Population Pharmacokinetics of Volasertib Administered in Patients with Acute Myeloid Leukaemia as a Single Agent or in Combination with Cytarabine

Bele´n P. Solans1,2 • Ange`le Fleury3 • Matthias Freiwald3 • Holger Fritsch3 •
Karin Haug3 • In˜aki F. Troco´niz1,2

© Springer International Publishing Switzerland 2017

Abstract
Background Volasertib, a potent and selective polo-like kinase inhibitor, has shown to increase response rates and improve survival with a clinically manageable safety pro- file, administered alone and in combination with cytarabine in patients with acute myeloid leukaemia.
Objectives The objectives of this analysis were to describe the pharmacokinetics of volasertib and cytarabine, admin- istered as single agents or in combination.

Electronic supplementary material The online version of this article (doi:10.1007/s40262-017-0566-9) contains supplementary material, which is available to authorized users.

& In˜aki F. Troco´niz [email protected]

1 Pharmacometrics and Systems Pharmacology, Department of Pharmacy and Pharmaceutical Technology, School of Pharmacy and Nutrition, University of Navarra, Irunlarrea 1, 31008 Pamplona, Spain
2 Navarra Institute for Health Research (IdisNA), University of Navarra, Pamplona, Spain
3 Translational Medicine and Clinical Pharmacology, Boehringer Ingelheim GmbH & Co. KG,
Biberach an der Riss, Germany

Methods Three thousand, six hundred and six plasma volasertib concentrations from 501 patients receiving either volasertib alone, or in combination with cytarabine, and 826 plasma cytarabine concentrations from 650 patients receiving cytarabine as multiple subcutaneous injections per cycle either alone, or in combination with volasertib, were analysed using NONMEM Version 7.3. Covariates evaluated included demographic and disease-related parameters.
Results The pharmacokinetics of volasertib were found to be dose independent from 150 to 550 mg. Body surface area and ethnicity showed significant effects in all the patients. This is reflected as an increase in drug exposure for Japanese patients, although this finding has to be interpreted with caution because only 7% of patients were part of that population group. Volasertib showed low-to- mild inter-individual variability in total clearance. For the case of cytarabine, its pharmacokinetics was affected by body surface area. Finally, volasertib and cytarabine did not influence the pharmacokinetic characteristics of each other.
Conclusions The pharmacokinetics of volasertib in patients with acute myeloid leukaemia alone or in combi- nation with cytarabine is predictable and associated with low-to-mild patient variability with the exception of the high variability associated with the volume of distribution of the central compartment, having no effect on the area under the plasma concentration–time curve.

Key Points
To the best of our knowledge, the pharmacokinetic characteristics of polo-like kinase 1 inhibitors in large target patient populations have not been reported so far.
Regarding the anti-cancer drug cytarabine, despite its common use in acute myeloid leukaemia, available information about its pharmacokinetic properties is very scarce, especially after multiple low doses injected subcutaneously.
This population analysis is of special relevance owing to the lack of models describing volasertib pharmacokinetics, and the possibility of using both volasertib and cytarabine models to predict the effect on white blood cell counts of patients receiving this therapy.

1 Introduction

Acute myeloid leukaemia (AML) is a heterogeneous haematological disorder characterised by clonal expansion of myeloid progenitors in the bone marrow and peripheral blood. This leads to infiltration and subsequent accumula- tion of cells at various stages of incomplete maturation in the bone marrow, peripheral blood and other tissues. The production of healthy haematopoietic elements is reduced [1]. The increased proliferation of these immature cells leads to cytopenias, representing a hurdle in treatment management [2].
The incidence of AML increases with age, with a median age at diagnosis of 65–70 years. The number of new cases of AML was 4 per 100,000 people per year, being more common in men. AML can be cured in 35–40% of patients aged 60 years or younger, and in 5–15% of patients aged older than 60 years [3].
Treatment in AML has changed little over the last 40 years, being the conventional treatment divided in two phases—induction, which aims for complete remission, and consolidation, which includes stem-cell transplanta- tion, designed to eliminate leukaemia cells that persist after induction [1]. For those not eligible for stem-cell trans- plantation, the standard therapy consists of induction with 7 days of cytarabine plus 3 days of an anthracycline, fol- lowed by consolidation with additional chemotherapy. Despite intensive therapy, many patients relapse with poor prognosis [2, 4]. Thus, the majority of patients are unable

to tolerate intensive chemotherapy without unaccept- able side effects, with a median survival of only 6–10 months; therefore, alternatives to the standard treat- ment are being sought [5, 6].
Cell-cycle inhibitors are particularly useful in rapidly proliferating malignancies, such as AML. In this context, polo-like kinases (PLKs) are increasingly recognised as key regulators of cell division tightly correlated with cell proliferation [7, 8]. It has been shown that PLK1 is highly expressed in leukemic cell lines from patients with AML compared with healthy patients [9].
As a consequence, PLK inhibitors represent a promising new class of anticancer drugs because their inhibition would translate into apoptotic death [10]. Volasertib (BI 6727) is one of the selective and potent PLK inhibitors [2]. Volasertib potently inhibits PLK1 (half-maximal inhibitory concentration of 0.87 nmol/L), but shows no inhibitory activity against other unrelated kinases at concentrations up to 10 nmol/L [11]. Preclinical studies of volasertib using a variety of tumour cell lines, including haematopoietic malignancies, have proven the inhibition of cell division resulting in cell death [12, 13]. In addition, volasertib has shown anti-leukaemic efficacy, either as a single agent or in combination regimens in multiple preclinical models of AML, including bone marrow samples in short-term cul- ture as well as subcutaneous in-vivo models in immune- deficient mice [14]. Volasertib has also been investigated in several clinical trials, increasing the response rate and improving survival with a clinically manageable safety profile [15, 16].
The objective of the current analysis was to develop a population model characterising the pharmacokinetic pro- files of volasertib and cytarabine in patients with AML, quantifying the magnitude of inter-individual variability, identifying the patient characteristics that might have a significant impact on either the volasertib or cytarabine pharmacokinetic profile and evaluating a possible drug– drug interaction between the two drugs at the pharma- cokinetic level.

2 Materials and Methods

All patients provided written informed consent consistent with the International Conference on Harmonisation of Technical Requirements for Registration of Pharmaceuti- cals for Human Use—Good Clinical Practice and local legislation. All studies were performed in accordance with the Declaration of Helsinki and were approved by the institutional review board of the ethics committee at each study site.

2.1 Overview of Study Designs

2.1.1 Study 1230.4 (NCT00804856)

Study 1230.4 is an open-label phase I/IIa, multicenter, international, randomised, dose-escalation study in patients with AML. One hundred and seventy-five patients were enrolled in this study. In the phase I part of the study, different doses with dose escalation of volasertib were given on days 1 and 15 in combination with low-dose cytarabine (LDAC) or alone. The objectives of phase IIa were to evaluate the efficacy and safety of LDAC plus volasertib vs. LDAC alone in na¨ıve patients with AML.

2.1.2 Study 1230.14 (NCT01721876)

Study 1230.14 is a phase III, randomised, double-blind, multicenter, international controlled parallel-group study of the administration of volasertib in combination with LDAC vs. placebo plus LDAC in patients with AML aged 65 years and above and ineligible for intensive induction therapy. The study was conducted in 666 patients, of whom 533 were treated. Objectives were to evaluate the efficacy and safety of the treatments.

2.1.3 Study 1230.26 (NCT01662505)

Study 1230.26 is an open-label phase I trial, with dose escalation of volasertib every 2 weeks in Japanese patients with AML. This study enrolled 19 patients. The objective was to determine the maximum tolerated dose in Japanese patients with AML.

2.2 Drug Administration and Sample Collection

In studies 1230.4 and 1230.14, volasertib was supplied as a 1-h intravenous infusion on days 1 and 15 of a 28-day cycle. In the combination treatment arms, on days 1–10 of each cycle, patients received 20 mg of subcutaneous LDAC twice daily. In study 1230.26, volasertib was supplied as a 2-h intravenous infusion. In study 1230.4, patients received either volasertib, LDAC or a combination of both. Doses of volasertib ranged from 150 to 550 mg. In study 1230.14, patients received LDAC combined either with 350 mg of volasertib or placebo. In study 1230.26, the patients were treated with 350, 400 or 450 mg of volasertib.
Blood samples to determine the concentration of vola- sertib or cytarabine in plasma were obtained in all of the studies prior to administration. For the quantification of volasertib, 3 mL of venous blood was taken in an ethylendiaminetetraacetic acid-anticoagulant blood-draw- ing tub on every sample taken from the patients, whereas for the quantification of cytarabine, 5 mL of venous blood

was taken in a lithium heparin-anticoagulant blood-draw- ing tube on every sample taken from the patients. In studies 1230.4 and 1230.26, at least seven samples were taken within the first 4 days after volasertib administration. In study 1230.14, plasma samples were obtained shortly before the end of the first and second infusion of volasertib. Table 1 lists the number of patients treated per dose level, and a detailed description of sampling time among other relevant information.

2.3 Analytical Determination of Volasertib and Cytarabine in Plasma

Plasma concentrations of volasertib and cytarabine were determined by a validated high-performance liquid chro- matography, tandem mass spectrometry (HPLC–MS/MS) assay. Volasertib analysis was performed at Boehringer Ingelheim Pharma GmbH & Co. KG, Drug Metabolism and Pharmacokinetics, Biberach, Germany. Cytarabine analysis was performed at SGS Cephac Europe, Saint- Benoˆıt Cedex, France. Volasertib was analysed by HPLC- MS/MS using [D3]volasertib as the internal standard; cytarabine was analysed by HPLC-MS/MS using [13C3]- cytarabine as the internal standard. Both drug samples were subjected to solid-phase extraction in the 96-well plate format followed by reversed-phase HPLC with gradient elution, and were detected and quantified by MS/MS using electrospray ionisation in the positive ion mode. The limits of quantification for volasertib and cytarabine were, respectively, 0.20 and 10.0 ng/mL.

2.4 Brief Description of the Data

A total of 3606 plasma volasertib concentrations from 501 patients receiving either volasertib alone, or in combination with cytarabine were used in the analysis. Less than 0.5% of the total concentrations was reported as below the limit of quantification (BLQ), and was therefore ignored.
A total of 826 values of cytarabine concentrations were obtained from 650 patients receiving cytarabine alone, or in combination with volasertib; 338 of those were reported as BLQ, and were taken into account during the analysis, as described below (Sect. 2.4). Table 1S of the Electronic Supplementary Material (ESM) lists the number of patients per study providing at least one concentration of volasertib and/or cytarabine.

2.5 Data Analysis

NONMEM Version 7.3 [17] software (Icon Development Solutions, Ellicott City, MD, USA) was used for the pop- ulation analysis. During model building, volasertib and cytarabine data were analysed separately. Once the

Table 1 Study design characteristics

Study no.

Design Patient
no.

Dose levela

1230.4 Phase I/IIa – open-label, dose-escalation interventional study Dosing and sampling schedule:
Volasertib: 1-h infusion on days 1 and 15 of a 28-day cycle
Cytarabine: subcutaneous injection twice daily (2 9 20 mg) on days 1–10 in a 28-day cycle Sample collection: 8 samples were taken within the 4 first days after Volt administration, at the
following times after start of Volt infusion in the first cycle: –0:05, 30 min, 1, 1.5, 2, 3, 4–6, 24 h, 4
and 13 days. After the second cycle, samples were taken 1 h after the start of drug administration. 7 samples were taken within the 4 first hours after LDAC administration in the first cycle: -0:05, 30 min, 1, 1.5, 2, 3 and 4 h. After the second cycle, samples were taken 1 h after the drug administration

1230.14 Phase III: randomised, double-blind, controlled, parallel-group interventional study Dosing and sampling schedule:
Volasertib: 1-h infusion on days 1 and 15 of a 28-day cycle
Cytarabine: subcutaneous injection twice daily (2 9 20 mg) on days 1–10 in a 28-day cycle Sample collection: samples were taken 50 min after start of drug administration, for either Volt or
cytarabine

1230.26 Phase I: open-label, uncontrolled, dose-escalation interventional study Dosing and sampling schedule:
Volasertib: 2-h infusion on days 1 and 15 of a 28-day cycle
Sample collection: 7 samples were taken within the 4 first days after start of Volt administration, at the following times: -0:05, 1, 2, 3, 4, 8, 24 h, 4, 6 and 9 days
LDAC low-dose cytarabine, Volt volasertib
a The number of patients is represented in round brackets

175 Phase I Schedule A (32):
150 mg
Volt ? LDAC (4)
200 mg
Volt ? LDAC (3)
250 mg
Volt ? LDAC (5)
300 mg
Volt ? LDAC (9)
350 mg
Volt ? LDAC (8)
400 mg
Volt ? LDAC (3)
Phase I Schedule B (56):
150 mg Volt (11)
200 mg Volt (2)
350 mg Volt (5)
400 mg Volt (6)
450 mg Volt (23)
500 mg Volt (5)
550 mg Volt (4)
Phase II (87):
350 mg
Volt ? LDAC (42)
LDAC (45)
533 Treatment arm A (356):
350 mg
Volt ? LDAC Treatment arm B
(177):
Placebo ? LDAC
19 350 mg Volt (7) 400 mg Volt (4) 450 mg Volt (8)

selected models were achieved for both drugs separately, a final analysis was performed on all data together exploring possible covariance of random effects of the two

compounds. Data were logarithmically transformed. Inter- individual variability (IIV) was modelled exponentially. Non-diagonal elements of the X variance-covariance

matrix were tested for significance. Residual error was described with an additive model in the logarithmic domain. Drug concentrations reported as BLQ were con- sidered as censored information and modelled on the basis of the M3 method [18] using the F_Flag functionality and the PHI function as described by Ahn et al. [19]. The first- order conditional estimation method with INTERACTION or LAPLACIAN was used in the case of dealing with just continuous data or continuous plus censored data, respectively.

2.5.1 Model Selection Criteria

Selection between models was mainly based on the mini-

usual dose adjustments in the clinic, in this case BSA, was used to study the covariate effects. The stepwise covariate model building procedure [21] implemented in the Pearl Speaks NONMEM software [22] was used to obtain the statistically significant covariates. The stepwise covariate model procedure is based on a forward inclusion followed by a backward deletion approach. During the forward inclusion and the backward deletion approaches, the levels of significance used to incorporate the model and to keep a covariate in the model were set to 0.05 and 0.001, respectively.
Covariates that showed significance were included in the parameters using the following general covariate model:

mum value of the objective function provided by NON-

Ym .

covm

Σhm

Yp .

ctg !

MEM, which is equal to -2 9 Log likelihood (-2LL);
-2LL differences of 3.84, 7.88 and 10.83 are considered

TVP ¼ hn ×

cov

m;ref

1 hp;ctg
1 2

ð1Þ

significant at the 5, 0.5 and 0.1% levels, respectively, for
nested models differing in one parameter, and visual exploration of the goodness-of-fit plots.

2.5.2 Model Development

First, the base models for volasertib and cytarabine were selected, and then the search for the significant covariate relationship was followed by the model evaluation exercise.

2.5.2.1 Base Population Model Disposition of volasertib and cytarabine in plasma was characterised using com- partmental models parameterised in apparent distribution volumes and distribution and elimination clearances. For volasertib, the presence of dose-dependent pharmacoki- netics was explored in all pharmacokinetic parameters.
Different absorption models including first- and zero- order processes, and transit compartments were considered to describe cytarabine input into the central compartment. Time dependencies in the pharmacokinetics of volasertib and cytarabine were explored empirically including time effects on the absorption and disposition parameters.

2.5.2.2 Covariate Selection Patient characteristics including genetic subgroups are listed in Table 2. The genetic subgroups relate to mutations in certain genes that are directly related to disease progression and response to treatment [20]. The interaction between volasertib and cytarabine was explored through the binary covariate combination therapy, which corresponded to 0 when the drugs were given alone or 1 when the two drugs were given in combination.
For those covariates that were correlated between them, as it was the case for weight, height and body surface area (BSA), only the most relevant covariate with regard to

where the typical value of a model parameter (TVP) is
described as a function of m continuous (covm) and p cat- egorical covariates including several categories (ctg). hn describes the nth typical parameter value for an individual with covariate values (covm) equal to the reference values: [(covm = covm;ref) and ctg = 1]. covm;ref refers usually to the median value across the studied populations. hm and hp;ctg are parameters quantifying the magnitude of the covariate-parameter relationship.

2.5.2.3 Model Evaluation Parameter precision was fur- ther evaluated performing 500 non-parametric bootstrap analyses using Pearl Speaks NONMEM [22], stratified by ethnicity for volasertib and study for cytarabine (to keep the same proportion of Japanese subjects, and those receiving cytarabine either alone or in combination, in each bootstrap dataset), and listing the 2.5th, 50th, and 97.5th percentiles of each parameter distribution.
Model performance was evaluated by performing pre- diction-corrected visual predictive checks [23]. A total of
500 datasets with the same study characteristics as the original were simulated. The 2.5th, 50th and 97.5th per- centiles of the simulated observations in each dataset were computed for all time intervals and the 95% prediction interval of each calculated percentile was obtained and plotted against the 2.5th, 50th, and 97.5th percentiles obtained from the raw data.

2.5.3 Model Exploration

Exposure measurements were derived from the selected population pharmacokinetic models, to highlight the clin- ical relevance of the included covariates. Different covariate groups were determined according to the covariates included, and for each group, 500 individual pharmacokinetic profiles were simulated. The following

Table 2 Summary of patient characteristicsa

1230.4 1230.14b 1230.26
Number of patients 175 533 19
Age (years)
Sexc 72.47 (7.70) [26–87] 75.24 (5.12) [65–93] 72.21 (7.66) [53–86]
Male 96 (54.86) 315 (59.10) 12 (63.16)
Female 79 (45.14) 218 (40.90) 7 (36.84)
Weight (kg) 72.94 (14.1) [43–118] 71.73 (14.18) [35.4–129.4] 55.4 (11.31) [40.4–88.4]
Height (cm) 168.46 (8.72) [142–192] 165.74 (9.14) [143–191] 160 (8.63) [144–176]
Body surface area (cm2)
Ethnicityc 1.82 (0.19) [1.42–2.39] 1.79 (0.20) [1.22–2.53] 1.56 (0.18) [1.34–2.02]
Caucasian and other non-Asian 171 (97.7) 448 (84.05)
Japanese 32 (6.00) 19 (100)
Other Asian 4 (2.2) 53(9.94)
Combination therapy (volasertib ? cytarabine)c
Yes 74 (42.28) 356 (66.79)
No 101 (57.71) 177 (33.21) 19 (100)
Genetic subgroupsc,d
Favourable 10 (5.71) 57 (10.69)
Intermediate I 83 (47.43) 208 (39.02)
Intermediate II 39 (22.29) 99 (18.57)
Adverse 43 (24.57) 169 (31.71)
a Values are represented as mean (SD) [range] unless specified otherwise
b Showing demographics in patients randomised C5 months at the time of the clinical cut-off
c Values are represented as number (%)
d Values for study 1230.26 are not specified

metrics were derived: maximum plasma concentration and area under the plasma concentration–time curve in the first cycle (AUC0–28). To evaluate differences in terms of drug bioequivalence, both metrics were compared with the 80–125% range of the distribution obtained from the group with covariate values equal to the median of the patient population studied.

3 Results

3.1 Data

Figure 1a shows the raw observed concentration data obtained for volasertib, and Fig. 1b shows the raw observed concentration data obtained for cytarabine, showing the BLQ concentrations.

3.2 Base Population Model

3.2.1 Volasertib

The plasma concentration vs. time profiles of volasertib could be best described by a four-compartment model

(Fig. S1A of the ESM). We obtained a drop in the mini- mum objective function of 238.475 points compared with the three-compartment model, when the fourth compart- ment was incorporated. The pharmacokinetics of volasertib was found to be time independent.
The typical estimate for total plasma clearance (CL) was
58.4 L/h. With respect to the apparent volumes of distri- bution of the central compartment (V1), second, third and fourth peripheral compartments (V2, V3 and V4), the estimates were 82.5, 2830, 6750 and 263 L, respectively. For a typical patient, the value of the distribution clear- ances between the central and the first, second and third peripheral compartments (Q2, Q3 and Q4) were, respec- tively 277, 43.9 and 254 L/h. Dose was found to affect Q3, and therefore, a parameter modifying the typical estimate of Q3, modulated by the dose given, was introduced in the model.
Inter-individual variability was included on CL, V1, V3, Q2 and Q3 (p \ 0.01) with a value of 34, 87, 41, 31 and 46%, respectively. Non-diagonal elements of the X vari- ance–covariance matrix were non-significant (p [ 0.05).
A different magnitude of residual error was estimated for the observations gathered during the phase III study with respect to the other two studies (0.24 vs. 0.31).

Fig. 1 Observed volasertib (a) and cytarabine (b) concentrations vs. time after dose (TAD). a Blue dots represent observed volasertib concentrations. b Grey dots represent observed cytarabine

concentrations and red dots represent below the limit of quantification observations. LN logarithmic transformation

3.2.2 Cytarabine

Attempts to model the absorption yielded a poorer description of the data than the model assuming an intra- venous bolus administration (p \ 0.001). Inclusion of a latency time did not improve the fit significantly (p [ 0.05).
Multi-compartmental disposition models (e.g. two-
compartment model) were non-significantly better than the simpler one compartment model (p [ 0.05). The structure of the model selected for cytarabine is represented in Fig. S1B of the ESM. Inter-individual variability was supported on total apparent plasma clearance (CL/F) and apparent volume of distribution of the central compartment (V/F). Typical estimate of the absolute bioavailability (F1) was fixed to 1 owing to the absence of data after intra- venous administration. Inclusion of IIV on F1 was non- significant (p [ 0.05). Typical estimates corresponding to CL/F and V/F were 207.9 L/h and 208.9 L, respectively. The estimates of IIV were 33% (CL/F) and 53% (V/F). A different magnitude of residual error was estimated for the observations gathered during the phase III study with respect to the other studies (0.42 vs. 0.19).

3.3 Covariate Model

3.3.1 Volasertib

The set of covariates tested for each parameter in the model were sex, age, ethnicity, BSA, combination therapy and genetic subgroups. The selected full covariate model comprised the following covariate effects: (1) ethnicity on CL, V1, V3, Q2 and Q3; and (2) BSA on CL, V1, V2, V3, V4, Q2, Q3 and Q4.

3.3.2 Cytarabine

Sex, age, ethnicity, genetic subgroups, combination ther- apy and BSA were tested for significance on CL/F and V/
F. After the stepwise covariate model analysis, BSA was significant on CL/F and V/F.

3.4 Final Model Evaluation

Table 3 lists the estimates of the model parameters corre- sponding to the final selected models for both volasertib and cytarabine. All parameters were obtained with

Table 3 Population pharmacokinetic parameter estimates of volasertib and cytarabine after intravenous and subcutaneous administrations, respectively

Parameter Estimate (RSE%) [2.5th–97.5th] IIV (RSE%) [SHR%] [2.5th– 97.5th]
Volasertib
CL (L/h) = HCL 9 (1 ? HETHNCL) 9 (BSA/1.77)HBSACL HCL = 58.40 (2.63) [54.89–61.21] 35 (21.64) [22]
[28–43]
HETHNCL = -0.31 [-0.37 to -0.21]
(13.98) (Jap)
HETHNCL = 0 (non-Jap) [0.68–1.28]
HBSACL = 0.91 (16.88)
V1 (L) = HV1 9 (1 ? HETHNV1) 9 (BSA/1.77)HBSAV1 HV1 = 82.5 (17.62) [59.64–117.43] 87 (31.44) [47] [55–121]
HETHNV1 = -0.39 [-0.51 to -0.21]
(20.17) (Jap)
HETHNV1 = 0 (non-Jap) [0.93–1.82]
HBSAV1 = 1.43 (17.14)
Q2 (L/h) = HQ2 9 (1 ? HETHNQ2) 9 (BSA/1.77)HBSAQ2 HQ2 = 277 (4.14) [255.87–299.64] 31 (23.31) [50] [22–37]
HETHNQ2 = -0.31 [-0.37 to -0.21]
(13.98) (Jap)
HETHNQ2 = 0 (non-Jap) [0.68–1.28]
HBSAQ2 = 0.91 (16.88)
V2 (L) = HV2 9 (BSA/1.77)HBSAV2 HV2 = 2830 (8.78) [2432.56–3468.48] NE
HBSAV2 = 1.43 (17.14) [0.93–1.82]
Q3 (L/h) = HQ3 9 (1 ? HETHNQ3) 9 (BSA/ HQ3 = 43.9 (11.25) [33.52–51.81] 45 (26.01) [53] [31–56]

1.77)HBSAQ3 9 (DOSE/350) HDOSE

HETHNQ3

= -0.31

[-0.37 to -0.21]

(13.98) (Jap)

[0.68–1.28]

HETHNQ3 = 0 (non-Jap)
HBSAQ3 = 0.91 (16.88)
HDOSE = -1.2 (15.09) [-1.57 to -0.82]
V3 (L) = HV3 9 (1 ? HETHNV3) 9 (BSA/1.77)HBSAV3 HV3 = 6750 (7.53) [5676.01–7636.85] 41 (39.12) [56] [22–54]
HETHNV3 = -0.39 [-0.51 to -0.21]
(20.17) (Jap)
HETHNV3 = 0 (non-Jap) [0.93–1.82]
HBSAV3 = 1.43 (17.14)
Q4 (L/h) = HQ4 9 (BSA/1.77)HBSAQ4 HQ4 = 254 (7.22) [216.23–288.97] NE
HBSAQ4 = 0.91 (16.88) [0.68–1.28]
V4 (L) = HV4 9 (BSA/1.77)HBSAV4 HV4 = 263 (9.64) [206.22–315.57] NE
HBSAV4 = 1.43 (17.14) [0.93–1.82]
Residual error [log(nmol/L)] 0.30 (5.56) [0.27–0.34] NE
Residual error: phase III [log(nmol/L)] 0.24 (17.89) [0.15–0.32] NE
Cytarabine
CL/F (L/h) = HCL 9 (BSA/1.77)HBSACL HCL = 208.73 (4.76) [190.63–229.64] 33 (30.86) [55] [22–44]
HBSACL = 1.78 (16.49) [1.21–2.39]
V/F (L) = HV 9 (BSA/1.77)HBSAV HV = 209.25 (7.94) [180.83–245.94] 53 (42.37) [52] [33–74]
HBSAV = 2.24 (24.56) [1.25–3.45]
Residual error [log(nmol/L)] 0.19 (12.28) [0.14–0.23] NE
Residual error: phase III [log(nmol/L)] 0.42 (10.89) [0.34–0.52] NE
Results from 500 bootstrap analysis are shown in square brackets and with the 95% confidence interval. Parameters and covariate terms are defined in the text
IIV inter-individual variability expressed as coefficient of variation, NE not estimated, Jap Japanese patients, RSE relative standard error, SHR
shrinkage

reasonable precision. Goodness-of-fit plots for volasertib (Fig. S2 of the ESM) show no apparent tendencies indi- cating the absence of major model misspecifications. Pre- diction-corrected visual predictive checks for volasertib, volasertib stratified by dose and cytarabine are shown in Figs. 2, 3 and 4, respectively. The models of both com- pounds performed adequately, and were able to describe the central tendency of the data and the spread. For cytarabine, the selected model was capable of handling BLQ data quite well (Fig. 4, bottom panel). For the case of volasertib, the profiles show, in general, an over-prediction of the variability of the lower plasma values.
For volasertib, mean values of CL and V1 in a reference Caucasian patient were 58.4 L/h and 82.5 L, respectively, while the corresponding values in a typical Japanese patient were 35.4 L/h, and 41.2 L, respectively. However, the number of Japanese patients compared with the number of non-Asian patients is low (35 vs. 423) and, therefore, this result should be interpreted with caution and needs further investigation in future studies.
With respect to the covariate BSA, CL is increased by 8.9% in the 75th percentile of BSA, and by 12.77% in the

95th percentile of BSA in Caucasian patients with respect to the typical Caucasian patient with the 50th percentile of BSA. With respect to the total apparent volume of distri- bution, it is increased by 13.16 and 20.79% in the 75th and 95th percentiles, respectively.

3.5 Model Exploration

Box plots exploring the impact of the selected covariates on drug exposure for subgroups in volasertib and cytara- bine are provided in Fig. 5. For both exposure measure- ments, median values and ranges for volasertib are higher for Japanese patients while for cytarabine no such tendency is seen. Furthermore, for volasertib, the interquartile range of these patients’ exposure measurements are outside the bioequivalence range (80–125%), taking the exposure of the typical non-Japanese patient as the reference.
In Fig. 6, in addition, the drug concentration vs. time after dose profiles for volasertib (Fig. 6a) and cytarabine (Fig. 6b) are depicted for typical subjects representing different values for each selected covariate. Differences within concentration profiles are greater for volasertib than

Fig. 2 Results of the prediction-corrected visual predictive checks from 500 simulated profiles for volasertib. Blue points represent raw data and blue lines correspond to the 2.5th, 50th and 97.5th percentiles. Grey shaded areas represent the 95% prediction intervals of the 2.5th, 50th, and 97.5th percentiles of 500 simulated datasets. LN logarithmic
transformation, TAD time after dose

Fig. 3 Results of the prediction-corrected visual predictive checks from 500 simulated profiles for volasertib stratified by dose. Blue points represent raw data and blue lines correspond to the 2.5th, 50th and 97.5th percentiles. Grey shaded areas represent the 95%

prediction intervals of the 50th percentile of 500 simulated datasets. Grey dotted lines correspond to the median of the 2.5th and 97.5th percentile of the simulated datasets. LN logarithmic transforma- tion, TAD time after dose

cytarabine; however, in both cases, the range of the typical profiles are inside the predicted variability associated with the typical individual in the population (Caucasian, and BSA = 1.77).
Finally, a last modelling exercise was performed ana- lysing together the pharmacokinetic data from volasertib and cytarabine to look for covariance in the parameter distributions. The results indicated an absence of signifi- cant covariance (p \ 0.05).

4 Discussion

The current evaluation reports for the first time the popula- tion pharmacokinetics of volasertib in patients with AML. From this analysis, three main properties were extracted. First, multi-compartmental disposition characterised by a large apparent volume of distribution of almost 10,000 L suggests that binding to tissue components exceeds the binding to plasma proteins. Volasertib distributes to the lysosomes of the cells, and it is a cationic amphiphilic compound with a tendency to accumulate in acidic

organelles. This fact can somehow explain the large volume of distribution of this compound. Interestingly, the apparent volume of distribution of BI2536, the first PLK1 inhibitor studied in clinics, was also estimated to be large (1600 L), although lower than volasertib [24]. The larger volume of distribution and longer half-life of volasertib, probably as a consequence of redistribution from either destroyed cells or loaded cytosomes, compared with other drugs used in the standard treatment of AML, such as idarubicin or daunoru- bicin [25], can lead to prolonged exposure in malignant cells, and therefore, higher treatment efficacy.
Second, results from this population analysis revealed that the magnitude of IIV is low to moderate but not high, with the exception of IIV in V1. Regarding predictability of exposure, that characteristic represents a benefit with respect to other drugs used in the treatment of AML. Pre- vious work reporting the pharmacokinetic profile of other PLK-inhibitory drugs investigated in AML, such as BI 2536 [24], defined a variability associated with PK parameters was somewhat higher (49–134%) than those described in the current report (35–41%), comparing sim- ilar parameters.

Fig. 4 Results of the prediction-corrected visual predictive checks from 500 simulated profiles for cytarabine. Upper panel: red points represent raw data and red lines correspond to the 2.5th, 50th and 97.5th percentiles. The grey shaded area represents the 95% prediction intervals of the 50th percentiles of 500 simulated datasets. Grey lines represent the median of the 2.5th–50th–97.5th percentiles of 500 simulated datasets. Horizontal grey line represents the value of

the lower limit of quantification. Lower panel: red open circles represent the percentage of raw data reported as a measurement below the lower limit of quantification (BLQ). The grey shaded area represents the 95% prediction intervals of percentages of the simulated observations below the BLQ from the 500 simulated datasets. At greater times, 100% of observations were BLQs, as predicted by the model. TAD time after dose

Fig. 5 Box plots after 500 simulations of patients receiving either volasertib (a) or cytarabine (b) in the following covariate groups: 1 Japanese patients with a body surface area (BSA) equal to 1.53 m2; 2 Japanese patients with a BSA equal to 1.77 m2; 3 non-Japanese patients with a BSA equal to the 95th percentile of the BSA distribution in non-Japanese patients; 4 non-Japanese patients with a BSA equal to the 75th percentile of the BSA distribution in non- Japanese patients; 5 non-Japanese patients with a BSA equal to the

25th percentile of the BSA distribution in non-Japanese patients; 6 non-Japanese patients with a BSA equal to the 5th percentile of the BSA distribution in non-Japanese patients; 7 non-Japanese patients with a median BSA equal to 1.77 m2. The exposure is assessed by area under the plasma concentration–time curve in the first cycle (AUC0–28) and maximum plasma concentration (CMAX) over the dosing interval. Shaded areas represent the 80–125% interval of the reference covariate group

Fig. 6 Typical concentration vs. time after dose (TAD) profiles for volasertib (a) and cytarabine (b) for individuals belonging to different covariate groups. Simulations were performed at a single dose of 350 mg of volasertib infused during 1 h, and a single 20-mg dose of cytarabine administered subcutaneously. The shaded area shows the

95% prediction interval for the average patient taking inter-individual variability into account and obtained from 500 simulated profiles at the doses described above. The black horizontal dotted line in b shows the limit of quantification for cytarabine observations. LN logarithmic transformation, P percentile

Third, covariate effects were tested in the different Asian subpopulations, but owing to the low number of patients in these ethnic subgroups, only Japanese and Korean patients could be thoroughly investigated. The pharmacokinetic parameters CL, V1, V3, Q2 and Q3 were found to be decreased in Japanese patients compared with non-Japanese patients, yielding higher maximum plasma concentration and AUC0–28 for these patients. The median BSA in these patients is also lower than in non-Japanese patients, decreasing as well their parameter values. Hence, ethnicity and BSA were identified as statistically and potentially clinically relevant covariates on the parameters where they were introduced. When comparing volasertib exposure in Japanese patients with non-Ja- panese patients (Fig. 5), it was found that indicators reflecting drug exposure were outside the 80–125% range of the typical values of a non-Japanese patient with a median BSA. In previous studies [15, 16], these differ- ences in exposure have been assumed to be owing to weight differences, but in this case, those differences are

taken into account with the inclusion of BSA as a covariate in the model. Nevertheless, the number of Japanese patients is low compared with the number of non-Asian patients (35 vs. 423), and the effect expected on efficacy and safety is unknown. This information should be interpreted with caution and needs further investigation in future studies when additional data in this subpopulation are available.
Similarly, the Korean ethnic group comprised only sparse sampling as it was observed exclusively in the phase III study. It was found to have an effect on volume terms estimated with a very high imprecision as well as a very high effect on inter-compartmental clearance terms, suggesting the disappearance of one compartment, prob- ably owing to the lack of information provided on dis- tribution from the sparse sampling. The Korean ethnic group was therefore not implemented as a covariate in the final model because of the lack of information about this population. Similarly to the finding in the Japanese patients’ subgroup, the effect of ethnicity on the

pharmacokinetics of volasertib remains to be confirmed when more data in Asian patients are available. The full covariate model is presented even though clinical rele- vance was not found. The reason for doing this is because the full model can be used in further studies with patients with different characteristics to see if clinical relevance can then be identified.
With respect to cytarabine, there have been previous reports on the population pharmacokinetics of cytarabine in an analysis of data from patients with AML (n = 23) receiving high-dose cytarabine following an intravenous bolus [26]. In the cited work, cytarabine was adminis- tered to patients receiving also etoposide and daunoru- bicin. The reported estimates of CL and V were 272 L/h and 138.2 L, respectively. In our study, the corre- sponding values were 208 L/h and 209 L, respectively. Those values could be considered within the same range given the differences in the administered dose, dosing schedule and the number of patients studied, and suggest that the absolute bioavailability of cytarabine is high. It has to be taken into account that in our analysis because of limited sampling, cytarabine disposition in plasma was described by a one-compartment model in contrast to the two-compartment model selected by Krogh-Mad- sen et al. [26]. The estimates of IIV for CL and V between the previous and the current work are compa- rable (45 vs. 33% in CL and 70 vs. 53% in V). As in the case of volasertib, no time dependencies were present in the data.
In the previous model developed for cytarabine [26], sex was found to be a significant covariate for CL, even though it was not included in the final model. These differences previously related to sex may in part be explained by weight differences, as it is related to BSA; these sex dif- ferences can be explained by the model currently devel- oped, which includes BSA as a covariate on CL/F and V/
F. This difference in covariates between the previous and the current model may actually describe the same effect to a certain extent, as weight is correlated to both sex and BSA.

5 Conclusion

To summarise, the pharmacokinetics of volasertib was studied in a group of 501 patients with AML receiving 150–550 mg of volasertib as a 1- or 2-h intravenous infusion, and was found to be time independent. Body surface area and ethnicity showed significant covariate effects reflected mainly as an increase in drug exposure for Japanese patients, although this finding has to be interpreted with caution because only 7% of the patients were part of that population subgroup. Volasertib showed

moderate IIV in total clearance. In the current analysis, the pharmacokinetics of cytarabine was also studied after multiple subcutaneous administrations, resulting in pre- dictable plasma concentration profiles with BSA affecting drug disposition. Finally, volasertib and cytarabine did not influence the pharmacokinetic characteristics of each other.

Compliance with Ethical Standards

Funding This work has been funded by Boehringer Ingelheim GmbH & Co.KG.

Conflict of interest Bele´n P. Solans and In˜aki F. Troco´niz have received research funding from Boehringer Ingelheim GmbH & Co.KG. Ange`le Fleury, Matthias Freiwald, Holger Fritsch an Karin Haug are employed by Boehringer Ingelheim GmbH & Co.KG.

References

1. Saultz J, Garzon R. Acute myeloid leukemia: a concise review. J Clin Med. 2016;5(3):E33. doi:10.3390/jcm5030033.
2. Hao Z, Kota V. Volasertib for AML: clinical use and patient consideration. Onco Targets Ther. 2015;8:1761–71.
3. Howlader N, Noone AM, Krapcho M, Garshell J, Miller D, Altekruse SF, Kosary CL, Yu M, Ruhl J, Tatalovich Z, Mariotto A, Lewis DR, Chen HS, Feuer EJ, Cronin KA, editors. SEER cancer statistics review, 1975–2012. National Cancer Institute, Bethesda, MD, USA. http://seer.cancer.gov/csr/1975_2012/. Accessed 2 June 2017.
4. Erba HP. Finding the optimal combination therapy for the treat- ment of newly diagnosed AML in older patients unfit for inten- sive therapy. Leuk Res. 2015;39(2):183–91.
5. Shafer D, Grant S. Update on rational targeted therapy in AML. Blood Rev. 2016;30(4):275–83.
6. Do¨ner H, Weisdorf D, Bloomfield C. Acute myeloid leukemia. N Engl J Med. 2015;373:1136–52.
7. Holtrich U, Wolf G, Bra¨uninger A, Karn T, Bo¨hme B, Ru¨bsa- men-Waigmann H, et al. Induction and down-regulation of PLK, a human serine/threonine kinase expressed in proliferating cells and tumors. Proc Natl Acad Sci USA. 1994;91(5):1736–40.
8. Barr FA, Sillje´ HHW, Nigg EA. Polo-like kinases and the orches- tration of cell division. Nat Rev Mol Cell Biol. 2004;5(6):429–40.
9. Ikezoe T, Yang J, Nishioka C, Yokoyama A. A novel treatment strategy targeting polo-like kinase 1 in hematological malignan- cies. Leukemia. 2009;23(1):1564–76.
10. Brandwein JM. Targeting polo-like kinase 1 in acute myeloid leukemia. Ther Adv Hematol. 2015;6(2):80–7.
11. Gjertsen BT, Scho¨ffski P. Discovery and development of the Polo-like kinase inhibitor volasertib in cancer therapy. Leukemia. 2015;29(1):11–9.
12. Rudolph D, Steegmaier M, Hoffmann M, Grauert M, Baum A, Quant J, et al. BI 6727, a polo-like kinase inhibitor with improved pharmacokinetic profile and broad antitumor activity. Clin Can- cer Res. 2009;15(9):3094–102.
13. Gorlick R, Kolb EA, Teir S, Maris J, Reynolds P, Kang M, et al. Initial testing (stage 1) of the Polo-Like kinase inhibitor vola- sertib (BI6727), by the Pediatric Preclinical Testing Program. Pediatr Blood Cancer. 2014;61(1):158–64.
14. Rudolph D, Impagnatiello MA, Blaukopf C, Sommer C, Gerlich DW, Roth M, et al. Efficacy and mechanism of action of vola- sertib, a potent and selective inhibitor of Polo-like kinases, in

preclinical models of acute myeloid leukemia. J Pharmacol Exp Ther. 2015;352(3):579–89.
15. Nokihara H, Yamada Y, Fujiwara Y, Yamamoto N, Wakui H, Nakamichi S, et al. Phase I trial of volasertib, a Polo-like kinase inhibitor, in Japanese patients with advanced solid tumors. Invest New Drugs. 2016;34(1):66–74.
16. Do¨hner H, Lu¨bbert M, Fiedler W, Fouillard L, Haaland A, Brandwein JM, et al. Randomized, phase 2 trial comparing low- dose cytarabine with or without volasertib in AML patients not suitable for intensive induction therapy. Blood. 2014;124(9):1426–34.
17. Beal S, Sheiner LB, Boeckmann A, Bauer RJ. NONMEM user’s guides (1989–2015). Ellicott City: Icon Development Solutions; 2015.
18. Beal SL. Ways to fit a PK model with some data below the quantification limit. J Pharmacokinet Pharmacodyn. 2001;28(5): 481–504.
19. Ahn J, Karlsson M, Dunne A, Ludden TM. Likelihood based approaches to handling data below the quantification limit using NONMEM VI. 2008;35(4):401–21.
20. Metzeler KH, Herold T, Rothenberg-Thurley M, Amler S, Sauerland MC, Gerlich D, et al. Spectrum and prognostic rele- vance of driver gene mutations in acute myeloid leukemia. Blood. 2016;128(5):686–98.

21. Jonsson E, Karlsson MO. Automated covariate model building within NONMEM. Pharm Res. 1998;15:1463–8.
22. Lindbom L, Pihlgren P, Jonsson EN. PsN toolkit: a collection of computer intensive statistical methods for nonlinear mixed effect modelling using NONMEM. Comput Methods Programs Biomed. 2005;79:241–57.
23. Bergstrand M, Hooker AC, Wallin JE, Karlsson MO. Prediction- corrected visual predictive checks for diagnosing nonlinear mixed-effects models. 2011;13(2):143–51.
24. Soto E, Staab A, Tillmann C, Trommeshauser D, Fritsch H, Munzert G, et al. Semi-mechanistic population pharmacokinetic/ pharmacodynamic model for neutropenia following therapy with the plk-1 inhibitor BI 2536 and its application in clinical devel- opment. Curr Microbiol. 2010;61(4):785–95.
25. Robert J, Rigal-Huguet F, Hurteloup P. Comparative pharma- cokinetic study of idarubicin and daunorubicin in leukemia patients. Hematol Oncol. 1992;10(2):111–6.
26. Krogh-Madsen M, Bender B, Jensen MK, Nielsen OJ, Friberg LE, Honore´ PH. Population pharmacokinetics of cytarabine, etoposide, and daunorubicin in the treatment for acute myeloid leukemia. Cancer Chemother Pharmacol. 2012;69(5):1155–63.