The Journal of Clinical

Pharmacology
http://www.jclinpharm.org

Effect of Different Durations of Ketoconazole Dosing on the Single-Dose Pharmacokinetics of

Midazolam: Shortening the Paradigm
S. A. Stoch, E. Friedman, A. Maes, K. Yee, Y. Xu, P. Larson, M. Fitzgerald, J. Chodakewitz and J. A. Wagner
J. Clin. Pharmacol. 2009; 49; 398 originally published online Feb 26, 2009;
DOI: 10.1177/0091270008331133

The online version of this article can be found at:

http://www.jclinpharm.org/cgi/content/abstract/49/4/398

Published by:

http://www.sagepublications.com

On behalf of:
American College of Clinical Pharmacology

Additional services and information for The Journal of Clinical Pharmacology can be found at:

Email Alerts: http://www.jclinpharm.org/cgi/alerts

Subscriptions: http://www.jclinpharm.org/subscriptions

Reprints: http://www.sagepub.com/journalsReprints.nav

Permissions: http://www.sagepub.com/journalsPermissions.nav

Downloaded from http://www.jclinpharm.org by Luciana Oliveira on April 19, 2009


Effect of Different Durations of Ketoconazole
Dosing on the Single-Dose Pharmacokinetics of
Midazolam: Shortening the Paradigm
S. A. Stoch, MD, E. Friedman, MS, A. Maes, PhD, K. Yee, PhD, Y. Xu, PhD,
P. Larson, MS, M. Fitzgerald, MD, J. Chodakewitz, MD, and J. A. Wagner, MD, PhD
Given the prominent role of cytochrome P450 3A (CYP3A)
in the metabolism of drugs, it is critical to determine
whether new chemical entities will be affected by the inhi-
bition of this enzyme system and result in clinically rele-
vant drug interactions. Ketoconazole interaction studies
are frequently performed to determine a given compound’s
sensitivity to CYP3A metabolism. The present study evalu-
ated whether probing a sensitive substrate (midazolam)
with a potent inhibitor (ketoconazole) at earlier time points
(days 1 or 2) might be used to reliably gauge the magnitude
of a meaningful interaction. The geometric mean ratios
(ketoconazole + midazolamday 5/ketoconazole + mida-
T he CYP3A family of enzymes plays a major role
in the biotransformation of many commonly
used drugs1 and is present in the liver and gut.
Therefore, the importance of assessing the impact of
CYP3A inhibition on the metabolism of a new chem-
ical entity (NCE) cannot be overemphasized. Drug-
drug interactions (DDIs) can lead to important severe
side effects and have previously resulted in the
withdrawal of approved drugs.2 The relevance of
assessing this early in drug development cannot be
underestimated, particularly in evaluating NCEs
with a projected narrow therapeutic index. Therefore,
the ability to reliably gauge the magnitude of a clini-
cally meaningful DDI early in development affords a
From Merck & Co, Inc, Whitehouse Station, New Jersey (Dr Stoch, E.
Friedman, Dr Maes, Dr Yee, Dr Xu, P. Larson, Dr Chodakewitz, Dr
Wagner) and ProMedica Clinical Research Center, Inc, Boston,
Massachusetts (Dr Fitzgerald). Submitted for publication July 28, 2008;
revised version accepted December 19, 2008. Address for correspond-
ence: S. Aubrey Stoch, MD, Merck & Co, Inc, RY34-A500, PO Box
2000, Rahway, NJ 07065-0900; e-mail: aubrey_stoch@merck.com.
DOI: 10.1177/0091270008331133
398 • J Clin Pharmacol 2009;49:398-406
title
zolamday 1 and ketoconazole + midazolamday 5/ketoco-
nazole + midazolamday 2) for midazolam AUC0-∞ were
1.36 and 1.06 with corresponding 90% confidence inter-
vals of (1.17, 1.57) and (0.83, 1.23), respectively. These
findings suggest that short-term drug-drug interaction
studies can predict the magnitude of change in AUC as
reliably as studies using longer duration treatments.
Keywords:  Midazolam; ketoconazole; drug-drug inter-
actions; cytochrome P450 3A
Journal of Clinical Pharmacology, 2009;49:398-406
© 2009 the American College of Clinical Pharmacology
unique opportunity to prioritize compounds and
select molecules with a lower likelihood of being a
sensitive CYP3A substrate. If an investigational drug
is a putative substrate, an assessment is made based
both on in vitro and in vivo interaction studies.
Based on these data, clinical studies are performed
to assess the likelihood and the magnitude of the
effect that may be anticipated in the clinic. From a
drug development standpoint, it is preferable to
make an informed decision as early as possible as to
whether a large drug interaction liability may exist.
It may be possible to evaluate this after a single or
limited multiple-dose administration.
To date, no prospective within-subject clinical study
data have been reported in the literature to evaluate the
changes in the pharmacokinetics of midazolam, a
sensitive CYP3A substrate, following the administration
of different durations of ketoconazole, a strong CYP3A
inhibitor. The present study was designed to explore
whether shortening the “traditional” duration of
ketoconazole dosing from 5 days to either 2 days or a
single-dose administration could be used to reliably
predict clinically meaningful effects on midazolam
 EFFECT OF KETOCONAZOLE DOSING ON MIDAZOLAM
pharmacokinetics. A 5-day ketoconazole dosing
paradigm with the administration of a 2 mg3 dose of
midazolam on day 5 has been used by certain sponsors
to study the effect of a potent CYP3A inhibitor on the
pharmacokinetics of a sensitive CYP3A substrate. It is
important to ascertain whether these trials can be
shortened without compromising their predictive
value. This would minimize the exposure of healthy
volunteers while at the same time realizing more
nimble drug development and more efficient decision
making. This study was also simulated using a
ketoconazole-interaction model developed by Chien
et al,4 in which the inhibitory effect of ketoconazole
was explicitly accounted for by changes in the
substrate’s bioavailability and systemic clearance
due to ketoconazole concentrations in the gut, plasma,
and liver. If the model is able to successfully replicate
results observed clinically, it may provide insight into
the mechanisms behind the experimentally observed
results and serve as a predictive tool to supplement
abbreviated ketoconazole DDI studies.
METHODS
Participants
A total of 12 healthy, white, male subjects were
enrolled in this study. All participants were non-
smokers with a mean age of 38.8 years (range, 21-54
years) and a body mass index of 18.5 to 27.0 kg/m2
with corresponding body weights between 64 and 89
kg. Participants were in good general health accord-
ing to routine medical history, physical examina-
tion, vital signs, 12-lead electrocardiogram, and
laboratory data.
Every participant gave written informed consent to
participate in this study. The protocol was approved
by the institutional review board (New England
Institutional Review Board) and was conducted in
accordance with the guidelines on good clinical
practice and with ethical standards for human
experimentation established by the Declaration of
Helsinki Principles.
Study Design
This was a 4-period, open-label, randomized, cross-
over study in 12 healthy male participants. The 4
treatments were as follows:
Treatment A: a single 2 mg dose of midazolam
Treatment B: a single 2 mg dose of midazolam coadmin-
istered simultaneously with a single dose of 400
mg ketoconazole
DRUG INTERACTIONS
Treatment C: multiple doses of 400 mg ketoconazole for 2
days with 2 mg midazolam coadministered on day 2
Treatment D: multiple doses of 400 mg ketoconazole for 5
days with 2 mg midazolam coadministered on day 5
Participants were randomly assigned to 1 of the 4
treatment sequences, and there was a minimum of 7
days between each treatment period. All doses of
midazolam were administered in the fasted state.
Ketoconazole was administered without regard to
food except on days when it was coadministered
with midazolam, in which case ketoconazole was
given fasted. All doses of ketoconazole and midazo-
lam were administered in the morning with 240 mL
of water. No acidification was employed in the
administration of either study drug.5
Midazolam
Midazolam (Roxane Laboratories, Columbus, Ohio) was
administered as a 2 mg oral dose (2.0 mg/mL midazo-
lam HCL syrup). All doses of midazolam were adminis-
tered with 240 mL of water. On days of coadministration
with ketoconazole, midazolam was given 5 minutes
after the ketoconazole dose. All doses of midazolam
were administered following an 8-hour fast.
Ketoconazole
Ketoconazole (Teva Pharmaceuticals, Petah Tikva,
Israel) was administered as oral 400 mg doses (1 ×
400‑mg ketoconazole tablet). All doses of ketocon-
azole were administered with 240 mL of water and
without regard to food except when coadministered
with midazolam, in which case both drugs were
given after an 8-hour fast.
Pharmacokinetic Assessments
Blood samples for midazolam pharmacokinetics were
collected predose and at various time points post-
dose. Immediately after collection, blood samples
were kept on ice, for no more than 30 minutes, until
centrifugation at 1500 g for 15 minutes at 4°C. The
plasma fraction was removed and stored at –20°C
until analyzed for midazolam concentrations.
When midazolam was administered alone, blood
samples were drawn, and plasma was collected
predose and at 0.25, 0.5, 1, 2, 4, 6, 8, and 12 hours after
dosing. When midazolam was coadministered with
ketoconazole, on day 1, 2, or 5, blood samples were
drawn, and plasma was collected predose and at 0.25,
0.5, 1, 2, 4, 6, 8, 12, 16, and 24 hours after dosing.
Plasma concentration of midazolam was determined
by reverse-phase high-performance liquid chroma­
tography (HPLC) coupled with tandem quadrupole
399
 STOCH ET AL
mass spectrometry (MS/MS). The sample analysis was
conducted at Pharmaceutical Product Development
(Wilmington, North Carolina). Midazolam and its
isotopically labeled internal standard were extracted
from human plasma through liquid-liquid extraction
using methyl-t-butyl ether and separated by reverse-
phase chromatography using a Thermo Aquasil C18
column (2.1 × 100 mm, 5 mm) with a gradient composed
of acetonitrile, water, and ammonium acetate. The
analytes were detected using an MDS Sciex API 3000
triple quadrupole mass spectrometer system. Ionization
occurred using positive ion Turbo IonSpray, and the
compounds were detected at the mass transitions m/z =
326 → 291 and m/z = 330 → 295 for midazolam and its
internal standard, respectively. The data system was
configured to calculate and annotate the areas of
midazolam and internal standard peaks automatically.
A calibration curve was constructed using peak area
ratios of the calibration standards by applying a linear
weighted (1/concentration squared) least squares
regression algorithm. The midazolam concentrations
in clinical samples were then back-calculated from
their peak area ratios of midazolam versus internal
standard against the calibration curve. The lower limit
of quantitation (LLOQ) for midazolam was 0.1 ng/mL.
Calibration standards ranged from 0.1 to 100 ng/mL.
With 0.3 ng/mL (3× LLOQ) midazolam quality control
samples, the accuracy and precision were 91.24% and
7.17% (% coefficient of variation [CV]), respectively,
across the analytical runs (n = 5).
Statistical Analyses
The effects of multiple oral doses of 400 mg ketocon-
azole once daily on the midazolam single-dose phar-
macokinetic parameters (AUC0-∞, Cmax, tmax, and
apparent terminal t1/2) were evaluated using a linear
mixed effects model with fixed effects for period and
duration of ketoconazole administration (0, 1, 2 or 5
days) and a random effect for subject. The assump-
tions of the linear model were tested and found to be
satisfied. AUC0-∞ and Cmax were log-transformed, tmax
was ranked, and the reciprocal of apparent terminal
t1/2 was taken prior to conducting the analyses. Back-
transformed summary statistics and inferential
results are reported.
The primary hypothesis of this study was that the
midazolam AUC0-∞ following a single 2 mg dose
coadministered on day 5 of once-daily dosing of 400
mg ketoconazole is less than 2-fold than that of a 2-mg
dose coadministered on day 1 or 2. That is, the
midazolam AUC0-∞ geometric mean ratio (ketoconazole
+ midazolamday 5/ketoconazole + midazolamday 1 or 2) is
400  •  J Clin Pharmacol 2009;49:398-406
contained within the interval (0.50, 2.00). To address
this hypothesis, we followed a closed testing procedure
starting with comparison of ketoconazole administra­
tion on day 5 versus day 2, then day 5 versus day 1.
Two-sided 90% confidence intervals for the midazolam
AUC0-∞ geometric mean ratios (GMRs, midazolam +
ketoconazoleMD, day 5/midazolam + ketoconazoleMD, day 1 or 2)
were calculated from the above mixed effects model. If
the 90% confidence interval for the AUC0-∞ GMR was
contained within the prespecified bounds of (0.50,
2.00), then no clinically meaningful decrease in AUC0-∞
after ketoconazole administration on day 2 (day 1) was
concluded.
Modeling and Simulation
Observed concentration-time profiles from adminis-
tration of a single 2 mg dose of midazolam without
ketoconazole were fit to a 2-compartment model to
determine mean absorption rate (Ka), clearance
(CL/F), central compartment volume (Vc/F), periph-
eral compartment volume (Vp/F), and transport
between central and peripheral compartments (Q)
using NONMEM (Version VI, GloboMax, Ellicot
City, Maryland). The demographics of this popula-
tion are described in the “Participants” section. The
optimal parameter fit was chosen based on visual
goodness of fit and minimization of the standard
error of the estimates. Due to the lack of pharma-
cokinetic sampling points beyond 12 hours postdose
for this data set, good estimates for intersubject vari-
ability could not be obtained. Nevertheless, based
on visual inspection, these rough parameter esti-
mates (see Table III) adequately capture the observed
midazolam pharmacokinetics (Figure 2) and are suf-
ficient for our purposes because the focus of the
simulation model is on ketoconazole and its inhibi-
tory effects rather than on midazolam population
pharmacokinetics.
The midazolam 2-compartment model was then inter­
faced with a model for ketoconazole pharmacokinetics
and inhibition effects previously developed by Chien et
al.4 This model incorporates the effects of changing
ketoconazole concentrations over time in the gut,
plasma, and hepatic enzyme site due to ketoconazole
dosing, absorption, and clearance by compartmental
modeling of ketoconazole pharmacokinetics. The
unbound concentrations of ketoconazole in the
gut and hepatic compartments at any given time
are used to estimate a corresponding midazolam
clearance. Parameter values for the ketoconazole
pharmacokinetic and inhibitory effect model were
the same as those used by Chien et al,4 whereas
 EFFECT OF KETOCONAZOLE DOSING ON MIDAZOLAM

Table I Single-Dose Pharmacokinetics of Midazolam With and Without 400 mg Ketoconazole

Pharmacokinetic Least Squares Geometric
parameter Treatment n Mean (SD) Mean Ratioa (95% CI)

AUC0-∞, ng/mL⋅h 400 mg ketoconazole + 2 mg midazolamday 5 10 265.9 (82.9) 13.96 (11.79, 16.53)
400 mg ketoconazole + 2 mg midazolamday 2 10 250.3 (105.6) 13.14 (11.08, 15.58)
400 mg ketoconazole + 2 mg midazolamday 1 10 195.9 ( 76.7) 10.28 (8.68, 12.19)
2 mg midazolam 11 19.0 (10.6)
Cmax, ng/mL 400 mg ketoconazole + 2 mg midazolamday 5 10 39.1(7.5) 5.42 (4.25, 6.90)
400 mg ketoconazole + 2 mg midazolamday 2 10 38.1 (12.3) 5.29 (4.15, 6.75)
400 mg ketoconazole + 2 mg midazolamday 1 10 36.1 (6.8) 5.01 (3.93, 6.39)
2 mg midazolam 11 7.2 (3.4)
tmax, hb 400 mg ketoconazole + 2 mg midazolamday 5 10 1.00 (0.50, 1.00)
400 mg ketoconazole + 2 mg midazolamday 2 10 0.75 (0.50, 2.00)
400 mg ketoconazole + 2 mg midazolamday 1 10 1.00 (0.50, 2.00)
2 mg midazolam 11 0.50 (0.25, 1.00)
t1/2, hc 400 mg ketoconazole + 2 mg midazolamday 5 10 6.89 (3.56)
400 mg ketoconazole + 2 mg midazolamday 2 10 6.07 (3.23)
400 mg ketoconazole + 2 mg midazolamday 1 10 4.91 (2.11)
2 mg midazolam 11 2.61 (1.07)
CI, confidence interval.
a. Ketoconazole + midazolamday x/midazolam alone.
b. Median (range).
c. Harmonic mean ± jackknife standard deviation.

gender, weight, and age distributions for the subject Pharmacokinetics

population were based on our clinical study
population. Concentration-time profiles for a single
dose of midazolam administered in the absence of
ketoconazole and with 1, 2, and 5 days of ketoconazole
treatment were simulated in Trial Simulator 2.2.1
(Pharsight Corp, Mountainview, California) using
the 2-compartment pharmacokinetic parameters
estimated in NONMEM and the ketoconazole
inhibition model described. Because the midazolam
pharmacokinetic parameters represent a composite
subject based on the population in the clinical study,
only intrasubject variability terms were included in
the ketoconazole-midazolam interaction model.
Intrasubject variability could not be estimated from
the midazolam data and was assumed to be normally
distributed with a CV of 30%.
RESULTS
Participant Demographics
Of the 12 participants enrolled in the study, 3 with-
drew consent for personal reasons prior to complet-
ing all 4 treatment periods. Data from 11 participants
who completed at least 2 treatments were used in the
evaluation of midazolam pharmacokinetics. Of these
11 participants, 9 completed all 4 treatment periods,
1 completed 3 periods, and 1 completed 2 periods.
Effect of different durations of ketoconazole dosing on
the single-dose pharmacokinetics of midazolam. The
effects of 400 mg ketoconazole, administered for 1, 2,
or 5 days, on the single-dose pharmacokinetics of
midazolam are summarized in Table I and Figure 1.
Compared with a single dose of 2 mg midazolam given
alone, administration of 400 mg ketoconazole for 1, 2,
or 5 days resulted in a 10.28-, 13.14-, and 13.96-fold
increase in midazolam AUC0–∞, respectively. The max-
imum concentration (Cmax) of midazolam was increased
5.01-, 5.29-, and 5.42-fold, respectively, when midazo-
lam was administered on day 1, 2, or 5 of once-daily
dosing of 400 mg ketoconazole. Ketoconazole, regard-
less of the dosing duration, did not affect the time to
Cmax (tmax) of midazolam. Coadministration of ketocon-
azole for 1, 2, and 5 days increased the apparent termi-
nal half-life (t1/2) of midazolam to 4.9, 6.1, and 6.9
hours, respectively, compared with 2.6 hours when
midazolam was dosed alone.
Comparative effect of 1, 2, and 5 days of ketocon-
azole dosing on the single-dose pharmacokinetics of
midazolam. The comparative effect that different
durations of ketoconazole dosing had on the single-
dose pharmacokinetics of midazolam is summarized
in Table II. The geometric mean ratio (ketoconazole
+ midazolamday 5/ketoconazole + midazolamday 2) for
midazolam AUC0-∞ was 1.06 with a corresponding

DRUG INTERACTIONS 401

 STOCH ET AL


administered on day 2 or day 5 of once-daily dosing

of 400 mg ketoconazole. However, the comparison of
  midazolam AUC0-∞ after a single dose of midazolam
0LGD]RODPSODVPDFRQFHQWUDWLRQ QJP/

coadministered on day 5 of dosing with ketoconazole






 
Figure 1. Mean midazolam plasma concentration (ng/mL) ver-
sus time following single-dose administration of 2 mg midazolam
with and without 400 mg ketoconazole in the fasted state to
young, healthy male participants.
90% confidence interval of (0.92, 1.23). The geomet-
ric mean ratio (ketoconazole + midazolamday 5/keto-
conazole + midazolamday 1) for midazolam AUC0-∞
was 1.36 with a corresponding 90% confidence
interval of (1.17, 1.57). Based on the prespecified
AUC0-∞ comparability bounds for the 90% confi-
dence interval (0.5, 2.00), there was no meaningful
difference observed for midazolam AUC0-∞ following
a single 2-mg oral dose of midazolam coadminis-
tered on day 1, 2, or 5 of once-daily dosing with 400
mg ketoconazole. There was no difference in mida-
zolam AUC0-∞ when a single dose of midazolam was
 versus a single dose of midazolam coadministered
with a single dose of ketoconazole reached statistical

significance (P > .002). Similarly, a statistically sig-
     
nificant treatment difference in midazolam AUC0-∞
0'=RQO\ Q 
0'=RQ'D\RI.HWR Q  was also observed for a single dose of midazolam
0'=RQ'D\RI.HWR Q 
0'=RQ'D\RI.HWR Q  coadministered on day 2 of ketoconazole dosing ver-
sus a single dose of midazolam coadministered with
a single dose of ketoconazole (P = .007). There were
   
no statistically significant between-treatment differ-
7LPH K ences in midazolam Cmax and tmax, but a statistically
significant difference in midazolam t1/2 was observed
when midazolam was given on day 1 versus day 5 of
once-daily dosing with 400 mg ketoconazole.
Model Validation and Application to
Experimental Results
The 2-compartment pharmacokinetic model for
midazolam described previously was interfaced with
the ketoconazole interaction model and used to
simulate the present experimental study (see Table
III). In general, the model was able to reproduce the
quantitative and qualitative results of the experi-
mental study, where the effects of 5 days of ketocon-
azole administration on the substrate AUC0-∞ and
Cmax may be determined after 1 or 2 days of ketocon-
azole administration. The simulated effects of keto-
conazole administration on single-dose midazolam

Table II Comparative Effect of Different Durations of Ketoconazole Dosing

on the Single-Dose Pharmacokinetics of Midazolam
AUC0-∞ Cmax tmax Apparent
Between- Between- Between- t1/2 Between-
AUC0-∞ GMR Treatment Cmax GMR Treatment Treatment Treatment
Treatment Comparison (90% CI) P Valuea (90% CI) P Valuea P Valuea P Valuea

Ketoconazole + 1.06 (0.92, 1.23) >.200 1.02 (0.83, 1.26) >.200 >.200 .164
midazolamday 5 vs
ketoconazole +
midazolamday 2
Ketoconazole + 1.36 (1.17, 1.57) .002 1.08 (0.88, 1.33) >.200 >.200 .005
midazolamday 5 vs
ketoconazole +
midazolam day 1
Ketoconazole + 1.28 (1.11, 1.47) .007 1.05 (0.86, 1.30) >.200 >.200 .097
midazolamday 2 vs
ketoconazole +
midazolamday 1
CI, confidence interval; GMR, geometric mean ratio.
a. P value < .050 indicates a statistically significant difference between treatment regimens.

402  •  J Clin Pharmacol 2009;49:398-406

 EFFECT OF KETOCONAZOLE DOSING ON MIDAZOLAM

Table III   Estimated Midazolam 2-Compartment

Model Pharmacokinetic Parameters
Standard
Error (%
Parameter Estimate of Estimate)

Ka, h–1 0.797 17

CL/F, L/h 111 16
Vc/F, L 76.7 24
Vp/F, L 47.6 24
Q, L/h 174 13
Proportional residual error 0.076 26
CL/F intersubject variability 0.186 51
Vc/F intersubject variability 0.569 66

pharmacokinetics were in good agreement with

experimentally observed concentration-time profiles
(Figure 2) and fold-changes in AUC0-∞ and Cmax fol-
lowing 1, 2, and 5 days of ketoconazole administra-
tion (Figure 3). The ability of the model to
quantitatively reproduce the experimental results
suggests that the model adequately captures the
Figure 2. Experimental and simulated midazolam plasma con-
mechanisms of ketoconazole inhibition of midazo- centration (ng/mL) versus time following single-dose administra-
lam metabolism and therefore may be a source of tion of 2 mg midazolam with and without 400 mg ketoconazole in
insight into the mechanisms behind the results of the fasted state to young, healthy male participants. The solid
the experimental study. and dotted lines are the median and 5th and 95th percentile
In both our within-subject clinical study and the simulated profiles for n = 1000 runs. The points correspond to
predicted results from the ketoconazole inhibition experimentally observed midazolam concentrations.
model, ketoconazole has exerted its maximal inhibitory
effect by day 2 of ketoconazole dosing. The effect of 5
consecutive days of ketoconazole dosing on systemic
clearance in our simulations is shown in Figure 4,
providing insight into the potential mechanism for the
time course of ketoconazole’s inhibitory activity. In our
simulations, it would appear that clearance has already
reached steady-state levels following a single dose
of ketoconazole administration despite predicted
accumulation of ketoconazole at the site of midazolam
metabolism in the liver throughout the 5 days of
dosing. Moreover, large fluctuations in gut ketoconazole
concentration following each daily dose of ketoconazole
have little effect on the overall clearance of midazolam.
Based on the ketoconazole interaction model, the Figure 3. Midazolam AUC0-∞ and Cmax ratios after 1, 2, and 5
concentration of ketoconazole in the gut after each days of once-daily ketoconazole administration compared to a
dose of ketoconazole increases the gut bioavailability single dose of midazolam only. Experimental results from the cur-
rent study are plotted as individual points representing geometric
(Fgut) of midazolam nearly 2-fold by inhibition of mean ratios (GMRs) and 95% confidence intervals (CIs) for each
CYP3A activity in the gut. However, given the rapid ketoconazole dosing regimen, whereas simulated results are plot-
absorption of ketoconazole, there is no accumulation ted as solid (median), dotted (25th and 75th percentile), and
in the gut, and the fluctuations in Fgut due to daily dashed (5th and 95th percentile) lines.

DRUG INTERACTIONS 403


50
Median
5th and 95th
Midazolam Plasma Clearance (L/hr)
Percentiles
40
30
20
10
0
0 20
Figure 4. Simulated midazolam plasma clearance after 5 days of
ketoconazole administration (inset: t = 0-24 hours). The solid line
represents the median (n = 1000) predicted plasma clearance,
whereas the dotted lines represent the 5th and 95th percentiles.
ketoconazole administration are nearly identical
following 1, 2, 5, or more days of dosing. Therefore,
administering additional doses of ketoconazole
through days 2 and 5 will likely have minimal additional
effect on the inhibitory potential of ketoconazole
because maximum changes in substrate bioavailability
and clearance are achieved by day 1. This is consistent
with what was observed in our clinical study.
Safety and Tolerability
Administration of ketoconazole and midazolam,
either separately or together, was well tolerated by
the study participants. Of the 11 participants
included in the pharmacokinetic analysis, none
reported any side effects.
DISCUSSION
Given the prominent role of CYP3A in the metabo-
lism of drugs, it is important to identify early in
development whether new chemical entities will be
affected by this enzyme system and produce clini-
cally relevant drug interactions. Because the expres-
sion and function of CYP3A enzymes are highly
variable, large intersubject variability in pharmacoki-
netic parameters can result. This could pose chal-
lenges for a drug with a narrow therapeutic index,
such as midazolam. Conducting DDI studies early in
development and exercising confident decision mak-
ing as early as possible cannot be overemphasized.
The objective of this study was to determine whether
performing shorter DDI studies may be used to reliably
404  •  J Clin Pharmacol 2009;49:398-406
STOCH ET AL
50 predict the data obtained from more “traditional,”
Inset: t = 0 – 24hours
lengthier interaction studies. The goal was to evaluate
40
Midazolam Plasma
whether a strong CYP3A inhibitor, ketoconazole,
Clearance (L/hr)
30
would affect the pharmacokinetics of a sensitive
20 substrate, midazolam, to a similar extent when
10 midazolam was coadministered with ketoconazole
on either day 1 or 2 when compared to probing on
0
0 5 10 15 20 day 5 to assess the magnitude of key drug-drug
Time (hr)
interactions. Given the importance of evaluating drug-
drug interaction potential in early development to
inform critical decision making, this study focused on
the timing of substrate administration relative to
40 60 80 100 120 140 160
interacting drug (single dose and multiple dose).
Time (hr) Midazolam, a short-acting benzodiazepine that is
selectively metabolized by CYP3A and widely used
as a phenotyping probe for CYP3A activity, was
administered at a dose of 2 mg orally to investigate the
maximal inhibitory effect in vivo.3,6 Ketoconazole, a
strong CYP3A inhibitor, was used at a dose of 400 mg
daily as an in vivo probe to evaluate the magnitude of
CYP3A-mediated drug interactions with midazolam on
days 1, 2, and 5. Midazolam pharmacokinetic parameters
have been shown to change dramatically during the
administration of ketoconazole with increases in
AUC0-∞ of ~16-fold and Cmax of ~5-fold7-10 and may be
accompanied by an almost 5-fold prolongation of the
midazolam elimination half-life with ketoconazole.9
Consistent with data reported in the literature,
probing with midazolam on day 5 of multiple-dose
ketoconazole administration resulted in a 13.96- and
5.4-fold increase in midazolam AUC0-∞ and Cmax,
respectively. When comparing these findings to the
magnitude of response on day 2 of coadministration of
ketoconazole with midazolam, the increase in AUC0-∞
and Cmax was indistinguishable at 13.14- and 5.3-fold,
respectively. Day 2 and day 5 were com­parable based
on the prespecified bounds (0.5 to 2.00) and would
also have met the more stringent bioequivalence (0.8,
1.25) bounds for AUC0-∞ and closely for Cmax had they
been employed. Similarly, if one coadministered
ketoconazole 400 mg and midazolam 2 mg on day 1, a
10.28- and 5.0-fold increase in midazolam AUC0-∞ and
Cmax, respectively, would be observed. Although the
AUC0-∞ data on day 1 do not exhibit the full degree of
inhibition as that observed on days 2 and 5, the
magnitude of effect seen on day 1 is likely clinically
relevant. The alteration in midazolam pharmaco­
kinetics seen with coadministration on day 1 is in
keeping with data reported by others following
pretreatment with ketoconazole for 4 days, which
resulted in a 10-fold increase in AUC0-∞.7 These data
support conducting shorter drug-drug interaction
studies using strong CYP3A inhibitors without
compromising data quality.
 EFFECT OF KETOCONAZOLE DOSING ON MIDAZOLAM
The notable strengths of this study include the
within-subject treatment comparisons and low
variability estimates resulting from the randomized
crossover design. Potential limitations of our study are
discussed below. This study lacks the use of midazolam
pharmacodynamic testing and 1′-hydroxymidazolam
analysis. The 1-hydroxy metabolite accounts for 50%
to 70% of metabolism of midazolam and is reported to
be pharmacologically active.11,12 Various authors have
investigated the ratio of 1′-hydroxymidazolam to
midazolam in plasma as an index for CYP3A4 activity
but with inconsistent results13,14 because the hydroxyl-
metabolites of midazolam undergo extensive glucuro­
nidation.15 Ketoconazole administration was given in
the fasted state on days when it was coadministered
with midazolam, but on all other study days,
ketoconazole was given without regard to food.
Although the potential impact of food on ketoconazole
pharmacokinetics was not assessed in this study, a
previous report concluded that only tmax is significantly
affected by food, and AUC, Cmax, and t1/2
pharmacokinetics are not altered to a statistically
significant extent.5 One should also take into
consideration the relative small, homogeneous
population restricted to white men employed in the
present study (n = 10 or 11), particularly when
compared with those evaluated in model-based
simulation work (n = 1000); therefore, it is possible
that both reported exposure (AUC0-∞) and Cmax observed
in an experimental study may not necessarily reflect
the true population mean. As a consequence, some
discrepancy may exist between experimental and
simulated data, and a direct comparison between the
two may not be possible.
Simulation models to predict the effect of
ketoconazole inhibition of CYP3A on midazolam
pharmacokinetics4,16-19 have the potential to be a
complementary tool to the clinical study described
here. In the ketoconazole interaction model reported
by Chien et al,4 biological details such as saturable
efflux from the hepatic enzyme site, gut wall
permeability, and metabolism in the gut were
considered. Based on this model, the inhibitory activity
of ketoconazole is mediated by the ketoconazole
concentration in the gastrointestinal and hepatic
compartments, which increases the oral bioavailability
and decreases the hepatic clearance of midazolam.
The observed agreement between our simulated and
experimental data suggests that this reliably captures
the effects of ketoconazole. The compartment-based
ketoconazole inhibition model also provides a view
into the mechanisms in the gut and liver that may
contribute to the clinically observed result. This model
DRUG INTERACTIONS
suggests that despite continued accumulation of
ketoconazole in the liver over 5 days of dosing, the
maximum change in midazolam clearance caused by
ketoconazole is achieved within 1 day of ketoconazole
dosing. Furthermore, daily fluctuations in ketoconazole
level in the gut give rise to large changes in the
bioavailability of midazolam over the course of 24
hours but are consistent between consecutive days
of dosing, so that the steady-state inhibitory effects of
ketoconazole are achieved after approximately 1 day
of dosing.
However, the enzyme site with the saturable
efflux model used here predicts that significant
concentrations of ketoconazole remain in the enzyme
site for a number of days following administration of
the last dose of ketoconazole. The present clinical
study results suggest that the accumulation of
ketoconazole in the enzyme site does not play a role
in the observed midazolam pharmacokinetics as
there were minimal carryover effects. Thus, although
the saturable efflux mechanism may be sufficient to
predict the inhibitory effects of ketoconazole, it may
not accurately reflect the true accumulation of
ketoconazole in the hepatic enzyme site in vivo.
Given the good agreement between simulated results
from the ketoconazole inhibition model and the clinical
study, the model may also serve as a predictive tool to
complement the abbreviated clinical study proposed
here. With clinical concentration-time profiles for a
single dose of substrate without ketoconazole and after
1 day of ketoconazole administration, substrate
pharmacokinetic parameters may be estimated and the
predictive ability of the model validated. The
concentration-time profile of the substrate after 2, 5, or
more days of ketoconazole dosing may then be predicted.
Moreover, with estimates for the Ki for ketoconazole
inhibition (fraction unbound for ketoconazole and the
substrate) and the Vmax and Km for CYP3A metabolism
of the substrate (the fraction of metabolism attributed to
CYP3A [fm] and the bioavailability in the gut), the
ketoconazole-inhibition model potentially may be
applied to substrates other than midazolam.
Conclusion
Although extensive CYP3A inhibition with <24 hours
of ketoconazole exposure is well recognized,20 this
study demonstrated that earlier administration of a
sensitive substrate (midazolam) relative to a potent
interacting drug (ketoconazole) may be used to reli-
ably predict the magnitude of drug-drug interaction
potential, which may be very enabling in early drug
development. The extent of the ­interaction observed
405
 STOCH ET AL
after earlier administration (day 1 or 2) was consis-
tent with that reported in the literature of key drug-
drug interaction studies of longer (5 days) duration.
It was remarkable how predictive day 2 data were of
day 5; in addition, data obtained on day 1 were also
a close approximate to those obtained on day 5. The
decision whether to implement probing on day 1
versus day 2 may rest on the degree of confidence
that would be required for decision making. Based
on this study, there is no difference between days 2
and 5, and day 1 would be a close approximate.
Although midazolam was selected as the sensitive
CYP3A probe substrate and ketoconazole as the
inhibitor, one could speculate that a similar relation-
ship may extend to other less sensitive CYP3A sub-
strates but not necessarily to less potent inhibitors.
The authors acknowledge Rita Chiou and Jin Zhang for coor-
dination of midazolam sample analysis. The authors wish to
thank Susi Li for her constructive input and review of this manu-
script. The authors thank the participants and study personnel at
Promedica CRC, Inc, who made this study possible.
Financial disclosure: This study was funded by Merck & Co,
Inc. Authors who are employees of Merck may potentially own
stock and/or hold stock options in the company. Authors who are
not employees of Merck have received grant support, consultant
fees, and/or lecture honoria.
REFERENCES
1. Kashuba A, Bertino J. Mechanisms of drug interactions I. In:
Pisceteill SC, Rodvold KA, eds. Drug Interactions in Infectious
Diseases. 2nd ed. Totowa, NJ: Humana; 2005:13-39.
2. Honig P, Wortham D, Zamani K, et al. Terfenadine-ketoconazole
interaction, pharmacokinetic and electrocardiographic conse-
quences. JAMA. 1993;269:1513-1518.
3. Lin Y, Lockwood G, Graham M, et al. In-vivo phenotyping for
CYP3A by a single-point determination of midazolam plasma
concentration. Pharmacogenetics. 2001;11:781-791.
4. Chien J, Lucksiri A, Ernest C, et al. Stochastic prediction of
CYP3A-mediated inhibition of midazolam clearance by ketocon-
azole. Drug Metab Dispos. 2006;34:1208-1219.
5. Daneshmend TK, Warnock DW, Ene MD, et al. Influence of food
on the pharmacokinetics of ketoconazole. Antimicrob Agents
Chemother. 1984;25:1-3.
6. Bjornsson T, Callaghan J, Einolf H, et al. The conduct of in vitro
and in vivo drug-drug interaction studies: a Pharmaceutical
Research and Manufacturers of America (PhRMA) perspective.
Drug Metab Dispos. 2003;31:815-832.
406  •  J Clin Pharmacol 2009;49:398-406
7. Olkkola K, Backman, J, Neuvonen M, et al. Midazolam should be
avoided in patients receiving the systemic antimycotics ketocon-
azole and itraconazole. Clin Pharmacol Ther. 1994;55:481-485.
8. Gorski J, Jones D, Haehner-Daniels B, et al. The contribution of
intestinal and hepatic CYP3A4 to the interaction between midazo-
lam and clarithromycin. Clin Pharmacol Ther. 1998;64:133-143.
9. Tsunoda S, Vlez R, Moltke L, et al. Differentiation of intestinal
and hepatic cytochrome P450 3A activity with use of midazolam
as an in vivo probe: effect of ketoconazole. Clin Pharmacol Ther.
1999;66:461-471.
10. McCrea J, Prueksaritanont T, Gertz B, et al. Concurrent admin-
istration of the erythromycin breath test (EBT) and oral midazo-
lam as in vivo probes for CYP3A activity. J Clin Pharmacol.
1999;39:1212-1220.
11. Zieler WH, Schalch E, Leishman B, et al. Comparison of the
effects of intravenously administered midazolam, triazolam and
their hydroxy metabolites. Br J Clin Pharmacol. 1983;19:271-278.
12. Mandema JW, Tuk B, van Steneninck AL, et al. Pharmacokinetic-
pharmacodynamic modeling of the central nervous system effects
of midazolam and its main metabolite alpha-hydroxymidazolam
in healthy volunteers. Clin Pharmacol Ther. 1992;51:715-728.
13. Lee LS, Bertino JS Jr, Nafziger AN. Limited sampling models
for oral midazolam: midazolam plasma concentrations, not the
ratio of 1-hydroxymidazolam to midazolam plasma concentra-
tions, accurately predicts AUC as a biomarker of CYP3A activity.
J Clin Pharmacol. 2006;46:229-234.
14. Rogers JF, Nafziger AN, Kashuba AD, et al. Single plasma
concentrations of 1′-hydroxymidazolam or the ratio of 1′-
hydroxymidazolam:midazolam do not predict midazolam clear-
ance in healthy subjects. J Clin Pharmacol. 2002;42:1079-1082.
15. Link B, Haschke M, Grignaschi N, et al. Pharmacokinetics of
intravenous and oral midazolam in plasma and saliva in humans:
usefulness of saliva as matrix for CYP3A phenotyping. Br J Clin
Pharmacol. 2008;66:473-484.
16. Zhou SF, Chan SY, Goh BC, et al. Mechanism-based inhibition
of cytochrome P450 3A4 by therapeutic drugs. Clin Pharmacokinet.
2005;44:279-304.
17. Brown HS, Galetin A, Hallifax D, et al. Prediction of in vivo
drug-drug interactions from in vitro data: factors affecting proto-
typic drug-drug interactions involving CYP2C9, CYP2D6 and
CYP3A4. Clin Pharmacokinet. 2006;45:1035-1050.
18. Shou MG. Prediction of pharmacokinetics and drug-drug
interactions from in vitro metabolism data. Curr Opin Drug
Discov Dev. 2005;8:66-77.
19. Yamano K, Yamamoto K, Katashima M, et al. Prediction of
midazolam-CYP3A inhibitors interaction in the human liver from
in vivo/in vitro absorption, distribution, and metabolism data.
Drug Metab Dispos. 2001;29:443-452.
20. von Moltke L, Greenblatt D, Harmatz J, et al. Triazolam
biotransformation by human microsomes in vitro: effects of
metabolic inhibitors and clinical confirmation of a predicted
interaction with ketoconazole. J Pharmacol Exp Ther. 1996;
276:370-379.