1

Research Article

Clinical pharmacology profile of an oral selective androgen

receptor down-regulator, AZD3514: Implications on the design

of ongoing castrate-resistant prostate cancer clinical studies

Angela W. Dymond1, Marc-Antoine Fabre1, Gareth D. James2, Masako Hirata3, Simon A. Smith4, Paul

A. Dickinson1,5, Michael Dymond6, Glen Clack7

1Quantitative Clinical Pharmacology, Early Clinical Development, AstraZeneca, Alderley Park,

Macclesfield, UK

2 PHASTAR, Unit 2, 2a Bollo Lane, London W4 5LE, UK

3Clinical Science Division, Research and Development, AstraZeneca K.K., Osaka, Japan

4Oncology Translational Medicine Unit, AstraZeneca, Melbourn Science Park, Melbourn, Royston,

UK

5Current Address, Seda Pharmaceutical Development Services®, The BioHub at Alderley Park,

Alderley Edge, UK

6Discovery Sciences Statistics, AstraZeneca, Alderley Park, Cheshire, Macclesfield, UK

7Oncology Translational Medicine Unit, AstraZeneca, Alderley Park, Macclesfield, UK

Corresponding author:

Simon A. Smith, Oncology Translational Medicine Unit, AstraZeneca, Melbourn Science Park,

Melbourn, Royston, UK

E-mail: Simon.A.Smith@astrazeneca.com

Telephone: +44 7557 540 988

2

Abstract

Purpose: To describe the pharmacokinetics (PK) and pharmacodynamics (PD) of AZD3514 and how

the design of the first-time-in-human study was adapted based on the emerging clinical PK.

Patients and methods: Data were collected from 77 patients with castrate-resistant prostate cancer

from two dose-escalation studies, in Europe (NCT01162395) and Japan (NCT01351688). PK

parameters were derived from plasma and urine data and exploration of PK-PD relationships were

performed. Post hoc analysis was conducted to investigate time-dependent changes and inter- and

intra-patient variability in PK.

Results: AZD3514 was rapidly absorbed and plasma levels declined in a bi-phasic manner with no

ethnic differences. Plasma exposure to AZD3514 was dose proportional. Generally, overall exposures

were similar between visits within each patient, but varied between patients within each cohort. A

switch to twice-daily dosing, to increase exposure, produced a marked time-dependent reduction in

area under the curve of 30% and an increase in apparent clearance (from 17 to 25 L/h) at steady state

compared to single doses. Emerging study data showed that low baseline testosterone may influence

prostate-specific antigen (PSA) reductions by AZD3514. Combination cohorts with abiraterone

acetate, a drug that decreases testosterone in CRPC patients, did not result in meaningful decreases in

PSA.

Discussion and conclusions: Despite adaptation of the clinical strategy from emerging PK and PD

data, the hypothesis around androgen receptor (AR) modulation through AR down-regulation could

not be tested due to the time-dependent effect on AZD3514 PK, which prevented coverage above the

target concentration. Further testing of this hypothesis is warranted.

Key words: clinical trial, phase 1, pharmacokinetics, pharmacodynamics, AZD3514, castrate-

resistant prostate cancer, adaptive design

3

Introduction

Prostate cancer is a common disease and is most often effectively treated with surgery or radiotherapy.

However, in patients with aggressive forms or metastatic disease, treatment is more difficult. In such

patients, the standard of care is androgen deprivation therapy achieved by either surgical castration or

administration of luteinizing hormone-releasing hormone agonists (1, 2). However, the positive effects

of androgen deprivation therapy are only temporary and in almost all patients the cancer inevitably

returns. Evidence has accumulated that tumor growth in castrate-resistant prostate cancer (CRPC) is

still dependent on the androgen receptor (AR) and the AR has become a target for drug development

with abiraterone and enzalutamide recently reaching the market (3-6). Unfortunately, despite castrate

levels of testosterone in CRPC patients, resistance ultimately develops, which in many cases is still

dependent on AR (7-10), and new drugs that are less prone to, or can over-come, the development of

resistance are needed.

AZD3514 is a new, first-in class compound that interferes with AR signaling, by binding to

AR, inhibiting its nuclear translocation, and ligand-dependent and independent transcriptional activity

(11). In the Hershberger rat model (12), AZD3514 caused a dose-dependent decrease in seminal vesicle

weight and a reduction of AR protein expression in ventral prostates of castrated adult rats dosed with

testosterone propionate (11). AZD3514 also inhibited the growth of androgen-dependent Dunning

R3327H prostate tumors in adult rats (AstraZeneca, data on file). AZD3514 differs from other drugs

directed at the AR receptor such as enzalutamide and bicalutamide in that it also induces AR down

regulation (11).

AZD3514 was developed as an immediate release formulation as a maleate salt (AZD3514

free base has a pKa of 6.2). The solubility of AZD3514 in pH 6.5 phosphate buffer (25°C) is 18

mg/mL, and 24 mg/mL in pH 1.2 simulated gastric fluid. The Biopharmaceutics Classification

System (BCS) guidance for AZD3514 is a tentative BCS4 at the doses being investigated in the

clinic.

Formatted: Left

4

The pre-clinical PK profile showed that AZD3514 was well absorbed in the dog with maximum

concentrations achieved at approximately 5 h with a terminal half-life of 5.6 h. Predictions from pre-

clinical species suggested that the target exposure of 2410 ng/ml for 18 h in a 24 h period was expected

to be achieved in humans at the 500 mg once-daily (QD) dose. In vitro experiments indicated that

AZD3514 is metabolized by CYP3A4 but the metabolic turnover was low. Pre-clinical safety

assessment identified gastrointestinal effects which were considered dose-limiting in both rats and

dogs (AstraZeneca, data on file) but the overall pre-clinical profile warranted testing in man.

Prostate-specific antigen (PSA) is a biomarker for prostate cancer used in diagnosis, for

monitoring diseases progression and assessing therapeutic response. Despite some limitations, notably

a lack of specificity for prostate cancer, PSA remains the primary clinical pharmacodynamic biomarker

available for prostate cancer used in the clinic (13). In this manuscript, pharmacokinetic and

pharmacodynamic (PSA) results from two first-time-in-man studies with AZD3514 on patients with

CRPC are reported. We describe how the unusual PK characteristics of AZD3514 and emerging PSA

biomarker data led to changes in the design of one of the clinical studies with inclusion of cohorts

dosed with AZD3514 in combination with abiraterone acetate. We present the challenges that can

occur in first-time-in-man studies that are often not predicted from pre-clinical species and how they

may be overcome to test novel mechanism of actions in CRPC. Post hoc analysis to investigate the

inter- and intra-patient variability in PK in patients who received AZD3514, and explore if dosage

influenced the variability is also presented. Due to the wealth of data gathered, the safety and efficacy

data have been reported elsewhere (14).

Formatted: Left, Indent: First line: 0", Line spacing: single

5

Methods

Patient population

Male patients older than 20 years of age with CRPC and a life expectancy of at least 12 weeks could

be enrolled. The patients had to have a documented evidence of metastatic prostate cancer for which

no standard therapy was considered appropriate. Details of the studies inclusion/and exclusion criteria

have been reported previously (14). All patients were required to provide signed and dated written

informed consent prior to any study specific procedures.

Study design

There were two phase I, open-label, multicenter, dose escalation studies. Study 1 was conducted in 5

centers in the UK, US and The Netherlands, and study 2 was conducted in 2 centers in Japan. All

patients in Study 1 were white and of European ancestry. Both studies followed similar methodology

but fewer doses were tested in Japan. Both studies were performed in accordance with the Declaration

of Helsinki and the International Conference on Harmonization Good Clinical Practice Guidelines and

approved by relevant regulatory and independent ethics committees.

At least 3 and up to 6 evaluable patients were required for each dose cohort. However, patient

cohorts at selected doses could be expanded to a maximum of 12 patients to investigate further the

tolerability, pharmacokinetics and biological activity of AZD3514. Each patient received a single dose

of AZD3514 on Day 1 and after a washout of 7 days multiple dosing was started on Day 8 onwards

until discontinuation. In study 1, the following single and multiple doses of AZD3514 were

investigated: 100, 250, 500 and 1000 mg QD and 1000 and 2000 mg twice daily (BID). Study 2 was

opened after the first 2 dose cohorts had been recruited in Study 1. Patients in Study 2 received single

doses of 250, 500 or 1000 mg and multiple doses of 250 or 500 mg QD and 500 mg BID. These doses

were selected based on findings from Study 1. The study drug was taken in the fasted state (food

restriction for at least 2 hours before and 1 hour after administration of AZD3514). In study 1, based

6

on emerging pharmacokinetic/pharmacodynamic modeling results, additional cohorts were included

investigating the effects on PSA from 500 mg BID AZD3514 in combination with 1000 mg QD

abiraterone acetate.

On study days 1, 8 and 29, venous blood samples for measuring AZD3514 plasma

concentrations were taken at the following time points for the once-daily dosing regimens: pre-dose,

0.5, 1, 2, 3, 4, 5, 6, 8, 10, 24 and on Day 1 only: 48, 72 and 96 hours post-dose. Similar pharmacokinetic

sampling was performed for the twice daily dosing regimens with the exception that on Days 8 and

29, the 10 h samples were replaced with a collection at 12 h post-dose. Urine collection for

pharmacokinetic purposes started immediately pre-dose until 10 h post-dose on Days 1 and 29. PSA

levels were measured at screening, on Days 8, 15, 29, 57 and 85 and once every 4 weeks thereafter

until discontinuation.

Pharmacokinetic analysis including post hoc assessment of temporal change in apparent clearance

and inter and intra patient variability in PK

Plasma samples were prepared by solid phase extraction and liquid chromatography. The

concentrations of AZD3514 in plasma and urine were quantified by tandem mass spectroscopy.

Pharmacokinetic parameters were determined by non-compartmental analysis using Phoenix™-

WinNonlin® v6.3. PK parameters were derived by the following methods: maximum observed

concentrations (Cmax, Cssmax) and time to maximum concentrations (tmax, tssmax) were determined by

inspection of the concentration-time profiles; λz was calculated by log-linear regression of the terminal

portion of the plasma concentration-time profiles; terminal half life (t½λz) was calculated as Ln(2)/λz;

the area under the plasma concentration-time curve up to the last quantifiable sample (AUC(0-t)), the

area under the plasma concentration-time curve up to the end of the dosing interval (AUCtau) and the

area under the curve at steady state (AUCss) were calculated using the linear up/log down trapezoidal

rule; where appropriate, AUC(0-t) was extrapolated to infinity using λz to obtain AUC0-; apparent

7

clearances (CL/F following the single dose and CLss/F following multiple dosing) were determined

from the ratio of dose/AUC0- or dose/AUCss. Apparent volume of distribution (V/F) was determined

from the mean residence time x CL/F; accumulation ratio was determined from the ratio of AUC(0-t)

on Day 29/AUC(0-t) on Day 1; temporal change parameter was calculated from the ratio of AUCtau Day

29/AUC Day 1.

A power model (15) was used to explore the dose proportionality of AZD3514 pharmacokinetics.

Possible effects of race on the pharmacokinetics of AZD3514 were analyzed descriptively and

explored graphically. AUC0-24h was chosen for this analysis in order to include patients for whom

AUC0- could not be calculated. Comparison of individual Day 1 against Day 29 apparent clearance

in Western patients was performed post hoc using a 2-sided paired t-test with a 5% significance level,

no statistical analysis was performed with the Japanese dose groups due to the low number of patients

(n = 4-5 per cohort). The inter and intra patient variability between Days 1 and 8 was assessed in terms

of Ln(AUC) and was conducted using SAS version 8.2. All patients who received doses of 100 mg to

1000 mg AZD3514 and had AUC measurements on Days 1 and 8 with no exclusions due to emesis

were used in the variability analysis. The correlation between the Ln(AUC) at Days 1 and 8 was

calculated for each dose, and a Bland Altman test (16) was conducted to determine any Ln(AUC) time

effects.

Population pharmacokinetic modeling

A retrospective population PK analysis was performed at the end of the studies to evaluate any

potential PK dependent covariates and to support the NCA findings.

This data was not used for decision making when the studies were ongoing. The population PK analysis

was based on multiple regressions using NONMEM program (version 7.2). Since the data set consisted

exclusively of rich data the first-order conditional estimation (FOCE) method performed well for the

stability, robustness and predictability of the model.

8

The model included four basic components as follows: (i) the structural PK model, which

predicts plasma concentration as a function of time and dose; (ii) the covariate model component,

which describes the influence of fixed effects (e.g. demography, laboratory data) on PK parameters;

(iii) the between-subject variance component, which describes the inter-individual variation in PK

parameters (after ‘correction’ for fixed effects); and (iv) the residual error model component, which

describes the underlying distribution of the error in the measured plasma concentrations.

The selection of the structural PK model and variance models for residual error was based on

the goodness-of-fit plots and on the difference in NONMEM objective function (-2LL: - 2xLog

Likelihood) between hierarchical models (i.e. the “likelihood ratio” test). The covariate models used

in this analysis represent shifts from the “typical” subject parameter value. Potential covariates were

selected by univariate analysis, testing the effect of each covariate on each of the relevant PK

parameters. A p value of 0.05 was chosen to retain one parameter, i.e. a difference in the objective

function ≥ 3.84 for one degree of freedom. The covariates identified by the univariate selection were

then included into a “full” model after ranking by the size of change in the objective function; rank 1

having the largest influence on the objective function. The covariates evaluated in this study were race,

age, bodyweight, body mass index, body surface area, creatinine clearance, alanine and aspartate

aminotransferases and bilirubin.

Population PK models were acceptable if they resulted in successful minimization and a

successful estimation of the covariance. Other requirements were that 95% confidence intervals of

structural parameters should not include zero, and the absolute value of correlation between two

structural parameters should not be >0.95. Diagnostic plots (population prediction versus observed

concentrations, individual prediction versus observed concentrations, weighted residual versus

population prediction and weighted residual versus time after dosing) were used to evaluate the

goodness-of-fit throughout the model-building procedure.

9

RESULTS

The patient demographics, clinical characteristics and details of prior anti-tumour therapies have been

reported previously (14). All patients had metastatic disease with 84% having metastasis to bone, 49%

metastasis to lymph nodes and 12% metastasis to viscera.

Pharmacokinetics of QD dosing

AZD3514 concentration-time profiles after single-dose administration in study 1 are shown in

Figure 1. The pharmacokinetic parameters from the non-compartmental analysis from both studies

following a single dosing and at steady state are summarized in Tables 1 and 2 respectively. Plasma

concentrations of AZD3514 peaked at 2 to 2.9 hours with no apparent lag in absorption, after which,

concentrations declined in a bi-phasic manner at all doses in both studies and the majority of AZD3514

(>90% of AUC0-) was cleared by 24 hours post-dose at all dose levels. The elimination phase was

characterized by a t1/2 of approximately 16 h in study 1, which appeared independent of dose (Table

1). At steady state on Day 29 and following QD dosing, values for Cssmax and tssmax were generally

similar to those observed after single-dose administration and there was no evidence of accumulation

(Tables 1 and 2). Increases in the mean apparent clearance at steady state (CLss/F) of ~7% to 31% at

100 to 1000 mg QD compared to single dose administration was detected, resulting in a small time-

dependent effect on AUC with mean temporal parameter change values ranging from 0.85 to 0.91.

Plasma exposure to AZD3514 increased proportionally with dose after both single- and

multiple-dose administration in both Western and Japanese patients (Table 3). The inter-patient

variability, based on group % CV, was low to moderate ranging from 1.3% to 44% for AUC0-∞ after

single AZD3514 doses (Table 1) and 8.9% to 35% for AUCtau on Day 29 (Table 2).

The post hoc variability assessment included 35 patients, 7 patients were excluded due to

emesis and two patients were excluded as their PK measurement was missing at either Day 1 or day

8. The results showed that for intra-patient variability, ln(AUC) between Day 1 and 8 in the AZD3514

monotherapy cohorts was highly correlated in the 100 mg, 250 mg and 500 mg QD cohorts (90, 75

10

and 95% respectively), and moderately correlated in the 1000 mg QD cohort (41%). A Bland Altman

test indicated there was no evidence of a time effect on ln(AUC) (95% confidence interval: -0.055,

0.078). Hence, ln(AUC) values were similar on Days 1 and 8 for the majority of patients indicating

low intra-patient variability. The ln(AUC) varied more between patients at the same dose than within

patients between Day 1 and 8, but overall the observed inter-patient variability was modest for a

compound metabolized exclusively by CYP3A4 and with low renal clearance. Post hoc analysis

results are shown in Supplementary Table 1 and Supplementary Figure 1. In all the above analyses,

one patient from the 100 mg cohort in study 1 was excluded from the descriptive statistics as he was

later found to have taken the prohibited co-medication diltiazem, a moderate CYP3A4 inhibitor (17),

which has the potential for PK interaction with AZD3514. At the 1000 mg QD cohort in study 1, one

patient’s Day 29 results were excluded due to a late dose administered on Day 28 which resulted in

increased AZD3514 plasma concentrations on Day 29.

Change to twice-daily dosing to increase exposure

The exposure of AZD3514 at 1000 mg QD dosing did not reach the target coverage of 2410 ng/mL

for 18 hours in a 24 h period, and simulations suggested that escalation to 2000 mg QD would not

reach the desired threshold either. The inability of QD dosing to reach the target coverage lead to a

change to twice-daily (BID) dosing beginning at the 1000 mg dose.

A marked temporal change in the pharmacokinetics of AZD3514 was apparent at 1000 mg BID

following multiple dosing, with a 30% lower in overall exposure on Day 29 compared Day 1

(individual temporal parameter change values ranged from 0.57 to 0.76) (Table 2). Post hoc statistical

comparison of individual changes in apparent clearance from Day 1 to Day 29 (Figure 2) indicate no

marked differences at 100 to 500 mg QD of AZD3514, although most patients showed some increase

with multiple dosing. Statistically significant increase in CLss/F on Day 29 at 1000 mg QD and 1000

mg BID compared to CL/F on Day 1 was detected, however, the small group sizes and the magnitude

11

of change with respect to clinical relevance should be taken into consideration. All patients dosed at

1000 mg BID showed increased CLss/F with a clinically relevant increase of ~50% in the group mean

(17.1 L/h on Day 1 to 25.0 L/h on Day 29) which would warrant a dose alteration. Consequently, a

further dose escalation at 2000 mg BID was tested.

Doses of 100 to 1000 mg QD and 1000 mg BID were well tolerated, however, on escalation to

2000 mg BID the occurrence of non-tolerated nausea and vomiting resulted in the discontinuation of

dosing prior to Day 29, hence no steady state data were obtained. The switch to BID dosing did not

permit attainment of the desired target coverage, and the clinical strategy changed to consider

alternative treatment strategies.

Renal excretion

The mean fraction of AZD3514 excreted in urine (fe) ranged from 2.6 to 6.6% of dose resulting in low

renal clearance (CLR), ranging from 0.45 L/hour to 1.17 L/hour. Similar renal excretion was detected

following single and multiple dosing and was independent of dose and schedule. At the 1000 mg BID

dose when marked temporal change in PK was detected, the fe and CLR remained low and were similar

following single and multiple dosing (fe at 3.8% and 2.2%; CLR at 0.88 and 0.54 L/hour, on Day 1 and

Day 29, respectively), so although overall clearance increased over time, renal clearance was not

affected.

Impact of ethnicity

Comparison of the systemic exposure of AZD3514 between Japanese and Western patients at

equivalent dose levels showed only minor ethnic differences with Cmax and AUC0- appearing

marginally higher in Japanese patients at the 500 and 1000 mg QD doses (Table 1). The mean

bodyweight of Japanese patients was approximately 17% lower than that of the Western patients, so

12

when AUC0-24h was normalized for body weight and dose, no difference in exposure was apparent

(Figure 3).

Population pharmacokinetic modeling

A two-compartment linear model with zero order absorption was found to adequately describe the

AZD3514 plasma concentration-time profile. The time dependency on pharmacokinetic parameters

was assessed as prior information during the development of the structure model (base model). The

models used to assess the time effect on PK parameters were a categorical and saturation model (Emax

or Hill model). The results of the model evaluation showed statistically significant changes in the

objective function, and parameters were well estimated for the simple categorical model on CL/F and

V2/F. The time dependency was also tested on bioavailability (F), in the same manner as described

aforesaid; however the statistical change was minor and thus not selected in the structure model.

During the covariate analysis, there were no correlations between apparent clearance, apparent initial

volume or apparent peripheral volume and the race, age, weight, body mass index or creatinine

clearance of the patients. Taking into account these covariates, the PK exposure from the population-

PK model was found to be the same between Western and Japanese patients confirming the conclusion

derived from the non-compartmental analysis. The parameters of the final model are provided in Table

4. The diagnostic goodness-of-fit visual predictive check (VPC) plots (Supplementary Figure 2), show

no bias in the population or individual predictions. The plots of conditional weighted residuals vs.

predictions and time show that most residuals were small and evenly distributed between –2 and 2

(data not shown), and therefore demonstrate the strong predictivity of the model.

Effects of AZD3514 on PSA

The effect of different doses of AZD3514 on the level of PSA during the first two months of

treatment in Western CRPC patients is shown in Figure 4. The 57-day time point was chosen as the

13

cut off for this graphical analysis because most patients were treated for at least 57 days whereas

treatment discontinuations were frequent thereafter. The mean baseline PSA value varied from 137 to

478 ng/mL between dose groups (Figure 4), whereas individual baseline PSA values varied from 6.5

to 5408 ng/mL (data not shown). No apparent effects of AZD3514 on the levels of PSA were noted

based on graphical examination (Figure 4). However, in some patients (12 out of 70 in the monotherapy

groups), treatment with AZD3514 resulted in a clinically interesting (>30%) transient decrease in

circulating PSA within 12 weeks of treatment (Figure 5), suggesting that modulation of AR signaling

was occurring albeit sub optimally. Similar observations were found in Japanese patients (study 2, data

not shown). To explain these observations, a systems pharmacology (mechanistic) modeling approach

was undertaken to determine if AZD3514 may be effective in a specific sub-population of patients,

with the expectation that it might help inform the ongoing clinical study design (18).

The systems pharmacology model indicated AZD3514 may be more effective in a low

dihydrotestosterone (DHT) environment and so administration of AZD3514 with abiraterone acetate

(which blocks the production of DHT) may result in greater efficacy even at lower exposures of

AZD3514 (18). As target exposure for efficacy was not achieved by AZD3514 alone, this alternative

dosing strategy may potentially provide a viable treatment option. Thus, abiraterone combination

cohorts dosed at 500 mg BID AZD3514 (a proposed long-term tolerated dose) were opened, initially

in abiraterone acetate naïve patients and later in patients on abiraterone acetate at the time of

progression. Adjustment of the patient population was made owing to this population being identified

as an area of unmet medical need and that evidence of resistance to abiraterone acetate provided an

opportunity to demonstrate the benefit of adding AZD3514. One of the 5 patients that had progressed

on abiraterone and was treated concomitantly with 500 mg BID AZD3514, had a much higher baseline

PSA value than the other 4 patients and this patient discontinued treatment within 2 weeks of coming

onto study. In the remaining 4 patients, AZD3514 did not lower PSA levels.

14

DISCUSSION

Challenges in the clinical drug development of AZD3514

The unexpected clinical pharmacokinetics of AZD3514 and significant issues with long-term

tolerability, namely nausea and vomiting (14) at high doses, presented challenges to the development

of AZD3514 resulting in adaptive changes to the design of the clinical studies (Figure 6).

In CRPC patients, the pharmacokinetic profile of AZD3514 is characterized by 1) rapid

absorption with tmax values of about 2 h, 2) biphasic disposition with a terminal t1/2 of about 16 h, 3)

dose proportionality in the once daily dose range investigated, 4) no accumulation after QD and BID

dosing, 5) low renal excretion of parent compound and 6) marked time-dependent pharmacokinetics

apparent at higher doses and shorter dosing interval. The apparent discrepancy between a t1/2 of about

16 h and the lack of accumulation may be explained by a first elimination phase accounting for the

majority of the elimination of AZD3514. The combination of a lack of accumulation and time-

dependent pharmacokinetics resulted in low trough concentrations that were insufficient to achieve the

target drug levels over a specified time predicted to be required for efficacy. The underlying

mechanism for the marked reduction in AZD3514 exposure after multiple-dose administration

compared to single doses is unknown. One possibility is that AZD3514 activates mechanisms that

facilitate its own metabolism by induction of CYP3A4 since AZD3514 is almost exclusively

metabolized by this route.

However, in vitro studies indicate that AZD3514 does not activate the pregnane X receptor, a

nuclear receptor involved in the regulation of a wide variety of genes involved in the elimination of

xenobiotics and activated by known inducers of CYP3A4 (19, 20). Furthermore, in cultured human

hepatocytes, AZD3514 did not induce CYP1A1/2, CYP2B6 or CYP3A4 at concentrations up to 100

µM (AstraZeneca, data on file), which is well above clinical exposures. These data suggest that auto

induction of metabolizing enzymes by AZD3514 is unlikely but this cannot be ruled out. To determine

whether hepatic CYP3A4 enzyme induction may be occurring in the clinic, measurement of the

Deleted: elimination

15

biomarker, 4β-hydroxycholesterol, was planned for new patients recruited onto study 1. However, the

study was stopped before samples could be collected and thus there is no clinical data to support the

potential for auto induction of metabolizing enzymes by AZD3514. As a point of note, the

measurement of 4β-hydroxycholesterol requires the collection of only a few blood samples and incurs

limited cost, and thus the early inclusion of 4β-hydroxycholesterol measurements to evaluate clinical

enzyme induction is worth consideration in first-time-in-man studies. Alternative in vivo mechanisms

of actions such as changes in absorption and hence bioavailability with time could not be ruled out as

these were not investigated.

Atypical low AZD3514 variability

An exploratory investigation showed that multiple oxidative metabolites were present in the blood

samples with parent constituting about 40% of overall drug related material at 3 h post dose

(AstraZeneca, data on file). Clinical evidence that CYP3A4 is substantially involved in the in vivo

metabolism of AZD3514 comes from the pharmacokinetic results of the one patient who was co-

administered with a moderate CYP3A4 inhibitor, diltiazem, in error. In this patient, the overall

exposure (AUC) to AZD3514 was about two-fold higher than the other patients with minimal effect

on Cmax. This suggests a small first-pass extraction of AZD3514 which is supported by the moderate

CL/F of 13 to 17 L/h in western patients which is ~ 17% of liver blood flow. Although Ln(AUC)

measurements varied between patients on the same doses, values on Day 1 and 8 for each patient were

generally similar. However, considering the apparent substantial contribution of CYP3A4 to the

clearance of AZD3514 the variability was rather low with a CV generally below 30%. Based on the

expectation that CYP3A4 is the main enzyme involved in the clearance of AZD3514, the PK profiles

and particularly the variability observed are somewhat surprising. The PK of compounds that are

thought to be exclusive substrates of CYP3A4 have been investigated extensively in an attempt to

identify a probe substrate that can be used to profile the CYP3A4 capacity of an individual to allow

16

dose individualization for all CYP3A4 substrates. Benet (21) and Masica et al. (22) investigated the

three benzodiazepines, alprazolam, triazolam and midazolam all of which are metabolized exclusively

by CYP3A4 and not subject to p-glycoprotein efflux. They reported mean oral clearance of 4.5, 28 and

92 L/h for alprazolam, triazolam and midazolam with a % standard deviation of 48, 42 and 50%

respectively. The mean oral clearance of AZD3514 in the dose proportional pharmacokinetic dose

range of 100 to 1000 mg was ~ 15 L/h which falls within the range of clearance observed by Masica

and co-workers. Other workers have reported high % standard deviations when measuring the

clearance of CYP3A4 substrates such as 73% and 63 % for alfenatil and midazolam respectively (24),

and 47% for midazolam (24).

Kato and co-workers (25) developed a method for predicting the inter-individual variability of

human exposure for CYP3A4 substrates using Monte Carlo simulation. Using this model they

concluded that inter-individual variability was related to clearance with higher clearance drugs (fh >

0.38) resulting in higher %CV. The majority of the variability was attributable to CYP3A4 expression

rather than variability in other physiological factors such as hepatic blood flow. To achieve a good fit

of the model data to the CYP3A4 substrates selected to assess the model against, Kato and co-workers

had to select the lowest reported %CV for CYP3A4 expression in liver microsomes (33%). This is

compared to other reported values that were in the range of 50 to 100%. It is suggested this is because

the % CV reported for CYP3A4 microsomal expression might be artificially high as a consequence of

recovery or degradation issues during the microsomal preparation. Although another potential

explanation maybe that some of the substrates selected are not exclusively cleared by CYP3A4 (for

instance: efavirenz (CYP2B6); rapaglinide (CYP2C8); loratidine (CYP2D6); diazepam (CYP2C19))

and consequently the impact of variable CYP3A4 expression is attenuated by the additional clearance

pathway. Although Kato et al. show a lower %CV for lower clearance CYP3A4 substrates than the

other reports discussed above, the %CV of AZD3514 in patients is still towards the lower end of

variability reported by Kato et al for drugs with similar clearance. Based on the literature review, the

17

PK variability of AZD3514 appears to be unusually low considering the relatively strong evidence that

CYP3A4 is predominately responsible for AZD3514 clearance.

Ethnicity is a demographic variable that may contribute to inter-individual variability in

pharmacokinetics and/or pharmacodynamics of drugs (26). Although ethnic differences in exposure

and/or response are not uncommon, in only a few cases have they led to population-specific prescribing

recommendations (27). In the present study, it was shown that exposure to AZD3514 was slightly

higher in Japanese patients when compared to Western patients; however, this difference was no longer

apparent after normalization for body weight. The observed smaller volume of distribution in Japanese

patients is consistent with this. Differences exist between Asian and Western patients regarding

CYP3A4 liver content (28) but the activity of this and that of other major drug metabolizing enzymes

was similar in these populations (29) and, therefore, do not appear to contribute to any observed

differences in pharmacokinetics.

Testing the androgen modulation hypothesis

The introduction of PSA as a biomarker for prostate cancer was an important step forward in the ability

to diagnose this disease and offer the patient earlier and more effective treatment (13). Not only is PSA

used to diagnosis this disease, but also to monitor the diseases progress and assess therapeutic response.

PSA is the primary clinical biomarker available for prostate cancer that is used outside the purely

research environment (13). Unfortunately, treatment with AZD3514 in patients did not result in a dose-

dependent and consistent decrease in PSA. From animal experiments it was predicted that in order for

AZD3514 to decrease PSA in patients, the plasma concentrations of this compound needed to be above

2410 ng/mL for at least 18 h during a 24 h period. At these exposures of AZD3514 in LNCaP cells,

PSA mRNA expression was reduced by 90-100% (30). It became clear that this goal could not be

achieved with QD dosing and the dosing frequency was increased to BID. However, due to a temporal

change in PK, the target coverage was not achieved. Simulations suggested that dose escalating to

18

2000 mg BID could have provided sufficient coverage for efficacy, but the occurrence of non-tolerated

gastrointestinal adverse events lead to discontinuation of 2000 mg BID dosing in patients (14).

Although AZD3514 did not decrease PSA in general, some patients showed clinically

significant decreases in PSA (Figure 5). A systems pharmacology modeling approach was undertaken

with the expectation that it might help inform the ongoing clinical study design. Initial empirical

modeling on the change in PSA over time found that no PK variables, such as the area under the curve,

and minimum and maximum plasma concentrations, were strong correlates of the growth constant.

Subsequently, other variables were investigated, such as baseline values of markers, co-medications,

age, pre-treatment PSA trajectory, and any other variables that had been collected. Counter-intuitively,

considering the mode of action of AZD3514, baseline PSA was identified as the only potentially

predictive covariate (18). Subsequently a mechanistic model was developed that indicated that

AZD3514 may be more active in a low dihydrotestosterone (DHT) environment (18). PSA gene

expression is reliably stimulated by androgens such as DHT (31), hence providing a potential

hypothesis why low baseline PSA may influence AZD3514 activity. This finding was the basis for

investigating a combination drug strategy which involved administration of AZD3514 in combination

with abiraterone, a drug which decreases DHT (and consequently PSA) in CRPC patients (32), to

cohorts of patients in study 1 (Figure 6). The combination dose for AZD3514 was 500 mg BID as this

was expected to be well tolerated for long term treatment and to provide a safety margin as no data

were available on the safety of the abiraterone and AZD3514 combination. Recruitment of patients

that had progressed on abiraterone proved slow and the mean PSA baseline in this group was no lower

than that of the 1000 mg QD AZD3514 monotherapy group (Figure 4). The addition of AZD3514 to

the treatment regimen of abiraterone-treated patients did not produce any meaningful decrease in PSA

in the 5 patients recruited into the cohort. As a result, an early futility analysis was performed which

gave a low probably of achieving target efficacy and the study was stopped. However, it can be argued

that the low PSA baseline hypothesis was not fully tested in the clinical study due to the lack of patients

Deleted: 0

Deleted: 1

19

recruited with low baseline PSA, the low AZD3514 exposure achieved and the small number of

patients recruited. Further exploration of this hypothesis with other drug therapies is worthy of

consideration in prostate cancer research.

In conclusion, the emerging unexpected clinical PK of AZD3514 and systems pharmacology

modeling (18) that suggested greater efficacy may be achieved with low baseline androgen resulted in

the transition to an adaptive clinical trial design to explore a drug combination strategy with

abiraterone. However, due to the time dependent decrease in exposure which was exacerbated with

twice daily dosing, coverage above the target concentration could not be achieved. Nevertheless, the

observation that in some patients a clinically meaningful decrease in PSA was observed may indicate

that follow-up compounds with a similar mechanism of action but with an improved pharmacokinetic

and safety profile may prove to be useful in the treatment of CRPC either as monotherapy or in

combination with abiraterone. Additionally, AZD3514 as a CYP3A4 substrate with low variability

and time dependent pharmacokinetics that do not seem to be explained by CYP3A4 induction may be

an interesting tool compound for academic research that may provide new insights into drug

disposition.

ACKNOWLEDGEMENTS

The authors thank AZD3514 study 1 and 2 clinical investigators and patients, Henk Poelman (PRA

International, Assen, The Netherlands) for the bioanalytical work and Paul van Giersbergen (Van

Giersbergen Consulting, Wuenheim, France) for editorial assistance.

Declaration of competing interests: All authors were at the time of study conduct employees of

AstraZeneca with the exception on Gareth D. James, who received payment from AstraZeneca for

services rendered. P. Dickinson owns shares in AstraZeneca and is Director of a company with a

contract to provide services to AstraZeneca.

20

REFERENCES

1. Acar O, Esen T, Lack NA. New therapeutics to treat castrate-resistant prostate cancer.

ScientificWorldJournal. 2013 May 27;2013:379641. doi: 10.1155/2013/379641.

2. Rove KO, Crawford ED. Traditional androgen ablation approaches to advanced prostate cancer:

new insights. Can J Urol. 2014 Apr;21(2 Supp 1):14-21.

3. Thomas BM, Smith C, Evans J, Button MR, Kumar S, Palaniappan N, et al. Time to prostate

specific antigen (PSA) nadir may predict rapid relapse in men with metastatic castration-resistant

prostate cancer (CRPC) receiving docetaxel chemotherapy. Med Oncol. 2013 Dec;30(4):719. doi:

10.1007/s12032-013-0719-0.

4. Thoreson GR, Gayed BA, Chung PH, Raj GV. Emerging therapies in castration resistant prostate

cancer. Can J Urol. 2014 Apr;21(2 Supp 1):98-105.

5. Vaishampayan U. Therapeutic options and multifaceted treatment paradigms in metastatic

castrate-resistant prostate cancer. Curr Opin Oncol. 2014 May;26(3):265-73. doi:

10.1097/CCO.0000000000000066.

6. Toren PJ, Gleave ME. Evolving landscape and novel treatments in metastatic castrate-resistant

prostate cancer. Asian J Androl. 2013 May;15(3):342-9. doi: 10.1038/aja.2013.38.

7. Montgomery RB, Mostaghel EA, Vessella R, Hess DL, Kalhorn TF, Higano CS, et al.

Maintenance of intratumoral androgens in metastatic prostate cancer: A mechanism for castration

resistant tumor growth. Cancer Res. 2008 Jun 1;68(11):4447-54. doi: 10.1158/0008-5472.CAN-

08-0249.

8. Knudsen KE, Penning TM. Partners in crime: Deregulation of AR activity and androgen synthesis

in prostate cancer. Trends Endocrinol Metab. 2010 May;21(5):315-24. doi:

10.1016/j.tem.2010.01.002.

21

9. Linja MJ, Savinainen KJ, Saramaki OR, Tammela TL, Vessella RL, Visakorpi T. Amplification

and overexpression of androgen receptor gene in hormone-refractory prostate cancer. Cancer Res.

2001 May 1;61(9):3550-5.

10. Taplin ME, Bubley GJ, Ko YJ, Small EJ, Upton M, Rajeshkumar B, et al. Selection for androgen

receptor mutations in prostate cancers treated with androgen antagonist. Cancer Res. 1999 Jun

1;59(11):2511-5.

11. Bradbury RH, Acton DG, Broadbent NL, Brooks AN, Carr GR, Hatter G, et al. Discovery of

AZD3514, a small-molecule androgen receptor downregulator for treatment of advanced prostate

cancer. Bioorg Med Chem Lett. 2013 Apr 1;23(7):1945-8. doi: 10.1016/j.bmcl.2013.02.056.

12. Owens W, Zeiger E, Walker M, Ashby J, Onyon L, Gray LE Jr. The OECD program to validate

the rat Hershberger bioassay to screen compounds for in vivo androgen and antiandrogen

responses. Phase 1: use of a potent agonist and a potent antagonist to test the standardized protocol.

Environ Health Perspect. 2006 Aug;114(8):1259-65.

13. Crawford ED, Ventii K, Shore ND. New biomarkers in prostate cancer. Oncology. (Williston

Park) 2014 Feb;28(2):135-42.

14. Omlin A, Jones RJ, van der Noll R, Satoh T, Niwakawa M, Smith SA, et al. AZD3514, an oral

selective androgen receptor down-regulator in patients with castration-resistant prostate cancer –

Results of two parallel first-in-human Phase I studies. Invest New Drugs. 2015 Jun;33(3):679-90.

doi: 10.1007/s10637-015-0235-5.

15. Gough K, Hutchison M, Keene O, Byrom B, Ellis S, Lacey L, McKellar J. Assessment of dose

proportionality: report from the statisticians in the pharmaceutical industry/ pharmacokinetics UK

joint working group. Drug Information Journal 1995; 29: 1039-48. doi:

10.1177/009286159502900324.

16. Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of

clinical measurement. Lancet. 1986 Feb 8;1(8476):307-10.

22

17. FDA Guidance for Industry. Drug interaction studies — study design, data analysis, implications

for dosing, and labeling recommendation. Draft Guidance, February, 2012. Available at:

http://www.fda.gov/downloads/drugs/guidancecomplianceregulatoryinformation/guidances/ucm

292362.pdf

18. Mistry H, Fabre MA, Young J, Clack G, Dickinson PA. Systems pharmacology modeling of

prostate-specific antigen in prostate cancer patients treated with an androgen receptor antagonist

and down-regulator. CPT Pharmacometrics Syst Pharmacol, 2016 e-published.

doi:10.1002/psp4.12066

19. Wang YM, Ong SS, Chai SC, Chen T. Role of CAR and PXR in xeno-biotic sensing and

metabolism. Expert Opin Drug Metab Toxicol. 2012;8(7):803–817.

20. Ma X, Idle JR, Gonzalez FJ. The pregnane X receptor: From bench to bedside. Expert Opin Drug

Metab Toxicol. 2008;4(7):895–908.

21. Benet LZ. A Holy Grail of clinical pharmacology: prediction of drug pharmacokinetics and

pharmacodynamics in the individual patient. Clin Pharmacol Ther. 2009 Aug;86(2):133-4. doi:

10.1038/clpt.2009.102.

22. Masica AL, Mayo G, Wilkinson GR. In vivo comparisons of constitutive cytochrome P450 3A

activity assessed by alprazolam, triazolam, and midazolam. Clin Pharmacol Ther. 2004

Oct;76(4):341-9. doi:10.1016/j.clpt.2004.07.003.

23. Kharasch ED, Walker A, Hoffer C, Sheffels P. Sensitivity of intravenous and oral alfentanil and

pupillary miosis as minimally invasive and noninvasive probes for hepatic and first-pass CYP3A

activity. J Clin Pharmacol. 2005 Oct;45(10):1187-97. doi: 10.1177/0091270005280077.

24. Eap CB, Buclin T, Cucchia G, Zullino D, Hustert E, Bleiber G. Oral administration of a low dose

of midazolam (75 microg) as an in vivo probe for CYP3A activity. Eur J Clin Pharmacol. 2004

Jun;60(4):237-46. doi: 10.1007/s00228-004-0762-z.

23

25. Kato M, Chiba K, Ito T, Koue T, Sugiyama Y. Prediction of interindividual variability in

pharmacokinetics for CYP3A4 substrates in humans. Drug Metab Pharmacokinet.

2010;25(4):367-78. doi: 10.2133/dmpk.DMPK-09-RG-038.

26. Xie HG, Kim RB, Wood AJ, Stein CM. Molecular basis of ethnic differences in drug disposition

and response. Annu Rev Pharmacol Toxicol. 2001;41:815-50. Doi:

10.1146/annurev.pharmtox.41.1.815.

27. Bjornsson TD, Wagner JA, Donahue SR, Harper D, Karim A, Khouri MS, et al. A review and

assessment of potential sources of ethnic differences in drug responsiveness. J Clin Pharmacol.

2003 Sep;43(9):943-67. doi: 10.1177/0091270003256065.

28. Ramamoorthy A, Pacanowski MA, Bull J, Zhang L. Racial/ethnic differences in drug disposition

and response: review of recently approved drugs. Clin Pharmacol Ther. 2015 Mar;97(3):263-73.

doi: 10.1002/cpt.61.

29. Myrand SP, Sekiguchi K, Man MZ, Lin X, Tzeng RY, Teng CH, et al. Pharmacokinetics/genotype

associations for major cytochrome P450 enzymes in native and first- and third-generation

Japanese populations: comparison with Korean, Chinese, and Caucasian populations. Clin

Pharmacol Ther. 2008 Sep;84(3):347-61. doi: 10.1038/sj.clpt.6100482.

30. Loddick SA, Ross SJ, Thomason AG, Robinson DM,Walker GE, Dunkley TPJ, et al. AZD3514:

A small molecule that modulates androgen receptor signaling and function in vitro and in vivo.

Mol Cancer Therapeutics 2013; 12(9): 1715-1727.

31. Zhu YS1, Cai LQ, You X, Cordero JJ, Huang Y, Imperato-McGinley J. Androgen-induced

prostate-specific antigen gene expression is mediated via dihydrotestosterone in LNCaP cells. J

Androl. 2003 Sep-Oct;24(5):681-7. doi: 10.1002/j.1939-4640.2003.tb02727.x.

32. Leibowitz-Amit R, Templeton AJ, Omlin A, Pezaro C, Atenafu EG, Keizman D, et al. Clinical

variables associated with PSA response to abiraterone acetate in patients with metastatic

Deleted: <#>¶

Formatted: Font: (Default) Times New Roman, 12 pt,Swedish (Sweden)

Formatted: Font: (Default) Times New Roman, 12 pt,Swedish (Sweden)

Formatted: Font: (Default) Times New Roman, 12 pt,Swedish (Sweden)

Formatted: Font: (Default) Times New Roman, 12 pt,Swedish (Sweden)

Formatted: Font: (Default) Times New Roman, 12 pt,Swedish (Sweden)

Formatted: Font: (Default) Times New Roman, 12 pt,Swedish (Sweden)

Formatted: Font: (Default) Times New Roman, 12 pt,Swedish (Sweden)

24

castration-resistant prostate cancer. Ann Oncol. 2014 Mar;25(3):657-62. doi:

10.1093/annonc/mdt581.

25

FIGURE LEGENDS

Figure 1 AZD3514 geometric mean single-dose plasma concentration-time profiles (linear scale and

semi-logarithmic scales).

Figure 2 Individual changes in apparent clearance from Day 1 to Day 29

Figure 3 Effect of race on AZD3514 exposure - individual single-dose body weight and dose-

normalized AUC0-24h

Figure 4 PSA responses after treatment with AZD3514 - Group mean standard deviation

Figure 5 Effect of AZD3514 on individual patients PSA responses - Semi-logarithmic scale

Figure 6 Clinical study design adaptation resulting from emerging study data

Supplementary Figure 1 Scatter plot of individual ln(AUC) values comparing Day 1 and Day 8 (top

panel) and a Bland-Altman plot of this data (bottom panel).

Supplementary Figure 2 Diagnostic plots of the final population pharmacokinetic model -Visual

Predictive Check for each day of treatment (A), at single dose and at steady-state (B) and for the 250

mg, 500 mg and 1000 mg dose groups (C)

26

Figure 1

0

1500

3000

4500

6000

7500

9000

0 1 2 3 4 5 6 8 10 24

Time after administration (h)

AZ

D3514 c

on

cen

trati

on

(n

g/m

l)

0 24 48 72 961

10

100

1000

10000

250 mg

500 mg

1000 mg

100 mg

Time after administration (h)

AZ

D3514 c

on

cen

trati

on

(n

g/m

l)

27

Figure 2

N=11P = 0.014

N=6P = 0.001

28

Figure 3

0

2000

4000

6000

8000

10000

250 500 1000

Western patients

Japanese patients

Single dose of AZD3514 (mg)

Bo

dy-w

eig

ht

an

d d

ose-n

orm

ali

zed

AU

C0

-24

h[(

ng

.h/m

L)/

mg

]

29

Figure 4

30

Figure 5

31

Figure 6

AZD3514 Monotherapy

Dose escalation at once daily dosing

Once daily dosing insufficient to achieve target exposure duration

Initiated twice daily dosing

Marked time dependent reduction in exposure became apparent with twice daily dosing

Continued dose escalation

High doses of AZD3514 not tolerated and target exposure for reduction of PSA was not achieved

SARD mechanism of action and PKPD modelling suggest greater benefit at low circulating testosterone setting, indicating greater

efficacy may be achieved at lower doses of AZD3514

Change in strategy from monotherapy to combination therapy with abiraterone – patients on abiraterone recruited

Early futility “Go/No go” criteria not achieved at 500 mg bid AZD3514 in combination with abiraterone

Project stopped

Explore possible reasons for time dependent change in PK:

Initiate measurement of biomarker 4-β hydroxycholesterol for indication of CYP enzyme

induction in the clinic in new patients

Studies closed prior to collection of samples

We

ste

rn a

nd

Ja

pa

ne

se

po

pu

latio

n

We

ste

rn p

op

ula

tio

n o

nly

Response to emerging study data

New emerging study data

Key

32

Supplementary Figure 1

8 9 10 11 128

9

10

11

12

100 mg

250 mg

500 mg

1000 mg

Ln(AUC) at day 1

Ln

(AU

C)

at

day 8

8 9 10 11 12-1.0

-0.5

0.0

0.5

1.0

100 mg

250 mg

500 mg

1000 mg

Mean Ln(AUC)

Ln

(AU

C)

at

day 8

- L

n(A

UC

) at

day 1

33

Supplementary Figure 2

(A) Day of treatment

34

(B) Single Dose vs. Steady-State

35

(C) For 250, 500 and 1000 mg

36

Table 1 Single-dose pharmacokinetic variables of AZD3514 in Western (study 1) and Japanese (study 2) CRPC patients

Study 1 Study 2

Variable 100 mg QD

(N=4)

250 mg QD

(N=6)

500 mg QD

(N=12)

1000 mg QD

(N=12)

1000 mg BID

(N=9)

2000 mg QD

(N=5)

250 mg QD

(N=4)

500 mg QD

(N=4)

1000 mg QD

(N=5)

AUC0- (ng.h/mL) 7270 (1.3) 19350 (20) 34110 (23) 70560 (28) 62470 (39) 228000 (25) 18850 (44) 41950 (20) 91750 (26)

Cmax (ng/mL) 982 (19) 3356 (31) 4679 (21) 9519 (21) 9350 (50) 19830 (35) 3451 (32) 5949 (15) 12970 (21)

tmax (h) 2.0 (1.0-2.0) 2.0 (0.5-2.0) 2.0 (2.0-4.0) 2.0 (0.6-3.1) 2.9 (2.0-4.0) 3.0 (1.0-5.0) 2.0 (0.9-2.1) 2.0 (1.0-3.0) 2.9 (2.0-3.0)

t1/2 (h) 16 (1.9) 19 (7.2) 16 (5.8) 16 (5.5) 12 (3.1) 16 (8.4) 8 (2.0) 11 (2.1) 12 (3.4)

CL/F (L/h) 14 (0.2) 13 (3.0) 15 (3.9) 15 (4.2) 17 (7.0) 9 (2.3) 14 (5.4) 12 (2.4) 11 (2.7)

V/F (L) 138 (4.9) 123 (36) 125 (30) 121 (46) 125 (34) 101 (43) 94 (24) 95 (14) 83 (22)

Data are expressed as geometric mean (%CV) for AUC and Cmax, median (range) for tmax, and arithmetic mean (SD) for t1/2, CL/F and V/F

QD, once daily; BID; twice daily; AUC0-, area under the curve from time zero to infinity; Cmax, maximum concentration; tmax, time to maximum concentration; t1/2, terminal

elimination half-life; CL/F, apparent clearance; V/F, apparent volume of distribution

37

Table 2 Steady state pharmacokinetic variables of AZD3514 in Western (study 1) and Japanese (study 2) CRPC patients

Study 1 Study 2

Variable

Monotherapy Combination

with abiraterone Monotherapy

100 mg QD

(N=5)

250 mg QD

(N=6)

500 mg QD

(N=12)

1000 mg

QD (N=12)

1000 mg BID

(N=9)

500 mg BID

(N=5)

250 mg QD

(N=4)

500 mg QD

(N=4)

500 mg BID

(N=5)

AUCtau (ng.h/mL) 6662 (8.9) 15640 (35) 27180 (29) 55830 (25) 36890 (23) 20650 (30) 16180 (34) 41330 (15) 23320 (20)

Cssmax (ng/mL) 1059 (4.6) 3054 (40) 5086 (27) 9194 (24) 9750 (36) 4914 (19) 3218 (17) 6809 (10) 5099 (26)

tssmax (h) 1.0 (0.9-1.0) 2.0 (1.0-2.1) 2.0 (1.0-4.0) 2.0 (1.0-3.3) 1.0 (1.0-2.2) 2.1 (2.0-2.3) 1.5 (0.5-3.0) 1.5 (0.5-2.0) 2.9 (1.9-3.0)

CLss/F (L/h) 15 (1.3) 17 (5.0) 19 (6.0) 18 (4.3) 25 (4.9) 25 (7.8) 16 (5.0) 12 (2.0) 22 (4.2)

Accumulation ratio 1.05 (0.05) 0.96 (0.25) 0.93 (0.20) 0.91 (0.19) 0.75 (0.08) NC 0.91 (0.06) 1.07 (0.20) NC

Temporal change

parameter 0.91 (0.08) 0.86 (0.23) 0.87 (0.19) 0.85 (0.16) 0.70 (0.07) NC 0.86 (0.08) 1.00 (0.18) NC

Cssmin (ng/mL) 38.9 (45) 8.74 (32) 112 (31) 219 (41) 519 (29) 324 (55) 67 (94) 189 (33) 535 (32)

Time above target

concentration (h) in

24-hour period

<1 1.5

(<1.0-2.5)

4.3

(1.0-5.5)

6.5

(4.5-10.0)

10.5

(4.0-12.0)

6.5

(3.0-10.0)

1.0

(1.0-2.0)

4.8

(4.0-6.0)

7.0

(6.0-10.0)

Data are expressed as geometric mean (%CV) for AUC, Cssmax and Cssmin, median (range) for tssmax, and arithmetic mean (SD) for CLss/F, accumulation ratio, temporal

change parameter and time above target concentration in 24-hour period

QD, once daily; BID; twice daily; AUCtau, area under the curve during a dose interval; Cssmax, maximum concentration at steady state; tssmax, time to maximum concentration

at steady state; Cssmin, minimum concentration at steady state

38

Table 3 Dose proportionality of AZD3514 pharmacokinetics

Cmax slope

(90% confidence interval)

AUC* slope

(90% confidence interval)

Study 1 Study 2 Study 1 Study 2

Day 1 0.92 (0.81-1.03) 0.96 (0.75-1.17) 0.96 (0.84-1.08) 1.14 (0.89-1.39)

Day 29 0.91 (0.80-1.01) NC 0.92 (0.81-1.03) NC

NC = not calculable

* AUC, area under the curve (AUC0- Day 1; AUCtau at steady state for day 29)

39

Table 4 Final population pharmacokinetic parameters

Pharmacokinetic model

Parameter Mean SE 95 CI CV%

Single-dose CL/F (L/h) 15.9 0.759 [14.4; 17.4]

Steady-state CL/F (L/h) 21.5 1.06 [19.4; 23.6]

V1/F (L) 84.9 4.3 [76.5; 93.3]

V2/F – single dose (L) 174 19.9 [135; 213]

V2/F – steady state (L) 317 37.9 [243; 391]

Inter-compartmental clearance (L/h) 3.91 0.267 [3.4; 4.4]

First-order absorption (h-1) 3.23 0.149 [2.94; 3.52]

Dose effect on first-order

absorption

0.25 0.0615 [0.13; 0.37]

Proportional error parameter 0.835 0.0447 [0.750; 0.920]

Additive error parameter 3.91 1.43 [1.11; 6.71]

Variance IIV of CL/F 0.0651 0.0187 [0.0284; 0.1018] 26%

Variance IIV of V1/F 0.168 0.0641 [0.0424; 0.2936] 41%

Variance IIV of V2/F 0.0362 0.0169 [0.0031; 0.0693] 19%

CI, confidence interval; CV%, coefficient of variation; IIV, inter-individual variability; TDD, total

daily dose; CL/F, apparent clearance; V1/F, volume of the central compartment, V2/F, volume of the

peripheral compartment;

40

Supplementary Table 1 Pharmacokinetic inter- and intra-patient variability

Summary statistics for ln(AUC) for each dose

Dose N Mean SE Min Max

100 mg 8 8.5 0.11 8.4 8.7

250 mg 12 9.5 0.32 8.9 10.1

500 mg 18 10.1 0.28 9.7 10.6

1000 mg 32 10.8 0.26 10.2 11.4

Statistics comparing day 1 and day 8 ln (AUC)

Correlation between day 1 and day 8 ln(AUC) Result

100 mg 0.90

250 mg 0.75

500 mg 0.95

1000 mg 0.41

Percentage change in ln(AUC) from day 1 to 8

Number (%) patients change ≤ 2%* 29 (82.9)

Number (%) patients change >2% to 5%* 4 (11.4)

Number (%) patients change >5% to 10%* 2 (5.7)

Number (%) patients change >10%* 0 (0.0)

Mean (%) change 0.13

Standard deviation % change 2.28

Largest % decrease -7.99

Largest % increase 5.31

Bland Altman statistics comparing day 1 and 8 ln(AUC)

Mean bias 0.012

Largest negative difference -0.914

Largest positive difference 0.551

Standard deviation 0.194

Standard error 0.033

95% confidence interval -0.055, 0.078

*Change from day 1 to day 8 ln(AUC) can be negative or positive.

AUC, area under the curve