed single-dose PK with the exception of Seasonique study 1. b Seasonique study 1 included both multiple-dose and single-dose assessments.
predictors of interindividual variability (IIV) in EE disposition; evaluated the performance of the developed PK model; and simulated the plasma concentration–time course for EE exposure over two 91-day cycles for Quartette and 2 other extended-regimen COCs (Seasonique and LoSeasonique) in order to compare estimates of EE exposure over various days within the 91-day cycle.
Design and Methodology A PK model for EE was developed using nonlinear mixedeffects modeling to characterize the PK of EE administered in various dosing regimens and formulations used in Quartette and other extended-regimen LNG/EE COCs. The PK data used to develop this model were obtained from 5 single-dose studies 1402
and 1 multiple-dose study of several extended-regimen COC products containing LNG and EE (Table 1).
Subjects, Drug Administration, and Blood Sampling Participants in each of the 6 studies were healthy women between the age of 18 and 51 years who had regular menstrual cycles. Ethinyl estradiol was administered as described in Table 1. In these studies, methods used to measure plasma EE concentrations included liquid chromatography–tandem mass spectrometry (with a lower limit of quantification [LLOQ] of 1-2 pg/mL) and gas chromatography–mass spectrometry–negative chemical ionization (with an LLOQ of 5.07 pg/mL). In each study, extensive PK sampling was performed, including samples at predose and up to 144 hours after the drug was first
Darwish et al
1403
Figure 1. Flowchart of model development process.
administered as detailed in Table 1. In the multiple-dose study, trough PK samples were collected prior to dosing on days 18, 19, 20, 81, 82, and 83.
Population PK Analysis The model was used to characterize the PK profile, assess the bioavailability and dose proportionality of EE, detect possible differences in EE PK between single- and multiple-dose regimens, and evaluate the impact of selected covariates as predictors of IIV in EE disposition. Figure 1 graphically depicts the general procedure used to develop the population PK model.
The population PK analysis was completed using nonlinear mixed-effects modeling software (NONMEM, version 6, level 2.0; ICON Development Solutions 2006, Hanover, Maryland). The first-order conditional estimation with interaction estimation method was used throughout the model development.
Exploratory Graphical Analysis Exploratory graphical analyses and data visualization techniques were used to guide the selection of a base model for the population PK analysis. The key objective of these analyses was to confirm whether the models to be tested were 1403
1404
Reproductive Sciences 21(11)
Table 2. Demographic Characteristics for the Population PK Analysis Data Set. Subject Characteristics
Overall Values
N Age, y Mean (SD) Median (range) Body mass index, kg/m2 Mean (SD) Median (range) Weight, kg Mean (SD) Median (range) Race, n (%) Caucasian Black Hispanic Other Smoker, n (%) No Yes
152 28.3 (7.7) 28.0 (18.0-51.0) 20.9 (3.6) 19.9 (15.5-29.8) 65.9 (9.4) 65.1 (48.2-88.6) 80 (52.6) 33 (21.7) 36 (23.7) 3 (2.0) 137 (90.1) 15 (9.9)
Abbreviations: PK, pharmacokinetics; SD, standard deviation.
appropriate and to verify the assumptions of the model. This exploratory analysis was also used to search for extreme values and potential outliers, assess possible trends in the data, and determine whether any errors were made in the handling of the data or the creation of the analysis data sets.
Selection of Base Structural PK Model Results of the exploratory analyses were used to determine the appropriate functional form of the base structural model of drug concentration versus time data. Previous literature10,11 and preliminary examination of plasma EE concentration profiles indicated that a linear, 2-compartment, open model with first-order absorption and elimination would adequately describe the data. Various structural models were applied and their capacities to characterize the population PK of EE were assessed. Each population PK model was described by estimating mean structural model parameters (eg, absorption rate constant [ka], bioavailability fraction [F], apparent volumes of distribution in central and peripheral compartments [Vc/F and Vp/F], apparent oral clearance [CL/F], and apparent intercompartmental clearance [Q/F]), the magnitude of IIV, and residual variability (RV). The IIV was estimated on CL/F, Vc/F, Q/F, Vp/F, ka, F, and duration of zero-order input into absorption (depot) compartment (D1) using exponential error models. The RV was estimated using a combined additive and constant coefficient of variation error model. Given that all data used in these analyses were obtained after oral drug administration, the F was assumed to be 100% and the PK parameters were considered apparent values. The impact of the extended-regimen COC (Quartette vs Seasonique vs Seasonale vs LoSeasonique) was tested on the relative F. The effect of the comparators was evaluated as a shift in the relative F compared to Quartette. 1404
Covariate Analysis Covariate analyses were performed to explore the influence of selected factors on the magnitude of IIV and RV in EE PK. These analyses were performed using a forward selection followed by a backward elimination process (Figure 1). Covariates were evaluated for their ability to explain IIV in CL/F, Vc/F, and Vp/F and included race, age, weight, BMI, and smoking. The covariate models were developed using graphical and statistical approaches to define the mathematical forms of the relationships in question and assess their statistical significance. The correlation between covariates was examined to avoid potential multicollinearity or confounding of effects in covariate submodels. If a covariate was highly correlated with another covariate (eg, body weight and BMI), only 1 of the highly correlated covariates was evaluated. Linear, exponential, power, additive, and proportional shift models were used to evaluate continuous and categorical covariates in NONMEM, as appropriate.
Population PK Model Refinement and Evaluation The reduced multivariable model, including all significant covariates, was evaluated for any remaining biases in the IIV and RV error models. A simulation-based, prediction-corrected, visual predictive check method was used to evaluate the adequacy of the final model.12 The final model was used to simulate 1000 replicates of the analysis data set with NONMEM. Statistics of interest (eg, the 5th, 50th, and 95th percentiles of the distributions of concentration) were calculated from the simulated and observed data for comparison. These percentiles were then plotted versus time to enable a visual assessment of concordance between the modelbased simulated data and the original observed data set and/or percentiles based on the observed data. Model-predicted PK parameters were calculated using the final population PK model. For subjects receiving Quartette, model-predicted values were then compared with values for area under the curve (AUC), maximum observed plasma drug concentration (Cmax), time to maximum observed concentration (tmax), and half-life calculated using noncompartmental analyses of the observed EE concentrations over time. The previous noncompartmental analysis was performed using WinNonlin (Pharsight Corporation, Mountain View, California) and the linear trapezoidal rule. The EE AUC from time zero to infinity (AUC0-1) was calculated using individual modelpredicted EE concentrations for each study day in which a full profile was obtained. Cmax, tmax, and half-life were also estimated.
Simulations Using Final Population PK Model The population PK model was used to predict EE concentration–time profiles at various days during the 91-day cycle for women enrolled in Phase 3 clinical studies of the extended regimens included in this analysis. Ethinyl estradiol concentrations
Darwish et al
1405
Figure 2. Goodness-of-fit scatterplots for the final population PK model of EE. EE indicates ethinyl estradiol; PK, pharmacokinetics.
were simulated over two 91-day cycles for Quartette, Seasonique, and LoSeasonique based on the dosing regimens used for each product (Table 1).
Results Demographics and Baseline Characteristics The population PK analysis included 6433 plasma EE concentrations from 152 female subjects of childbearing age (Table 2). On average, subjects were aged 28.3 (+7.7) years and weighed 65.9 (+9.4) kg. Mean BMI was 20.9 (+3.6) kg/m2 and 9.9% were smokers.
Between 20 and 55 samples were obtained from the majority of subjects. A small percentage of samples obtained more than 50 hours after the last dose was associated with measurable concentrations (n ¼ 106, 1.6%). These concentrations were considered to be extreme values and due to their potential to unduly influence the model, they were excluded from the analysis data set. After these data were excluded, the final analysis data set included 6327 concentration values from 152 subjects.
Pharmacokinetic Data Analysis Ethinyl estradiol concentrations declined from peak in a biphasic manner. Initially, a 2-compartment, open model with a 1405
1406 combination of zero-order input into a depot compartment, first-order absorption into the central compartment, and firstorder elimination was fit to the data. The analysis data set consisted of PK data from 4 products, and potential differences in F between these products were evaluated during the development of the base model. In the base structural model, the bioavailability of Seasonique and Seasonale was determined to be 119% relative to Quartette and LoSeasonique (F ¼ 1). All parameters were estimated with good precision (standard error of the mean expressed as a percentage [%SEM] < 20%), with the exception of the IIV for ka, which was estimated with slightly poorer precision (%SEM ¼ 31.2%). Goodness-offit plots were unbiased and demonstrated the adequacy of the base model. The effect of EE dose on clearance was evaluated using the base structural model; however, no evidence for a lack of dose proportionality was detected. A covariate analysis was performed to explore the sources of variability in EE PK. The effects of multiple dosing and demographic covariates (multiple dosing, race, age, weight, BMI, and smoking status) on CL/F, Vc/F, and Vp/F were evaluated. After 4 rounds of forward selection, 3 statistically significant predictors of variability were identified. Multiple dosing was a statistically significant predictor of Vc/F and CL/F, and weight was a statistically significant predictor of Vp/F (P < .05). Univariate stepwise backward elimination proceeded after all adjustments had been made to the IIV and RV error models. None of the demographic covariates evaluated (race, age, weight, BMI, or smoking status) resulted in a statistically significant effect after forward selection (P > .05) and backward elimination (P > .001). However, there was a trend for decreased EE exposure with increasing weight and among smokers versus nonsmokers. By comparing the simulated data to the raw data, the visual predictive check indicated no apparent biases in overall model fit. The simulated median and the simulated 5th and 95th percentiles were highly correlated with the corresponding percentiles of the observed data. The percentage of the observed concentrations below the 5th percentile was 4.3% and the percentage above the 95th percentile was 4.9%.
Population PK of EE The final population PK model was a 2-compartment model with a combination of zero-order input into a depot compartment, first-order absorption into the central compartment, and first-order elimination. A term estimating the relative bioavailability of Seasonique and Seasonale relative to Quartette and LoSeasonique was also included in the model. Residual variability was relatively low and ranged from 70% at predicted EE concentrations of 2 pg/mL to 16.3 at predicted EE concentrations of 325 pg/mL. The goodness-of-fit scatterplots in Figure 2 from the final model demonstrated the unbiased fit to the model. Scatterplots illustrating the relationship between the model-predicted estimates and the 1406
Reproductive Sciences 21(11)
Figure 3. Scatterplot of model-predicted versus noncompartmental area under the plasma concentration versus time curve from time zero to infinity (AUC0-1) from the final model. The black line is a line of identity (unit slope). EE indicates ethinyl estradiol.
Figure 4. Scatterplot of model-predicted versus noncompartmental maximum observed plasma drug concentration (Cmax) from the final model. The black line is a line of identity (unit slope). EE indicates ethinyl estradiol.
noncompartmental parameters are shown for AUC0-1 in Figure 3 and Cmax in Figure 4. These plots demonstrate that the model-based estimates of exposure and those based on noncompartmental analysis of the observed concentration data were comparable. As a result, this summary reports parameter estimates obtained using the population model. Table 3 provides the PK parameter estimates and standard errors from the final model. All parameters were estimated with good precision (%SEM < 25%) and the magnitude of IIV was small in all parameters. There was no relationship between EE dose and CL/F
Darwish et al
1407
Table 3. PK Estimates and Standard Errors From the Final Model. Final Parameter Estimate Parameter
Population Mean
%SEM
Magnitude of IIV (%CV) Final Estimate
CL/F, L/h 48.1 3.7 27.55a b Shift on CL/F, L/h 7.8 19.0 Vc/F, L 368 4.3 35.21a b Shift on Vc/F, L 74.9 15.4 Vp/F, L 505 4.1 29.39a Q/F, L/h 61.0 4.1 24.52a ka, 1/h 1.92 3.6 31.40 D1, h 0.68 3.1 36.88 F 1.17 4.6 NE RVc 1.36, 0.163 25.8, 4.4 NA Minimum value of the objective function ¼ 33 182.693
%SEM 12.3 15.6 12.7 17.1 23.7 15.6 NA NA
Abbreviations: CL/F, apparent oral clearance; %CV, coefficient of variation; D1, duration of zero-order input into absorption (depot) compartment; EE, ethinyl estradiol; F, bioavailability of Seasonique and Seasonale products relative to Quartette and LoSeasonique; IIV, interindividual variability; ka, first-order absorption rate constant; NA, not applicable; NE, not estimated; PK, pharmacokinetics; Q/F, apparent intercompartmental clearance; %SEM, percentage standard error of the mean; RV, residual variability; Vc/F, apparent central volume of distribution; Vp/F, apparent volume of the peripheral compartment. a Estimates (%SEM) of covariance terms: (IIV on CL/F, IIV on Vc/F) ¼ 0.0689 (16.3%), (IIV on CL/F, IIV on Q) ¼ 0.0478 (17.8%), (IIV on Vc/F, IIV on Q) ¼ 0.0558 (19.0%), (IIV on CL/F, IIV on Vp) ¼ 0.0472 (18.0%), (IIV on Vc/F, IIV on Vp) ¼ 0.0487 (20.7%), (IIV on Q, IIV on Vp) ¼ 0.0690 (14.8%). b Additional shift for multiple dosing versus single dose. c RV expressed as standard deviations of the additive and constant coefficient of variation components. These estimates correspond to a range of RV from 69.9%CV at an individual predicted EE concentration of 2 pg/mL to 16.3%CV at an individual predicted EE concentration of 325 pg/mL.
(P > .05). Model-based estimates of AUC0-1 and Cmax dose proportionality are shown in Figure 5. In the final model, a statistically significant reduction in the values of Vc/F and CL/F with multiple dosing versus single dosing was observed. Specifically, the Vc/F was 74.9 L (20.4%) smaller and CL/F was 7.8 L/h (16.2%) smaller during multiple dosing than with single dosing (P < .001). The terminal elimination half-life was 16.5 hours after a single dosing and 17.8 hours after multiple dosing. At steady state, the volume of distribution was 798 L following multiple dosing. Model-predicted PK parameters were calculated using the final population PK parameters reported in Table 3 and observed sample times. Most model-predicted means and medians were within 10% of the corresponding noncompartmental values. The largest difference between modelpredicted and noncompartmental estimates occurred with median Cmax after 40-mg EE, with the model-based estimate for the median Cmax being 17% lower than the corresponding noncompartmental estimate.
Simulation of Exposure Based on simulations of EE-concentration data following the dosing regimens for Quartette, Seasonique, and LoSeasonique
used in Phase 3 studies, estimates of EE exposure were calculated on various days over the 91-day cycle after accounting for the slight differences in demographic characteristics between the study populations (Table 4). Figure 6 illustrates the mean predicted single-dose and steady-state plasma concentration versus time profiles of EE following administration of a 20-mg dose within Quartette. Figure 7 depicts the population model-predicted EE concentration versus time profiles on days 42, 63, 84, and 91 with the use of Quartette and represents the profile of each dose of EE within Quartette over 24 hours. Increases in exposure occur in a stepwise fashion with increasing EE dose over the first 84 days of the extended cycle followed by lower concentrations of EE with the 10-mg daily dose of estrogen that replaces the traditional hormonefree interval. Predicted EE trough concentrations for Quartette on the day before the scheduled change in EE dose during the extended cycle days 1 through 84 were 9.67 pg/mL on day 42, 12.08 pg/mL on day 63, and 14.5 pg/mL on day 84. In contrast, predicted EE trough concentrations remained constant for the other extended-regimen COCs at 16.97 pg/mL on days 42, 63, and 84 for Seasonique and 9.67 pg/mL on days 42, 63, and 84 for LoSeasonique. On day 91, predicted trough EE concentrations were 4.85 pg/mL for Quartette, 5.67 pg/mL for Seasonique, and 4.84 pg/mL for LoSeasonique.
Discussion The PK of EE was well characterized by a 2-compartment model with first-order elimination and was consistent with the known profile of EE.13 Systemic exposure to EE after administration of Quartette increased in a dose-proportionate manner. The appropriateness of the population model was confirmed by the lack of bias in the goodness-of-fit diagnostic plots and the results of the visual predictive check. This analysis elucidates several aspects of the PK of EE. First, the bioavailability of Seasonique and Seasonale was determined to be 117% relative to Quartette and LoSeasonique. Given that the patient populations enrolled in the Phase 1 PK studies were rather homogenous, this difference in bioavailability may be due to differences in assay performance across various laboratories over time and lack of cross-validation of these assays across laboratories. Another finding was that there was no evidence of dose-dependent elimination. There was a statistically significant reduction in the values of Vc/F and CL/F following multiple dosing as compared to those after a single dose. Taken together, changes in CL/F and Vc/F suggested a predicted EE elimination half-life of 16.5 hours after single dosing and 17.8 hours after multiple dosing. The impact of covariates on EE PK was less substantial. Evaluation of the effects of race, age, weight, BMI, and smoking failed to detect statistically significant effects of these covariates on EE PK. Nonetheless, a trend toward decreased EE exposure with increasing weight and among smokers versus nonsmokers was observed—findings consistent with other studies.14-16 1407
1408
Reproductive Sciences 21(11)
Figure 5. Model-predicted ethinyl estradiol (EE) area under the plasma concentration versus time curve from time zero to infinity (AUC 0-1; A) and maximum observed plasma drug concentration (Cmax; B) by EE dose. Table 4. Demographic and Baseline Characteristics of Subjects in the Pooled PD Analysis of Phase 3 Studies of LNG/EE Extended Regimens. Quartette (n ¼ 3066) Median age, y (min, max) Age
A Comparison of the Pharmacokinetic Profile of an Ascending-Dose, ExtendedRegimen Combined Oral Contraceptive to Those of Other Extended Regimens
Reproductive Sciences 2014, Vol. 21(11) 1401-1410 ª The Author(s) 2014 Reprints and permission: sagepub.com/journalsPermissions.nav DOI: 10.1177/1933719114526472 rs.sagepub.com
Mona Darwish, PhD1, Mary Bond, MS, MBA2, Nancy Ricciotti, MSN3, Jennifer Hsieh, MS4, Jill Fiedler-Kelly, MS5, and Thaddeus Grasela, PharmD, PhD5
Abstract Quartette (levonorgestrel [LNG]/ethinyl estradiol [EE] and EE) is an ascending-dose, extended-regimen combined oral contraceptive (COC) that consists of a constant dose of LNG 150 mg on days 1 to 84 with EE 20 mg on days 1 to 42, 25 mg on days 43 to 63, 30 mg on days 64 to 84, and 10 mg of EE monotherapy on days 85 to 91. A population pharmacokinetic (PK) model for EE was developed using nonlinear mixed-effects modeling to characterize the PK profile of EE administered in Quartette and other extended-regimen LNG/EE COCs. Model-predicted plasma concentration–time profiles demonstrated a stepwise increase in systemic exposure to EE during the first 84 days of the cycle following each EE dose change. Lower concentrations of EE were noted during the final 7-day period of EE 10 mg. Gradual increases in EE seen with Quartette may decrease the incidence of unscheduled bleeding frequently observed during early cycles of extended-regimen COCs. Keywords combined oral contraceptive, ascending-dose, extended regimen, bleeding, ethinyl estradiol
Introduction Extended regimens of combined hormonal oral contraceptives (COCs) are an increasingly popular option for women seeking contraception because they reduce the frequency of scheduled withdrawal bleeding compared to traditional 28-day COCs.1,2 However, unscheduled bleeding and spotting with extendedregimen COCs is common in early COC cycles3-5 and may contribute to poor adherence and early treatment discontinuation.6 Two strategies for stabilizing the endometrium and reducing unscheduled bleeding include gradually increasing estrogen doses during the COC cycle and using a low-dose ethinyl estradiol (EE) during the traditional hormone-free interval.5,7-9 Quartette (levonorgestrel [LNG]/EE and EE) is a novel ascending-dose, extended-regimen COC that has been approved by the US Food and Drug Administration. The regimen consists of a continuous LNG dose combined with a gradually increasing EE dose for 84 days followed by a low dose of EE for 7 days. This design is intended to provide an increasing ratio of estrogen to progestin across the extended cycle, with the ratio increasing just prior to the times during the cycle when unscheduled bleeding most frequently occurs.9 The pharmacokinetics (PK) of LNG and EE are well studied, and while Quartette (Teva Women’s Health, Inc., Sellersville, Pennsylvania) uses LNG doses comparable to those in
other extended-regimen COCs, the PK of EE in the context of Quartette has not been assessed. Because LNG dosages (100 mg or 150 mg) and exposure are constant during the first 84 days of the extended regimens included in this report, this analysis focuses on the PK of EE. The objectives of this analysis, which used a population modeling approach, were to characterize the overall PK profile of EE after administration of Quartette, including its relative bioavailability, dose proportionality, and single- and multiple-dose EE PK. This analysis also explored the effects of selected covariates, such as weight, body mass index (BMI), age, race, and smoking status as
1
Clinical Pharmacology, SCI-MED BRIDGE, Malvern, PA, USA Phase 1 & Clinical Pharmacology, Teva Branded Pharmaceutical Products, R&D, Inc., Frazer, PA, USA 3 Teva Women’s Health, Teva Branded Pharmaceutical Products, R&D, Inc., Frazer, PA, USA 4 Global Biostatistics, Teva Branded Pharmaceutical Products, R&D, Inc., West Chester, PA, USA 5 Pharmacometric Services, Cognigen Corporation, Buffalo, NY, USA 2
Corresponding Author: Mary Bond, Phase I & Clinical Pharmacology, Teva Branded Pharmaceutical Products, R&D, Inc., 41 Moores Road, Frazer, PA 19355, USA. Email: [email protected]
1402
Reproductive Sciences 21(11)
Table 1. Description of Included Studies. Subjects With Usable Subjects EE-Concentration Data, na Enrolled, n
Dosing Regimen
Quartette
Days 1-42: LNG 150 mg/ Predose (0) and at 0.5, 1, 1.25, 1.5, 1.75, 2, 2.5, 3, 4, 6, 8, 10, 12, 16, 24, 36, 48, EE 20 mg 72, and 96 hours after dose Days 43-63: LNG 150 mg/ EE 25 mg Days 64-84: LNG 150 mg/ EE 30 mg Days 85-91: EE 10 mg Days 1-84: LNG 150 mg/ Predose (0) and at 0.5, 1.0, 1.25, 1.5, 1.75, 2, 2.5, 3, 4, 6, 8, 10, 12, 16, 24, EE 30 mg 36, 48, 72, and 96 hours after dose Days 85-91: placebo
18
17
Healthy, nontobaccousing women with normal menstrual cycle; 18- to 45-year-old; BMI 18-30 kg/m2
30
29
Days 1-84: LNG 150 mg/ Days 1, 21: predose (0) and at 0.5, 1, 1.33, 1.67, 2, 3, 4, 6, 8, 11, 15, and 24 EE 30 mg hours after dose Days 85-91: EE 10 mg Day 84: predose (0) and at 0.5, 1, 1.33, 1.67, 2, 3, 4, 6, 8, 11, 15, 24, 36, 48, 72, 96, 120, and 144 hours after dose Day 91: predose (0) and at 0.5, 1, 1.33, 1.67, 2, 3, 4, 6, 8, 11, 15, 24, 36, 48, 72, and 96 hours after dose Days 1-84: LNG 150 mg/ Predose (0) and at 0.33, 0.67, 1, 1.33, 1.67, 2, 2.5, 3, 4, 6, 8, 12, 16, 24, 36, EE 30 mg 48, and 72 hours after dose (EE and Days 85-91: EE 10 mg LNG) 96 and 120 hours after dose (LNG only)
30
30
Healthy women; 18- to 35-year-old; Within 15% of their ideal body weight Healthy women with normal menstrual cycle; 19- to 51-year-old; 119-191 lb
30
29
1-84: LNG 150 mg/ Predose (0) and at 0.33, 0.67, 1, 1.33, 1.67, 2, 2.5, 3, 4, 6, 8, 10, 12, 16, 24, 30 mg 36, 48, 72, and 96 hours after dose 85-91: EE 10 mg 1-84: LNG 100 mg/ Predose (0) and at 0.5, 1.0, 1.25, 1.5, 1.75, 2, 2.5, 3, 4, 6, 8, 10, 12, 16, 24, 20 mg 36, 48, 72, and 96 hours after dose 85-91: EE 10 mg
18
17
35
30
Seasonale
Seasonique Study 1b
Seasonique study 2
Seasonique study 3
Days EE Days LoSeasonique Days EE Days
PK Sampling
Subject Characteristics
Study
Healthy, nontobaccousing women with normal menstrual cycle; 19- to 51-year-old; BMI 18-30 kg/m2 Healthy women; 18- to 47-year-old; 53-87 kg Healthy women with regular menstrual cycle; 18- to 35-year-old; 48-75 kg
Abbreviations: EE, ethinyl estradiol; LNG, levonorgestrel; PK, pharmacokinetics; BMI, body mass index. a All studies assessed single-dose PK with the exception of Seasonique study 1. b Seasonique study 1 included both multiple-dose and single-dose assessments.
predictors of interindividual variability (IIV) in EE disposition; evaluated the performance of the developed PK model; and simulated the plasma concentration–time course for EE exposure over two 91-day cycles for Quartette and 2 other extended-regimen COCs (Seasonique and LoSeasonique) in order to compare estimates of EE exposure over various days within the 91-day cycle.
Design and Methodology A PK model for EE was developed using nonlinear mixedeffects modeling to characterize the PK of EE administered in various dosing regimens and formulations used in Quartette and other extended-regimen LNG/EE COCs. The PK data used to develop this model were obtained from 5 single-dose studies 1402
and 1 multiple-dose study of several extended-regimen COC products containing LNG and EE (Table 1).
Subjects, Drug Administration, and Blood Sampling Participants in each of the 6 studies were healthy women between the age of 18 and 51 years who had regular menstrual cycles. Ethinyl estradiol was administered as described in Table 1. In these studies, methods used to measure plasma EE concentrations included liquid chromatography–tandem mass spectrometry (with a lower limit of quantification [LLOQ] of 1-2 pg/mL) and gas chromatography–mass spectrometry–negative chemical ionization (with an LLOQ of 5.07 pg/mL). In each study, extensive PK sampling was performed, including samples at predose and up to 144 hours after the drug was first
Darwish et al
1403
Figure 1. Flowchart of model development process.
administered as detailed in Table 1. In the multiple-dose study, trough PK samples were collected prior to dosing on days 18, 19, 20, 81, 82, and 83.
Population PK Analysis The model was used to characterize the PK profile, assess the bioavailability and dose proportionality of EE, detect possible differences in EE PK between single- and multiple-dose regimens, and evaluate the impact of selected covariates as predictors of IIV in EE disposition. Figure 1 graphically depicts the general procedure used to develop the population PK model.
The population PK analysis was completed using nonlinear mixed-effects modeling software (NONMEM, version 6, level 2.0; ICON Development Solutions 2006, Hanover, Maryland). The first-order conditional estimation with interaction estimation method was used throughout the model development.
Exploratory Graphical Analysis Exploratory graphical analyses and data visualization techniques were used to guide the selection of a base model for the population PK analysis. The key objective of these analyses was to confirm whether the models to be tested were 1403
1404
Reproductive Sciences 21(11)
Table 2. Demographic Characteristics for the Population PK Analysis Data Set. Subject Characteristics
Overall Values
N Age, y Mean (SD) Median (range) Body mass index, kg/m2 Mean (SD) Median (range) Weight, kg Mean (SD) Median (range) Race, n (%) Caucasian Black Hispanic Other Smoker, n (%) No Yes
152 28.3 (7.7) 28.0 (18.0-51.0) 20.9 (3.6) 19.9 (15.5-29.8) 65.9 (9.4) 65.1 (48.2-88.6) 80 (52.6) 33 (21.7) 36 (23.7) 3 (2.0) 137 (90.1) 15 (9.9)
Abbreviations: PK, pharmacokinetics; SD, standard deviation.
appropriate and to verify the assumptions of the model. This exploratory analysis was also used to search for extreme values and potential outliers, assess possible trends in the data, and determine whether any errors were made in the handling of the data or the creation of the analysis data sets.
Selection of Base Structural PK Model Results of the exploratory analyses were used to determine the appropriate functional form of the base structural model of drug concentration versus time data. Previous literature10,11 and preliminary examination of plasma EE concentration profiles indicated that a linear, 2-compartment, open model with first-order absorption and elimination would adequately describe the data. Various structural models were applied and their capacities to characterize the population PK of EE were assessed. Each population PK model was described by estimating mean structural model parameters (eg, absorption rate constant [ka], bioavailability fraction [F], apparent volumes of distribution in central and peripheral compartments [Vc/F and Vp/F], apparent oral clearance [CL/F], and apparent intercompartmental clearance [Q/F]), the magnitude of IIV, and residual variability (RV). The IIV was estimated on CL/F, Vc/F, Q/F, Vp/F, ka, F, and duration of zero-order input into absorption (depot) compartment (D1) using exponential error models. The RV was estimated using a combined additive and constant coefficient of variation error model. Given that all data used in these analyses were obtained after oral drug administration, the F was assumed to be 100% and the PK parameters were considered apparent values. The impact of the extended-regimen COC (Quartette vs Seasonique vs Seasonale vs LoSeasonique) was tested on the relative F. The effect of the comparators was evaluated as a shift in the relative F compared to Quartette. 1404
Covariate Analysis Covariate analyses were performed to explore the influence of selected factors on the magnitude of IIV and RV in EE PK. These analyses were performed using a forward selection followed by a backward elimination process (Figure 1). Covariates were evaluated for their ability to explain IIV in CL/F, Vc/F, and Vp/F and included race, age, weight, BMI, and smoking. The covariate models were developed using graphical and statistical approaches to define the mathematical forms of the relationships in question and assess their statistical significance. The correlation between covariates was examined to avoid potential multicollinearity or confounding of effects in covariate submodels. If a covariate was highly correlated with another covariate (eg, body weight and BMI), only 1 of the highly correlated covariates was evaluated. Linear, exponential, power, additive, and proportional shift models were used to evaluate continuous and categorical covariates in NONMEM, as appropriate.
Population PK Model Refinement and Evaluation The reduced multivariable model, including all significant covariates, was evaluated for any remaining biases in the IIV and RV error models. A simulation-based, prediction-corrected, visual predictive check method was used to evaluate the adequacy of the final model.12 The final model was used to simulate 1000 replicates of the analysis data set with NONMEM. Statistics of interest (eg, the 5th, 50th, and 95th percentiles of the distributions of concentration) were calculated from the simulated and observed data for comparison. These percentiles were then plotted versus time to enable a visual assessment of concordance between the modelbased simulated data and the original observed data set and/or percentiles based on the observed data. Model-predicted PK parameters were calculated using the final population PK model. For subjects receiving Quartette, model-predicted values were then compared with values for area under the curve (AUC), maximum observed plasma drug concentration (Cmax), time to maximum observed concentration (tmax), and half-life calculated using noncompartmental analyses of the observed EE concentrations over time. The previous noncompartmental analysis was performed using WinNonlin (Pharsight Corporation, Mountain View, California) and the linear trapezoidal rule. The EE AUC from time zero to infinity (AUC0-1) was calculated using individual modelpredicted EE concentrations for each study day in which a full profile was obtained. Cmax, tmax, and half-life were also estimated.
Simulations Using Final Population PK Model The population PK model was used to predict EE concentration–time profiles at various days during the 91-day cycle for women enrolled in Phase 3 clinical studies of the extended regimens included in this analysis. Ethinyl estradiol concentrations
Darwish et al
1405
Figure 2. Goodness-of-fit scatterplots for the final population PK model of EE. EE indicates ethinyl estradiol; PK, pharmacokinetics.
were simulated over two 91-day cycles for Quartette, Seasonique, and LoSeasonique based on the dosing regimens used for each product (Table 1).
Results Demographics and Baseline Characteristics The population PK analysis included 6433 plasma EE concentrations from 152 female subjects of childbearing age (Table 2). On average, subjects were aged 28.3 (+7.7) years and weighed 65.9 (+9.4) kg. Mean BMI was 20.9 (+3.6) kg/m2 and 9.9% were smokers.
Between 20 and 55 samples were obtained from the majority of subjects. A small percentage of samples obtained more than 50 hours after the last dose was associated with measurable concentrations (n ¼ 106, 1.6%). These concentrations were considered to be extreme values and due to their potential to unduly influence the model, they were excluded from the analysis data set. After these data were excluded, the final analysis data set included 6327 concentration values from 152 subjects.
Pharmacokinetic Data Analysis Ethinyl estradiol concentrations declined from peak in a biphasic manner. Initially, a 2-compartment, open model with a 1405
1406 combination of zero-order input into a depot compartment, first-order absorption into the central compartment, and firstorder elimination was fit to the data. The analysis data set consisted of PK data from 4 products, and potential differences in F between these products were evaluated during the development of the base model. In the base structural model, the bioavailability of Seasonique and Seasonale was determined to be 119% relative to Quartette and LoSeasonique (F ¼ 1). All parameters were estimated with good precision (standard error of the mean expressed as a percentage [%SEM] < 20%), with the exception of the IIV for ka, which was estimated with slightly poorer precision (%SEM ¼ 31.2%). Goodness-offit plots were unbiased and demonstrated the adequacy of the base model. The effect of EE dose on clearance was evaluated using the base structural model; however, no evidence for a lack of dose proportionality was detected. A covariate analysis was performed to explore the sources of variability in EE PK. The effects of multiple dosing and demographic covariates (multiple dosing, race, age, weight, BMI, and smoking status) on CL/F, Vc/F, and Vp/F were evaluated. After 4 rounds of forward selection, 3 statistically significant predictors of variability were identified. Multiple dosing was a statistically significant predictor of Vc/F and CL/F, and weight was a statistically significant predictor of Vp/F (P < .05). Univariate stepwise backward elimination proceeded after all adjustments had been made to the IIV and RV error models. None of the demographic covariates evaluated (race, age, weight, BMI, or smoking status) resulted in a statistically significant effect after forward selection (P > .05) and backward elimination (P > .001). However, there was a trend for decreased EE exposure with increasing weight and among smokers versus nonsmokers. By comparing the simulated data to the raw data, the visual predictive check indicated no apparent biases in overall model fit. The simulated median and the simulated 5th and 95th percentiles were highly correlated with the corresponding percentiles of the observed data. The percentage of the observed concentrations below the 5th percentile was 4.3% and the percentage above the 95th percentile was 4.9%.
Population PK of EE The final population PK model was a 2-compartment model with a combination of zero-order input into a depot compartment, first-order absorption into the central compartment, and first-order elimination. A term estimating the relative bioavailability of Seasonique and Seasonale relative to Quartette and LoSeasonique was also included in the model. Residual variability was relatively low and ranged from 70% at predicted EE concentrations of 2 pg/mL to 16.3 at predicted EE concentrations of 325 pg/mL. The goodness-of-fit scatterplots in Figure 2 from the final model demonstrated the unbiased fit to the model. Scatterplots illustrating the relationship between the model-predicted estimates and the 1406
Reproductive Sciences 21(11)
Figure 3. Scatterplot of model-predicted versus noncompartmental area under the plasma concentration versus time curve from time zero to infinity (AUC0-1) from the final model. The black line is a line of identity (unit slope). EE indicates ethinyl estradiol.
Figure 4. Scatterplot of model-predicted versus noncompartmental maximum observed plasma drug concentration (Cmax) from the final model. The black line is a line of identity (unit slope). EE indicates ethinyl estradiol.
noncompartmental parameters are shown for AUC0-1 in Figure 3 and Cmax in Figure 4. These plots demonstrate that the model-based estimates of exposure and those based on noncompartmental analysis of the observed concentration data were comparable. As a result, this summary reports parameter estimates obtained using the population model. Table 3 provides the PK parameter estimates and standard errors from the final model. All parameters were estimated with good precision (%SEM < 25%) and the magnitude of IIV was small in all parameters. There was no relationship between EE dose and CL/F
Darwish et al
1407
Table 3. PK Estimates and Standard Errors From the Final Model. Final Parameter Estimate Parameter
Population Mean
%SEM
Magnitude of IIV (%CV) Final Estimate
CL/F, L/h 48.1 3.7 27.55a b Shift on CL/F, L/h 7.8 19.0 Vc/F, L 368 4.3 35.21a b Shift on Vc/F, L 74.9 15.4 Vp/F, L 505 4.1 29.39a Q/F, L/h 61.0 4.1 24.52a ka, 1/h 1.92 3.6 31.40 D1, h 0.68 3.1 36.88 F 1.17 4.6 NE RVc 1.36, 0.163 25.8, 4.4 NA Minimum value of the objective function ¼ 33 182.693
%SEM 12.3 15.6 12.7 17.1 23.7 15.6 NA NA
Abbreviations: CL/F, apparent oral clearance; %CV, coefficient of variation; D1, duration of zero-order input into absorption (depot) compartment; EE, ethinyl estradiol; F, bioavailability of Seasonique and Seasonale products relative to Quartette and LoSeasonique; IIV, interindividual variability; ka, first-order absorption rate constant; NA, not applicable; NE, not estimated; PK, pharmacokinetics; Q/F, apparent intercompartmental clearance; %SEM, percentage standard error of the mean; RV, residual variability; Vc/F, apparent central volume of distribution; Vp/F, apparent volume of the peripheral compartment. a Estimates (%SEM) of covariance terms: (IIV on CL/F, IIV on Vc/F) ¼ 0.0689 (16.3%), (IIV on CL/F, IIV on Q) ¼ 0.0478 (17.8%), (IIV on Vc/F, IIV on Q) ¼ 0.0558 (19.0%), (IIV on CL/F, IIV on Vp) ¼ 0.0472 (18.0%), (IIV on Vc/F, IIV on Vp) ¼ 0.0487 (20.7%), (IIV on Q, IIV on Vp) ¼ 0.0690 (14.8%). b Additional shift for multiple dosing versus single dose. c RV expressed as standard deviations of the additive and constant coefficient of variation components. These estimates correspond to a range of RV from 69.9%CV at an individual predicted EE concentration of 2 pg/mL to 16.3%CV at an individual predicted EE concentration of 325 pg/mL.
(P > .05). Model-based estimates of AUC0-1 and Cmax dose proportionality are shown in Figure 5. In the final model, a statistically significant reduction in the values of Vc/F and CL/F with multiple dosing versus single dosing was observed. Specifically, the Vc/F was 74.9 L (20.4%) smaller and CL/F was 7.8 L/h (16.2%) smaller during multiple dosing than with single dosing (P < .001). The terminal elimination half-life was 16.5 hours after a single dosing and 17.8 hours after multiple dosing. At steady state, the volume of distribution was 798 L following multiple dosing. Model-predicted PK parameters were calculated using the final population PK parameters reported in Table 3 and observed sample times. Most model-predicted means and medians were within 10% of the corresponding noncompartmental values. The largest difference between modelpredicted and noncompartmental estimates occurred with median Cmax after 40-mg EE, with the model-based estimate for the median Cmax being 17% lower than the corresponding noncompartmental estimate.
Simulation of Exposure Based on simulations of EE-concentration data following the dosing regimens for Quartette, Seasonique, and LoSeasonique
used in Phase 3 studies, estimates of EE exposure were calculated on various days over the 91-day cycle after accounting for the slight differences in demographic characteristics between the study populations (Table 4). Figure 6 illustrates the mean predicted single-dose and steady-state plasma concentration versus time profiles of EE following administration of a 20-mg dose within Quartette. Figure 7 depicts the population model-predicted EE concentration versus time profiles on days 42, 63, 84, and 91 with the use of Quartette and represents the profile of each dose of EE within Quartette over 24 hours. Increases in exposure occur in a stepwise fashion with increasing EE dose over the first 84 days of the extended cycle followed by lower concentrations of EE with the 10-mg daily dose of estrogen that replaces the traditional hormonefree interval. Predicted EE trough concentrations for Quartette on the day before the scheduled change in EE dose during the extended cycle days 1 through 84 were 9.67 pg/mL on day 42, 12.08 pg/mL on day 63, and 14.5 pg/mL on day 84. In contrast, predicted EE trough concentrations remained constant for the other extended-regimen COCs at 16.97 pg/mL on days 42, 63, and 84 for Seasonique and 9.67 pg/mL on days 42, 63, and 84 for LoSeasonique. On day 91, predicted trough EE concentrations were 4.85 pg/mL for Quartette, 5.67 pg/mL for Seasonique, and 4.84 pg/mL for LoSeasonique.
Discussion The PK of EE was well characterized by a 2-compartment model with first-order elimination and was consistent with the known profile of EE.13 Systemic exposure to EE after administration of Quartette increased in a dose-proportionate manner. The appropriateness of the population model was confirmed by the lack of bias in the goodness-of-fit diagnostic plots and the results of the visual predictive check. This analysis elucidates several aspects of the PK of EE. First, the bioavailability of Seasonique and Seasonale was determined to be 117% relative to Quartette and LoSeasonique. Given that the patient populations enrolled in the Phase 1 PK studies were rather homogenous, this difference in bioavailability may be due to differences in assay performance across various laboratories over time and lack of cross-validation of these assays across laboratories. Another finding was that there was no evidence of dose-dependent elimination. There was a statistically significant reduction in the values of Vc/F and CL/F following multiple dosing as compared to those after a single dose. Taken together, changes in CL/F and Vc/F suggested a predicted EE elimination half-life of 16.5 hours after single dosing and 17.8 hours after multiple dosing. The impact of covariates on EE PK was less substantial. Evaluation of the effects of race, age, weight, BMI, and smoking failed to detect statistically significant effects of these covariates on EE PK. Nonetheless, a trend toward decreased EE exposure with increasing weight and among smokers versus nonsmokers was observed—findings consistent with other studies.14-16 1407
1408
Reproductive Sciences 21(11)
Figure 5. Model-predicted ethinyl estradiol (EE) area under the plasma concentration versus time curve from time zero to infinity (AUC 0-1; A) and maximum observed plasma drug concentration (Cmax; B) by EE dose. Table 4. Demographic and Baseline Characteristics of Subjects in the Pooled PD Analysis of Phase 3 Studies of LNG/EE Extended Regimens. Quartette (n ¼ 3066) Median age, y (min, max) Age
