Clin Pharmacokinet (2014) 53:479–488 DOI 10.1007/s40262-013-0130-1
ORIGINAL RESEARCH ARTICLE
Population Pharmacokinetic Modeling of Veliparib (ABT-888) in Patients with Non-Hematologic Malignancies Ahmed Hamed Salem • Vincent L. Giranda Nael M. Mostafa
•
Published online: 23 January 2014 Ó Springer International Publishing Switzerland 2014
Abstract Background and Objective Veliparib (ABT-888) is a potent oral inhibitor of Poly(ADP-ribose) polymerase enzyme that is currently in development for the treatment of non-hematologic and hematologic malignancies. This analysis characterizes the population pharmacokinetics of veliparib, including developing a structural pharmacokinetic model and testing patient demographics and covariates for potential influence on veliparib pharmacokinetics in patients with non-hematologic malignancies. Methods The analysis dataset included 3,542 veliparib concentration values from 325 patients with non-hematologic malignancies enrolled in three phase I and one phase II studies. Population pharmacokinetic modeling was performed using NONMEM. The likelihood ratio test was used for comparison of nested models, and visual predictive check was employed for model qualification. Covariates tested included body size measures, creatinine clearance (CLCR), formulation, age, sex, race, liver function tests, and coadministration with temozolomide. Results A one-compartment model with first-order absorption and elimination adequately described veliparib pharmacokinetics. The final model included fixed effects for CLCR on veliparib oral clearance (CL/F) and lean body mass (LBM) on volume of distribution (Vd/F). CL/F and Vd/F were 20.9 L/h (for a CLCR of 100 mL/min) and 173 L (for an LBM of 56 kg), respectively. A. H. Salem (&) V. L. Giranda N. M. Mostafa Clinical Development, AbbVie Inc., 1 North Waukegan Road, North Chicago, IL 60064, USA e-mail:
[email protected] A. H. Salem Department of Clinical Pharmacy, Faculty of Pharmacy, Ain Shams University, Cairo, Egypt
Conclusion Only LBM and CLCR were found to be determinants of veliparib Vd/F and CL/F, respectively. Dosage adjustments of veliparib on the basis of body size, age, sex, race, liver function, and temozolomide coadministration are not necessary in patients with non-hematologic malignancies. This is the first study to characterize the population pharmacokinetics of veliparib, and the developed model will be used to conduct simulations and evaluate veliparib exposure-response relationships.
1 Introduction Poly(ADP-ribose)-polymerase (PARP) is a nuclear enzyme that recognizes deoxyribonucleic acid (DNA) damage and facilitates repair of both single-stranded and double-stranded DNA breaks [1, 2]. Cancer cells are more dependent than normal cells on PARP for DNA repair and their higher PARP expression has been linked to drug resistance and the overall ability to sustain genotoxic stress [3–6]. Thus, PARP inhibition was hypothesized to result in less efficient DNA repair following an insult by DNA-damaging agents such as cytotoxic chemotherapy and radiation therapy. Veliparib (ABT-888) is a potent oral inhibitor of the PARP1 and 2 enzymes. In preclinical models of melanoma, and breast and colon cancer, veliparib potentiated the antitumor activity of several DNA-damaging agents such as temozolomide, cyclophosphamide, platinums, and radiation [7–9]. Because of its broad spectrum chemosensitizing and radiosensitizing properties, veliparib is being investigated for the treatment of malignancies in several phase I and II clinical studies (http://www.clinicaltrials. gov). In clinical studies on subjects with refractory solid tumors and lymphomas, veliparib caused significant reductions in the Poly (ADP-ribose) (PAR) levels [10–12].
PK samples were collected at 0, 0.25, 0.5, 1, 1.5, 2, 3, 4, 6, 8, 10, 12, and 24 h after dosing on each study period Phase I, crossover bioavailability study of three formulations of veliparib 4
PK pharmacokinetic, HCC hepatocellular carcinoma, bid twice daily
40 mg single dose
Phase II, placebo-controlled study of veliparib in combination with temozolomide 3 [32]
Subjects with solid tumors (N = 27)
PK samples were collected at 0.5, 1, 2, and 3 h after dosing on day 1 of cycle 1, and at least 1 h postdose on day 1 of cycle 2 through cycle 6 20 and 40 mg bid
Phase I, dose-escalation study of veliparib in combination with whole-brain radiation therapy 2 [31]
Subjects with metastatic melanoma (N = 229)
10, 20, 30, 50 mg bid
Phase I, dose-escalation study of veliparib in combination with temozolomide 1
Subjects with brain metastases (N = 28)
10, 20, 40, 60, and 80 mg bid
Study description Study no.
The population pharmacokinetic analysis included veliparib plasma concentration data from patients with nonhematologic malignancies who participated in four phase I/II clinical studies. The study protocols were approved by the Institutional Review Boards of the individual study sites and written informed consent was obtained from each subject prior to enrollment. All subjects were C18 years of age with a histologically confirmed malignancy that was metastatic or unresectable, and had adequate bone marrow, renal, and hepatic function. Table 1 summarizes the clinical studies used in this analysis and their dosing and sampling schemes. Study 1 was a phase I, multicenter, open-label, dose-escalating trial that evaluated the tolerability and pharmacokinetics of veliparib in combination with temozolomide in subjects with non-hematologic malignancies, including metastatic melanoma, BRCA-deficient breast, ovarian, primary peritoneal, or fallopian tube cancer, and hepatocellular carcinoma. Potential subjects who have received prior therapy with dacarbazine or temozolomide were ineligible. In each 28-day cycle, subjects received veliparib twice daily on days 1 through 7, and temozolomide on days 1 through 5, with a 21-day rest period between cycles. Study 2 was a phase I, multicenter, dose-escalation study evaluating the safety, tolerability, and pharmacokinetics of veliparib administered twice daily concurrently with conventional whole-brain radiation therapy (WBRT) in the treatment of subjects with solid primary tumors metastatic to the brain. Subjects had histologically or cytologically confirmed non-central nervous system primary solid malignancy and pathologically- or radiographically-confirmed metastatic disease in the brain. Subjects enrolled in the study received WBRT once-daily for a total of 15 fractions administered over 15 consecutive
Table 1 Overview of the clinical studies included in the analysis
2.1 Clinical Studies and Patient Population
Population
Veliparib doses
2 Methods
Subjects with non-hematologic malignancies, including metastatic melanoma, BRCA-deficient breast, ovarian, primary peritoneal, or fallopian tube cancer, and HCC (N = 41)
Veliparib PK sampling
In animal studies, veliparib showed good oral bioavailability, extensive distribution in the tissues, and crossed the blood-brain barrier [8, 9, 13, 14]. Preliminary pharmacokinetic assessments in cancer patients showed that veliparib is rapidly absorbed following oral administration and is mainly eliminated in the urine as the unchanged parent drug [12]. The objectives of this analysis were to characterize the population pharmacokinetics of veliparib, including developing its structural pharmacokinetic model, estimating pharmacokinetic parameters and associated inter-individual variability, and testing patient demographics and covariates for potential influence on veliparib pharmacokinetics.
PK samples were collected at 0, 0.5, 1, 2, 4, 6, and 24 h after dosing on days 1, 6, and 15
A. H. Salem et al. PK samples were collected at 0, 0.5, 1, 1.5, 2, 4, and 6 h after dosing on days 3 and 7
480
Population Pharmacokinetics of Veliparib (ABT-888)
days, excluding weekends and holidays. Veliparib was administered continuously throughout the entire course of WBRT with one additional day of veliparib dosing following the final session of WBRT. Study 3 was a multicenter, randomized, double-blind, placebo-controlled, phase II study evaluating the efficacy of veliparib in combination with temozolomide versus temozolomide alone in subjects with metastatic melanoma. Subjects were randomized into 1 of 3 arms in a 1:1:1 ratio. One cohort of subjects received temozolomide with placebo. Subjects in the second and third arms received temozolomide, and 20 or 40 mg of veliparib twice daily, respectively. Temozolomide was administered at 150 mg/m2 per day in the first cycle; the dose may have been escalated to 200 mg/m2 in subsequent cycles if the subject did not experience grade 3 or 4 neutropenia or thrombocytopenia, or other clinically significant common terminology criteria for adverse events (CTCAE) greater than grade 3. Temozolomide was administered orally on days 1 through 5 of each 28-day cycle, and oral veliparib or oral placebo was administered on days 1 through 7 of each cycle. Finally, study 4 was a multicenter, single-dose, randomized, four-period, crossover, phase I study that tested the bioavailability of three different formulations of veliparib as well as the effect of food on veliparib bioavailability in subjects with advanced solid tumors. 2.2 Sample Collection and Quantification Blood samples were collected via venipuncture or central line into ethylenediaminetetraacetic acid (EDTA) tubes and stored on ice until centrifugation. Plasma samples were then stored at approximately -20 °C until analysis. Analyses for the four studies were conducted at two laboratories, two studies at each laboratory. Both laboratories used solid-phase extraction followed by liquid chromatography with tandem mass spectrometric detection. The assay was selective for veliparib, with lower limit of quantitation (LLQ) established at 1.07 ng/mL in 0.1 mL plasma. Samples quantified above the highest standard were diluted with blank plasma and re-assayed. Only 9 of the 3,542 samples were below LLQ and were included in the analysis as LLQ/2. Quality-control samples supplemented with concentrations of 2.75, 34.4, and 430 ng/mL of veliparib were analyzed with the unknown samples. The coefficient of variation values ranged from 3.0 % to 7.3 %; the mean bias values ranged from –0.1 % and 0.7 %. 2.3 Modeling Methodology Concentration–time data were analyzed using a non-linear mixed-effects population analysis approach with
481
NONMEM (version 7.2.0, ICON Development Solutions, Ellicott City, MD, USA) and PsN (version 3.5.3) [15, 16]. The First-Order Conditional Estimation (FOCE) method P with ETA–EPSILON (g– ) interaction was employed throughout the model development. Graphic processing of the NONMEM output was performed with R (version 3.0) and Xpose (version 4.3.5) [17]. 2.3.1 Development of the Base Model After dose proportionality was established, both one- and two-compartment models with first-order absorption and elimination (ADVAN 2 and ADVAN 3 subroutines in NONMEM) were fitted to the data. Two residual error models were assessed; proportional plus additive error model (combined error model) and proportional error model. Individual pharmacokinetic parameters were assumed to be log-normally distributed, and the interindividual variability in pharmacokinetic parameters was modeled using an exponential error model. Different structures of the X matrix were explored. 2.3.2 Development of the Covariate Model Covariates screened for their possible effect on pharmacokinetic parameters included coadministration with temozolomide, creatinine clearance (CLCR), veliparib formulation, liver function markers (bilirubin, blood urea nitrogen, AST, ALT), age, sex, race, total body weight (WT), body surface area (BSA), body mass index (BMI), lean body mass (LBM) [18], normal-fat mass [19], and fatfree mass (FFM) [20]. CLCR and BSA were calculated using Cockcroft–Gault and Mosteller formulas, respectively [21, 22]. The impact of continuous covariates (COVs) on the pharmacokinetic parameters was explored using Eq. 1: COV hcon TVP = h ð1Þ STD where TVP is the typical value of the estimated population pharmacokinetic parameter, h is the pharmacokinetic parameter population estimate at the standard or median value (STD) of the COV, and hcon is the covariate effect exponent. The impact of categorical covariates on the pharmacokinetic parameters was explored using Eq. 2: TVP ¼ h hlevel cat
ð2Þ
where hcat is the change in the pharmacokinetic parameter population estimate (h) associated with a change in a dichotomous covariate level from 0 to 1. More than one hcat was used for categorical covariates with more than two
482
levels. Multiplicative equations were used to describe the combined effect of multiple covariates on the same parameter. Covariate modeling was performed using the forwardinclusion (0.01 significance level, objective function values [OFV] drop [6.63 for a change of one degree of freedom), backward-elimination approach (0.001 significance level, OFV increase of 10.83 for a change of one degree of freedom) and was guided by evaluation of the empiric Bayesian pharmacokinetic parameter estimates versus covariate plots as well as changes in the estimates of pharmacokinetic parameter variability and residual variability. Nested models were compared using Likelihood ratio test, while non-nested models were compared using Akaike information criterion. The models throughout the population analysis were evaluated by the examination of the convergence of the estimation and covariance routines and the visual assessment of the diagnostic plots, including, but not limited to, the agreement in scatterplots of the population and individual predicted versus measured observations, and the lack of trends or patterns in scatterplots of conditional weighted residuals versus predicted observations and versus time. In addition, quantile-quantile plots for ETAs were examined to assess the underlying normal distribution assumption, and shrinkage in ETAs was also evaluated [23]. Precision of the final model parameter estimates was assessed using the asymptotic standard errors obtained by the covariance routine in NONMEM, as well as by the bootstrap confidence intervals. In bootstrapping, patients were randomly sampled with replacement from the dataset that was used in model development to obtain 1,000 datasets that have the same number of patients as the original dataset. The final model was then fitted to each of these datasets and the parameter estimates were compared with the estimates from the original dataset. 2.3.3 Model Qualification The final model was qualified by prediction and variancecorrected visual predictive check (PC-VPC) where the final parameter estimates were used to simulate 1,000 replicates of the observed dataset. Both observations and the simulated data were normalized on the typical model prediction for the median independent variable in each bin in order to account for variation in sampling times and predictive covariates introduced by binning of the observations [24]. The median, 5th and 95th percentile concentrations of the simulated datasets were then plotted against the original observations.
A. H. Salem et al.
3 Results A total of 3,542 plasma concentrations from 325 patients with non-hematologic malignancies were analyzed. The demographic and clinical characteristics of the patient population included in the pharmacokinetic analysis are summarized in Table 2. A one-compartment disposition model with first-order absorption and elimination adequately described the data. The model was parameterized in terms of absorption rate constant (ka), oral clearance (CL/F) and apparent volume of distribution (Vd/F). Interindividual variability in all three structural model parameters was estimated with high precision indicating the richness of the data. Estimating the covariance matrix between all three parameters was attempted; however, the covariance parameters were estimated with poor precision. Therefore, the analysis allowed for covariance between CL/F and Vd/F only, with their correlation estimated to be 0.8 in the base model. Proportional error model best accounted for the residual unexplained variability of the observed concentrations. When covariates were added univariately in the forward inclusion process, body size covariates on Vd/F were found to be the most significant covariates. LBM and FFM explained larger fractions of the inter-individual variability in Vd/F, and caused greater reductions in the OFV than WT, BSA and BMI. LBM was chosen due to the ease of its calculation. After inclusion of LBM, the only other significant covariate was CLCR on CL/F. Stepwise backward elimination did not result in removal of either of the two covariates. After reaching the final model, we tested the need for estimating the exponent of the allometric model for both LBM and CLCR to ensure model parsimony. Fixing the exponents to 1, as suggested by some researchers [25], resulted in significant increases in OFVs and hence the exponents were estimated from the data. Covariates explained 13 % and 28 % of the variability in CL/F and Vd/F, respectively, relative to the base model. Figure 1 shows that the inclusion of covariates in the final model accounted for the trend observed in the base model between the post hoc ETA estimates for veliparib pharmacokinetic parameters and CLCR, WT, and age. Figure 2 shows the similarity in veliparib CL/F with and without temozolomide coadministration. Estimates from the final model parameters and the precision associated with their estimation are shown in Table 3. All structural and variability parameters were estimated with high precision (relative standard error [RSE] B10%) with minimal shrinkage associated with random effects (6–23 %). The final equations for CL/F and Vd/F were as follows (Eqs. 3, 4):
Population Pharmacokinetics of Veliparib (ABT-888)
483
Table 2 Summary of subjects’ characteristics at baseline Parameter (unit)
Study 1 (N = 41)
Study 2 (N = 28)
Study 3 (N = 229)
Study 4 (N = 27)
Total (N = 325)
Age (years)
55.6 ± 12.8 (33.0–79.0)
55.2 ± 12.0 (31.0–77.0)
60.5 ± 13.1 (24.0–88.0)
57.0 ± 14.4 (29.0–79.0)
59.1 ± 13.2 (24–88)
Weight (kg)
77.6 ± 19.3 (48.0–127) 52.5 ± 10.4 (35.6–80.1)
76.1 ± 16.6 (53.0–115) 53.3 ± 9.9 (40.9–76.1)
82.0 ± 19.3 (42.0–170) 57.3 ± 10.4 (34.6–80.3)
71.6 ± 12.0 (52.0–94.0) 50.1 ± 8.80 (36.7–72.0)
80.1 ± 18.8 (42–170) 56.8 ± 11.5 (34.6–84.6)
27.7 ± 5.7 (18.1–42.1)
26.3 ± 4.5 (18.3–39.3)
28.1 ± 6.1 (16.6–53.7)
26.2 ± 4.34 (18.4–39.1)
27.7 ± 5.8 (16.6–53.7)
Male
14 (34.1)
9 (32.1)
147 (64.2)
6 (22.2)
176 (54.2)
Female
27 (65.9)
19 (67.9)
82 (35.8)
21 (77.8)
149 (45.8)
40 (97.6) 1 (2.4)
25 (89.3)
226 (98.7) 1 (0.4)
27 (100)
318 (97.8) 2 (0.6)
Demographics
Lean body mass (kg) Body mass index (kg/m2) Sex
Race White Asian Black
3 (10.7)
Other
3 (0.9) 2 (0.9)
2 (0.6)
Clinical Creatinine clearance (mL/min)
99.3 ± 32.5 (33.9–168)
96.1 ± 25.5 (56.0–158)
94.2 ± 34.3 (26.6–226)
102 ± 36.9 (38.2–188)
95.7 ± 33.6 (26.61–225.8)
Creatinine (lmol/L)
0.86 ± 0.22 (0.60–1.3)
0.85 ± 0.19 (0.60–1.4)
0.96 ± 0.24 (0.50–2.2)
0.76 ± 0.20 (0.40–1.3)
0.92 ± 0.24 (0.4–2.2)
Albumin (g/L)
4.1 ± 0.55 (2.7–5.5)
3.5 ± 0.44 (2.1–4.3)
4.2 ± 0.44 (2.3–5.4)
3.9 ± 0.41 (3.1–4.8)
4.1 ± 0.49 (2.1–5.5)
ALT (U/L)
29.3 ± 18.9 (9.0–94.0)
40.8 ± 50.1 (13.0–269)
24.2 ± 27.1 (5.0–346)
28.9 ± 33.4 (8.0–184)
26.7 ± 29.7 (5–346)
AST (U/L)
39.3 ± 33.9 (13.0–185)
27.6 ± 23.5 (4.0–108)
23.5 ± 16.1 (8.0–209)
28.6 ± 20.9 (11.0–122)
26.3 ± 20.8 (4–209)
Total bilirubin (mg/dL)
0.50 ± 0.34 (0.10–1.7)
0.43 ± 0.24 (0.10–1.3)
0.65 ± 0.90 (0.20–10.8)
0.54 ± 0.19 (0.30–1.2)
0.60 ± 0.77 (0.1–10.8)
Data are presented as mean ± SD (range) or N (%)
CLCR mL=min 0:48 CL/FðL/hÞ ¼ 20:9 ð3Þ 100 LBM kg 0:73 Vd =FðLÞ ¼ 173 ð4Þ 56 The goodness-of-fit to the observed concentrations is demonstrated in Fig. 3. Population-predicted concentrations were predicted using population parameter estimates and covariate information, while individual-predicted concentrations are based on post hoc empiric Bayes estimates of the pharmacokinetic parameters. The conditional weighted residuals plots showed symmetrical distribution and no time- or concentration-related trends. In order to evaluate the precision of estimated pharmacokinetic parameters, a non-parametric bootstrap analysis was performed and 97.1 % of the bootstrap replicates converged successfully. The bootstrap showed narrow confidence intervals for all parameters. The median, 5th
and 95th percentiles of the parameter estimates from the fit of the final model to the bootstrap samples are shown in Table 3. The asymptotic estimates obtained from the original dataset showed close agreement with the median and were all included within the 5th to 95th percentile of the bootstrapping values, indicating model stability. For the VPC, observed plasma concentration–time data and 90 % prediction intervals derived from model-simulated data are shown in Fig. 4. About 92 % of the original data fit within the 5th and 95th percentiles of the simulated datasets.
4 Discussion This is the first study to report the effect of patient demographics and clinical characteristics on the pharmacokinetics of a PARP inhibitor. Using a non-linear mixedeffect modeling approach, we developed a veliparib
484
b 1.0
1.0
0.5
0.5
ETA on CL/F
ETA on CL/F
a
0.0
−0.5
−1.0
0.0
−0.5
−1.0
50
100
150
200
50
Creatinine clearance (mL/min)
1.0
1.0
0.5
0.5
0.0
200
0.0
−0.5
−0.5
−1.0
−1.0
40
60
80
40
60
80
Age (Year)
Age (Year)
e
f 1.0
1.0
0.5
0.5
ETA on CL/F
ETA on CL/F
150
d
ETA on CL/F
ETA on CL/F
100
Creatinine clearance (mL/min)
c
0.0
0.0
−0.5
−0.5
−1.0
−1.0
40
50
60
70
80
40
Lean body mass (Kg)
50
60
70
80
Lean body mass (Kg)
g
h 0.5
ETA on Vd/F
0.5
ETA on Vd/F
Fig. 1 ETA covariate plots for the base model (left panel) and final model (right panel). a, b Post hoc ETA on CL/F vs. creatinine clearance for the base and final models, respectively. c, d Post hoc ETA on CL/F vs. age for the base and final models, respectively. e, f Post hoc ETA on CL/F vs. lean body mass for the base and final models, respectively. g, h Post hoc ETA on Vd/F vs. lean body mass for the base and final models, respectively. CL/F oral clearance, Vd/F apparent volume of distribution
A. H. Salem et al.
0.0
−0.5
0.0
−0.5
40
50
60
70
Lean body mass (Kg)
80
40
50
60
70
Lean body mass (Kg)
80
Population Pharmacokinetics of Veliparib (ABT-888)
Fig. 2 Relationship between veliparib oral clearance and temozolomide coadministration The boxes represent 25th, 50th and 75th percentiles, whiskers represent the lowest datum still within 1.5 IQR of the lower quartile, and the highest datum still within 1.5 IQR of the upper quartile range, and the open circles represent outliers. IQR interquartile range
population pharmacokinetic model in patients with different tumor types. The analysis confirmed the linearity of veliparib pharmacokinetics across the studied doses in alignment with previous reports on dose proportionality of veliparib exposure [10, 12]. The final pharmacokinetic model is a one-compartment model with first-order absorption and elimination. The estimated population mean for CL/F was 20.9 L/h when scaled to a CLCR of 100 mL/ min. The estimated population mean for the Vd/F was
485
173 L when scaled to the median LBM of the subjects (56 kg). These estimates are similar to those reported in cancer subjects taking veliparib and cyclophosphamide at various doses [10]. They are also consistent with predictions based on preclinical studies [9, 14]. Veliparib is highly permeable and has shown wide tissue distribution in mice, rats, dogs, and monkeys including the central nervous system [8, 9, 13, 14]. In addition, it has a moderate protein binding of 51 % in humans [14]. Therefore, the relatively large Vd/F estimate in this analysis, together with its high oral bioavailability, indicates the ability of veliparib to extensively distribute into human tissues. The CL/F and Vd/F indicate a veliparib half-life of 6.1 h, which supports the current twice-daily dosing schedule. Since adipose tissue has minimal contribution to clearance and unique distribution characteristics, different fractions of fat mass were proposed as predictors of the effect of body size on pharmacokinetic parameters [25]. In our study, we assessed the association between veliparib pharmacokinetic parameters and the following body size measures: WT, BSA, BMI, LBM, normal-fat mass, and FFM. LBM and FFM explained larger fractions of the inter-individual variability in Vd/F and caused greater reductions in the OFV than the other body size measure. LBM and FFM were found to explain similar magnitudes of the variability in Vd/F, and LBM was used throughout the model development because of its simple equation. LBM has been previously employed to explain interindividual variability in pharmacokinetics of anti-cancer drugs [26, 27]. Good correlations between LBM and Vd/F have been reported for hydrophilic drugs, and LBM was proposed as a better predictor of drug dosage in the obese [28]. A large proportion of the subjects included in our analysis were overweight or obese, with about 66 % of the subjects having a BMI higher than 25. Therefore, the
Table 3 Estimates of population pharmacokinetic parameters obtained after fitting the final model to the original dataset and to 1,000 bootstrap samples Parameter
Estimate (RSE%)
Bootstrap estimatea [median (95 % CI)]
CL/F (L/h)
20.9 (2.3)
20.9 (20.0–21.9)
CL/F exponent
0.48 (10.3)
0.48 (0.38–0.58)
Vd/F (L)
173 (1.7)
174 (168–179)
Vd/F exponent
0.73 (8.8)
0.72 (0.57–87)
2.4 (5.9)
2.4 (2.1–2.7)
Inter-individual variability of CL/F (CV%)
36.7 (5.2)
36.5 (32.9–40.2)
Inter-individual variability of Vd/F (CV%)
19.8 (7.8)
19.7 (16.6–22.5)
Inter-individual variability of ka (CV%)
84.0 (6.6)
83.6 (72.0–94.5)
Proportional error (SD)
0.37 (4.3)
0.37 (0.36–0.39)
First-order absorption rate constant (h-1)
RSE% percentage of relative standard error, CL/F oral clearance, Vd/F apparent volume of distribution, CV% percentage of coefficient of variation a
Based on 971/1,000 successful runs
486
A. H. Salem et al.
Fig. 3 Goodness-of-fit plots for the final model
stronger correlation of veliparib Vd/F and LBM over WT could be attributed to the inability of WT to account for differences in body composition between the subjects. Similarity in results between FFM and LBM could be explained by the minor difference between the two body size measures. LBM differs from FFM in that it includes lipids of cellular membranes, the CNS and bone marrow, which only accounts for 3–5 % of WT [29]. Veliparib CL/F was not associated with any of the body size measures. This corroborates the current fixed dosing of
veliparib and precludes the need for weight-based dosing. Veliparib CL/F was associated with CLCR but not with liver function parameters. According to our covariate model, a 10 % decrease in CLCR is associated with a 5 % increase in veliparib exposure. The impact of increases in exposure in subjects with renal impairment will depend on the maximum tolerated dose and the tolerability profile of veliparib and the coadministered drugs in the target population. In addition, the tolerability, pharmacokinetics, and pharmacodynamics of veliparib in cancer patients with
Population Pharmacokinetics of Veliparib (ABT-888)
Fig. 4 Prediction and variance-corrected visual predictive check for the final model. Circles represent prediction-corrected observations, and shaded areas represent 95 % confidence intervals of the 5th, 50th, and 95th percentiles of prediction-corrected simulated data
varying degrees of renal or hepatic dysfunction are being assessed in an ongoing study (http://clinicaltrials.gov/ct2/ show/NCT01366144). Inclusion of CLCR in the model completely accounted for the apparent association between CL/F and age in the base model. Such association is attributed to the colinearity between age and CLCR. The use of CLCR rather than age is more physiological and was supported by the predominant elimination of veliparib through the kidney. In bile ductcannulated rats and dogs, veliparib and its metabolites were primarily eliminated in urine, with 40 % of the dose as parent drug [13]. In humans, about 70 % of unchanged veliparib was recovered in the urine in 24 h at the 50-mg dose level [12]. Assuming renal clearance accounts for 70 % of veliparib CL/F, the veliparib renal clearance is much higher than the filtration clearance corrected for veliparib plasma protein binding. This indicates that active secretion plays a role in veliparib elimination. In organic cation transporter (Oct)1/Oct2 double-knockout mice, the plasma exposure of veliparib was increased 1.5-fold, and the renal clearance was decreased 1.8-fold compared with wild-type mice, indicating that veliparib active secretion may be mediated by OCTs [30]. Veliparib is being tested as a chemosenitizer to temozolomide in subjects with non-hematologic malignancies. Our analysis demonstrated that temozolomide does not affect veliparib pharmacokinetics. Therefore, no veliparib dose adjustment is recommended when used in combination with temozolomide. Other clinical studies have shown that veliparib was not affected by topotecan or
487
cyclophophamide [10, 11], which is consistent with the preclinical studies that indicated low potential of veliparib to experience drug interactions [13, 30]. This could be explained by the minor role metabolism plays in its elimination and its multiple metabolic pathways [13]. Furthermore, veliparib is a weak P-glycoprotein (P-gp) substrate and does not inhibit P-gp, organic anion transporter (OAT)1, OAT3, or OCT2 at clinically relevant concentrations [13, 30]. We observed a decline in veliparib CL/F in subjects older than 65 years of age. However, after accounting for the LBM and CLCR effects, age was not a predictor of veliparib clearance. No differences between races were observed. However, the subjects included in the analysis were predominantly White, which may not have allowed a robust assessment of the race effect. The remaining interindividual variability in veliparib CL/F and Vd/F were 37 % and 20 %, respectively. The residual unexplained variability in veliparib concentrations was 37 %, which may reflect, in addition to model misspecification, assay noise, and other unknown sources, the heterogeneity among the studied populations due to their different cancer types and various prior and current treatments.
5 Conclusion We characterized the pharmacokinetics of veliparib in patients with different types of malignancies and evaluated the influence of patient covariates and clinical characteristics on veliparib disposition. Only LBM and CLCR were found to be determinants of veliparib Vd/F and CL/F, respectively. No other clinically relevant covariates were identified as predictors of veliparib pharmacokinetics. Dosage adjustments of veliparib on the basis of body weight, age, race, temozolomide coadministration, and liver dysfunction are not necessary in patients with nonhematologic malignancies. The developed population pharmacokinetic model describes veliparib pharmacokinetics adequately and will be used to conduct simulations and evaluate veliparib exposure-response relationship. Acknowledgments This study was sponsored by AbbVie, who contributed to the study design, research, and interpretation of data, and writing, reviewing, and approving the publication. Ahmed Hamed Salem, Vincent L. Giranda, and Nael M. Mostafa are employees of AbbVie. The authors would like to thank Teresa Turner (AbbVie) for medical writing support.
References 1. Chiarugi A. Poly (ADP-ribose) polymerase: killer or conspirator? The ‘suicide hypothesis’ revisited. Trends Pharmacol Sci. 2002;23(3):122.
488 2. Vira´g L, Szabo´ C. The therapeutic potential of poly (ADP-ribose) polymerase inhibitors. Pharmacol Rev. 2002;54(3):375–429. 3. Wielckens K, Garbrecht M, Kittler M, Hilz H. ADP-ribosylation of nuclear proteins in normal lymphocytes and in low-grade malignant non-Hodgkin lymphoma cells. Eur J Biochem/FEBS. 1980;104(1):279. 4. Tomoda T, Kurashige T, Moriki T, Yamamoto H, Fujimoto S, Taniguchi T. Enhanced expression of poly (ADP-ribose) synthetase gene in malignant lymphoma. Am J Hematol. 1991;37(4): 223–7. 5. Shiobara M, Miyazaki M, Ito H, Togawa A, Nakajima N, Nomura F, et al. Enhanced polyadenosine diphosphate-ribosylation in cirrhotic liver and carcinoma tissues in patients with hepatocellular carcinoma. J Gastroenterol Hepatol. 2001;16(3): 338–44. 6. Hirai K, Ueda K, Hayaishi O. Aberration of poly (adenosine diphosphate-ribose) metabolism in human colon adenomatous polyps and cancers. Cancer Res. 1983;43(7):3441–6. 7. Palma JP, Wang YC, Rodriguez LE, Montgomery D, Ellis PA, Bukofzer G, et al. ABT-888 confers broad in vivo activity in combination with temozolomide in diverse tumors. Clin Cancer Res. 2009;15(23):7277–90. 8. Penning TD, Zhu G-D, Gandhi VB, Gong J, Liu X, Shi Y, et al. Discovery of the poly (ADP-ribose) polymerase (PARP) inhibitor 2-[(R)-2-methylpyrrolidin-2-yl]-1 H-benzimidazole-4-carboxamide (ABT-888) for the treatment of cancer. J Med Chem. 2008;52(2):514–23. 9. Donawho CK, Luo Y, Penning TD, Bauch JL, Bouska JJ, Bontcheva-Diaz VD, et al. ABT-888, an orally active poly(ADPribose) polymerase inhibitor that potentiates DNA-damaging agents in preclinical tumor models. Clin Cancer Res. 2007;13(9):2728–37. 10. Kummar S, Ji J, Morgan R, Lenz HJ, Puhalla SL, Belani CP, et al. A phase I study of veliparib in combination with metronomic cyclophosphamide in adults with refractory solid tumors and lymphomas. Clin Cancer Res. 2012;18(6):1726–34. 11. Kummar S, Chen A, Ji J, Zhang Y, Reid JM, Ames M, et al. Phase I study of PARP inhibitor ABT-888 in combination with topotecan in adults with refractory solid tumors and lymphomas. Cancer Res. 2011;71(17):5626–34. 12. Kummar S, Kinders R, Gutierrez ME, Rubinstein L, Parchment RE, Phillips LR, et al. Phase 0 clinical trial of the poly (ADPribose) polymerase inhibitor ABT-888 in patients with advanced malignancies. J Clin Oncol. 2009;27(16):2705–11. 13. Li X, Delzer J, Voorman R, de Morais SM, Lao Y. Disposition and drug-drug interaction potential of veliparib (ABT-888), a novel and potent inhibitor of poly(ADP-ribose) polymerase. Drug Metab Dispos. 2011;39(7):1161–9. 14. Muscal JA, Thompson PA, Giranda VL, Dayton BD, Bauch J, Horton T, et al. Plasma and cerebrospinal fluid pharmacokinetics of ABT-888 after oral administration in non-human primates. Cancer Chemother Pharmacol. 2010;65(3):419–25. 15. Lindbom L, Pihlgren P, Jonsson N. PsN-Toolkit—a collection of computer intensive statistical methods for non-linear mixed effect modeling using NONMEM. Comput Methods Programs Biomed. 2005;79(3):241–57.
A. H. Salem et al. 16. Beal S, Sheiner L, Boeckmann A, Bauer R. NONMEM user’s guide. Ellicott City: Icon Development Solutions; 1989–2009. 17. Jonsson EN, Karlsson MO. Xpose–an S-PLUS based population pharmacokinetic/pharmacodynamic model building aid for NONMEM. Comput Methods Programs Biomed. 1999;58(1): 51–64. 18. Cheymol G. Effects of obesity on pharmacokinetics. Clin Pharmacokinet. 2000;39(3):215–31. 19. Duffull SB, Dooley MJ, Green B, Poole SG, Kirkpatrick CM. A standard weight descriptor for dose adjustment in the obese patient. Clin Pharmacokinet. 2004;43(15):1167–78. 20. Janmahasatian S, Duffull SB, Ash S, Ward LC, Byrne NM, Green B. Quantification of lean bodyweight. Clin Pharmacokinet. 2005;44(10):1051–65. 21. Cockcroft DW, Gault MH. Prediction of creatinine clearance from serum creatinine. Nephron. 1976;16(1):31–41. 22. Mosteller R. Simplified calculation of body-surface area. N Engl J Med. 1987;317(17):1098. 23. Savic RM, Karlsson MO. Importance of shrinkage in empirical bayes estimates for diagnostics: problems and solutions. AAPS J. 2009;11(3):558–69. 24. Bergstrand M, Hooker AC, Wallin JE, Karlsson MO. Predictioncorrected visual predictive checks for diagnosing nonlinear mixed-effects models. AAPS J. 2011;13(2):143–51. 25. Anderson BJ, Holford NH. Mechanistic basis of using body size and maturation to predict clearance in humans. Drug Metabol Pharmacokinet. 2009;24(1):25–36. 26. Ekhart C, Rodenhuis S, Schellens JH, Beijnen JH, Huitema AD. Carboplatin dosing in overweight and obese patients with normal renal function, does weight matter? Cancer Chemother Pharmacol. 2009;64(1):115–22. 27. Prado CM, Lima IS, Baracos VE, Bies RR, McCargar LJ, Reiman T, et al. An exploratory study of body composition as a determinant of epirubicin pharmacokinetics and toxicity. Cancer Chemother Pharmacol. 2011;67(1):93–101. 28. Morgan D, Bray K. Lean body mass as a predictor of drug dosage: implications for drug therapy. Clin Pharmacokinet. 1994;26(4):292. 29. Roche A, Heymsfield S, Lohman T, editors. Human body composition. Champaign (IL): Human Kinetics Press; 1996. 30. Kikuchi R, Lao Y, Bow DA, Chiou WJ, Andracki ME, Carr RA, et al. Prediction of clinical drug–drug interactions of veliparib (ABT-888) with human renal transporters (OAT1, OAT3, OCT2, MATE1, and MATE2K). J Pharm Sci. 2013;102(12):4426–32. 31. Mehta MP, Curran WJ, Wang D, Wang F, Kleinberg L, Brade A, et al. Phase 1 safety and pharmacokinetic (PK) study of veliparib in combination with whole brain radiation therapy (WBRT) in patients (PTS) with brain metastases. J Thorac Oncol. 2012;84(3S):S269. 32. Middleton M, Friedlander P, Hamid O, Daud A, Plummer R, Schuster R, et al. Efficacy of veliparib (ABT-888) plus temozolomide versus temozolomide alone: a randomized, doubleblind, placebo-controlled trial in patients with metastatic melanoma. Eur J Cancer. 2011;47:8.
