Prediction of Irinotecan Pharmacokinetics by Use of
Cytochrome P450 3A4 Phenotyping Probes
Ron H. J. Mathijssen, Floris A. de Jong, Ron H. N. van Schaik, Erin R. Lepper,
Lena E. Friberg, Trinet Rietveld, Peter de Bruijn, Wilfried J. Graveland,
William D. Figg, Jaap Verweij, Alex Sparreboom
Background: Irinotecan is a topoisomerase I inhibitor that
has been approved for use as a first- and second-line treat-
ment for colorectal cancer. The response to irinotecan is
variable, possibly because of interindividual variation in the
expression of the enzymes that metabolize irinotecan, includ-
ing cytochrome P450 3A4 (CYP3A4) and uridine diphos-
phate glucuronosyltransferase 1A1 (UGT1A1). We prospec-
tively explored the relationships between CYP3A phenotype,
as assessed by erythromycin metabolism and midazolam
clearance, and the metabolism of irinotecan and its active
metabolite SN-38. Methods: Of the 30 white cancer patients,
27 received at least two treatments with irinotecan adminis-
tered as one 90-minute infusion (dose, 600 mg) with 3 weeks
between treatments, and three received only one treatment.
Before the first and second treatments, patients underwent
an erythromycin breath test and a midazolam clearance test
as phenotyping probes for CYP3A4. Erythromycin metabo-
lism was assessed as the area under the curve for the flux of
radioactivity in exhaled CO2within 40 minutes after admin-
istration of [N-methyl-14C]erythromycin. Midazolam and
irinotecan were measured by high-performance liquid chro-
matography. Genomic DNA was isolated from blood and
screened for genetic variants in CYP3A4 and UGT1A1. All
statistical tests were two-sided. Results: CYP3A4 activity
varied sevenfold (range ? 0.223%–1.53% of dose) among
patients, whereas midazolam clearance varied fourfold
(range ? 262–1012 mL/min), although intraindividual vari-
ation was small. Erythromycin metabolism was not statisti-
cally significantly associated with irinotecan clearance (P ?
.090), whereas midazolam clearance was highly correlated
with irinotecan clearance (r ? .745, P<.001). In addition, the
presence of a UGT1A1 variant with a (TA)7repeat in the
promoter (UGT1A1*28) was associated with increased ex-
posure to SN-38 (435 ng · h/mL, 95% confidence interval
[CI] ? 339 to 531 ng · h/mL in patients who are homozygous
for wild-type UGT1A1; 631 ng · h/mL, 95% CI ? 499 to 762
ng · h/mL in heterozygous patients; and 1343 ng · h/mL, 95%
CI ? 0 to 4181 ng · h/mL in patients who are homozygous for
UGT1A1*28) (P ? .006). Conclusion: CYP3A4 phenotype, as
assessed by midazolam clearance, is statistically significantly
associated with irinotecan pharmacokinetics. Evaluation of
midazolam clearance combined with UGT1A1*28 genotyp-
ing may assist with optimization of irinotecan chemotherapy.
[J Natl Cancer Inst 2004;96:1585–92]
The topoisomerase I inhibitor irinotecan has been approved in
the United States as a second-line treatment for advanced colo-
rectal cancer that is refractory to fluorouracil and as first-line
treatment in combination with fluorouracil–leucovorin for met-
astatic colorectal cancer (1). Apart from antitumor activity in
colorectal cancer, single-agent irinotecan is also moderately
active in several other solid malignancies, including breast can-
cer (2), relapsed or refractory non-Hodgkin lymphoma (3), and
lung cancer (4). Irinotecan is unique among camptothecin ana-
logues in that it must first be converted by a carboxylesterase-
converting enzyme to the active metabolite 7-ethyl-10-
hydroxycamptothecin (SN-38) (5). Despite much research into
the complex pharmacokinetic profile and pharmacodynamic ef-
fects of irinotecan, unpredictable and severe side effects are still
commonly observed (6,7).
Irinotecan pharmacokinetic variability between individuals
is large. Pathways that eliminate irinotecan contain the cyto-
chrome P450 3A subfamily members CYP3A4 and CYP3A5
(collectively referred to as CYP3A); the uridine diphosphate
glucuronosyltransferase 1A subfamily members UGT1A1,
UGT1A3, UGT1A7, and UGT1A9 (collectively referred to as
UGT1A); and drug transporting proteins, including ABCB1
(P-glycoprotein) (5) (Fig. 1). Recent investigations have also
indicated that genetic polymorphisms (8), herbal supplements
(9), and concomitantly administered allopathic drugs (10) can
alter the activity and/or expression levels of these proteins
and change the rate of irinotecan elimination. Thus far, only
limited attempts have been made to incorporate this knowl-
edge into clinical practice. Because dosing strategies that are
based on body surface area do not reduce interindividual
variability in irinotecan pharmacokinetics (11–13), other
measures to predict the pharmacologic profile of irinotecan in
individual patients are needed. One possibility is to assess the
phenotype of CYP3A because CYP3A is involved in the
metabolism of about half of all prescribed drugs (14) and
plays a principal role in the metabolism of irinotecan (5). In
vivo probe drugs such as cortisol, dextromethorphan, eryth-
romycin, and midazolam are widely used for evaluating
CYP3A activity in humans (15–17), and such probes accu-
rately predict the activity of CYP3A (18–20) and the clear-
ance of docetaxel, another anticancer drug that is a CYP3A
substrate (21–24). Consequently, we prospectively explored
the relationship between CYP3A phenotype, as assessed with
Affiliations of authors: Departments of Medical Oncology (RHJM, FADJ,
PDB, JV), Clinical Chemistry (RHNVS), Internal Medicine (TR), and Biosta-
tistics (WJG), Erasmus University Medical Center, Rotterdam, The Netherlands;
Department of Pharmaceutical Biosciences, Uppsala University, Uppsala, Swe-
den (LEF); Clinical Pharmacology Research Core, National Cancer Institute,
Bethesda, MD (ERL, WDF, AS).
Correspondence to: Alex Sparreboom, PhD, Clinical Pharmacology Research
Core, National Cancer Institute, 9000 Rockville Pike, Bldg. 10, Rm. 5A01,
Bethesda, MD 20892 (e-mail firstname.lastname@example.org).
See “Notes” following “References.”
Journal of the National Cancer Institute, Vol. 96, No. 21, © Oxford University
Press 2004, all rights reserved.
Journal of the National Cancer Institute, Vol. 96, No. 21, November 3, 2004ARTICLES 1585
by guest on June 4, 2013
the probe drugs erythromycin and midazolam, and the me-
tabolism of irinotecan and its active metabolite SN-38 by
white patients with cancer.
PATIENTS AND METHODS
Treatment of Patients
Eligible patients had a solid tumor for which irinotecan was
considered the treatment of choice or for which no standard
treatment option was available. Patients were treated with irino-
tecan (Aventis, Hoevelaken, The Netherlands) once every 3
weeks as a 90-minute intravenous infusion at a fixed dose of 600
mg. If excessive toxicity (e.g., diarrhea and neutropenia) was
detected, the next course was postponed for a week and/or the
dose was reduced by 25%. Thirty cancer patients received iri-
notecan chemotherapy; all but three received two treatments.
Eligibility and exclusion criteria (including acceptable liver
functions and a World Health Organization performance score
of ?1), premedication, and protocols for treatment of drug-
induced side effects (e.g., diarrhea with the use of loperamide
and/or antibiotics) were identical to those used earlier (25). No
patient was allowed to use drugs, food supplements, and/or
herbal preparations that were known to interfere with the func-
tion or expression of proteins involved in irinotecan disposition.
The clinical protocol was approved by the Erasmus Medical
Center Ethics Board, and all patients provided written informed
consent before study entry.
Erythromycin Breath Test
Five microcuries (?Ci; 0.07 mg or 0.1 ?mol) of [N-methyl-
14C]erythromycin (American Radiolabeled Chemicals, St.
Louis, MO) in 4 mL of a solution of 2.5% glucose and 0.45%
sodium chloride (final erythromycin concentration, 17.5 ?g/mL
or 1.19 ?Ci/g) was injected intravenously through an infusion
set in less than 30 seconds, 4–8 days before the first treatment
with irinotecan. In half of the patients, this test was repeated
before the second course; in the other half, a second midazolam
clearance test was given. The metabolism of erythromycin in the
liver involves the CYP3A4-catalyzed cleavage of the N-methyl
group from erythromycin and, after a series of non–rate-limiting
steps, the formation of CO2from the cleaved formaldehyde (20).
results in the production of14CO2(26). Patients exhaled14CO2
through a drinking straw into a solution of 2.00 mL of 1 M
hyamine hydroxide in methanol (Packard Instrument, Meriden,
CT) and 2 mL of thymolphthalein (60 mg/L in ethanol).14CO2
and CO2in the breath sample were absorbed by the hyamine
hydroxide, and the indicator thymolphthalein (60 mg/L in eth-
anol) changed color from blue to clear when the hyamine hy-
droxide was saturated with CO2. Aspiration of liquid as the
patient breathed was prevented by safety valves. After combin-
ing the 4-mL breath sample solution with 5 mL of Insta-Gel Plus
liquid scintillation fluid (Packard), we determined the amount of
14CO2in a breath sample by liquid scintillation counting on a
Packard TRI-CARB Liquid Scintillation Analyzer 1900 TR, as
described elsewhere (27). The amount of radioactivity was ex-
pressed as disintegrations per minute (dpm). Eight breath sam-
ples were taken over a 40-minute period (i.e., immediately
before treatment and 5, 10, 15, 20, 25, 30, and 40 minutes after
infusion ended). The flux of radioactivity (14C) in exhaled CO2
at each time tx (CERtx), expressed as a percentage of the eryth-
romycin dose of radioactivity per minute, was calculated as
14C-labeled N-methyl moieties in erythromycin
2.222 ? 106dpm/?Ci?
1.19 ?Ci/g ? ?weight? 5 mmol CO2/min/m2? BSA,
2.00 ml ? 0.972 M CO2
where the first two terms, (dpmtx? dpmt0)/2.222 ? 106dpm/
?Ci and 1/2.00 ml ? 0.972 M CO2, refer to the measured dose
in microcuries (first term) per captured millimole (second term)
of CO2at time tx, and in the third term 1.19 ?Cig ? ?weight is
the administered dose in microcuries, and 100 corrects for per-
centage. In the equation, dpmtxis the amount of radioactivity
(expressed in disintegrations per minute) for each breath sample
obtained, and weight is the mass (expressed in grams) of solu-
tion injected. The product of the first three terms is the percent-
age of administered dose per exhaled millimole of CO2at time
x. Multiplying this product by the fourth term (5 mmol CO2/
min/m2? BSA) corrects for CO2output of individuals on the
basis of 5 mmol of CO2exhaled per minute per square meter of
body surface area (BSA), to give the flux of exhaled14C ex-
pressed as percentage of the dose per minute (28). The most
commonly used parameter from the erythromycin breath test is
CER0–40, the area under the curve for the flux of radioactivity
(14C) in exhaled CO2from 0 to 40 minutes (19), and it served as
the primary parameter to predict total body clearance of irinotecan.
Fig. 1. Schematic diagram of a human liver showing the main drug-metabolizing
enzymes and ATP-binding cassette drug transporters involved in elimination
routes of irinotecan and its metabolites. CPT-11 ? irinotecan; SN-38 ? 7-ethyl-
10-hydroxycamptothecin; SN-38G ? SN-38 glucuronide; APC ? 7-ethyl-10-
[4-N-(5-aminopentanoic acid)-1-piperidino]-carbonyloxy-camptothecin; NPC ?
7-ethyl-10-[4-amino-1-piperidino]-carbonyloxycamptothecin; hCE1/2 ? human
carboxylesterase isoforms 1 and 2; CYP3A4/5 ? cytochrome P450 isoforms
3A4 and 3A5; UGT1A1/7/9 ? UDP glucuronosyltransferase isoforms 1A1,
1A7, and 1A9; ABCB1 ? P-glycoprotein; ABCC2 ? multidrug resistance–
associated protein 2 (cMOAT); ABCG2 ? breast cancer–resistance protein.
Irinotecan is eliminated in all those compartments (blood, hepatocyte, and bile)
where the mentioned metabolites (black names) are formed by use of enzymes
and excreted by ABC transporters (white names).
1586 ARTICLESJournal of the National Cancer Institute, Vol. 96, No. 21, November 3, 2004
by guest on June 4, 2013
Midazolam Clearance Test
Midazolam (0.025 mg/kg of body weight, Roche Laborato-
ries, Mijdrecht, The Netherlands) was injected intravenously
within a 30-second period, 4–8 days before the first treatment
with irinotecan. In patients not undergoing a second erythromy-
cin breath test, the midazolam clearance test was repeated before
the second treatment with irinotecan. Blood samples of 7 mL
were collected immediately before infusion and 5 and 30 min-
utes and 1, 2, 4, 5, and 6 hours after the infusion ended. These
blood samples were centrifuged immediately after collection for
10 minutes at 2000g (4 °C), and plasma supernatants were stored
at ?20 °C on the day of collection and then at ?80 °C until the
day of analysis. After the addition of 25 ?L of the internal
standard (a solution of lorazepam at 4 ?g/mL of methanol), 600
?L of plasma was extracted in one step with ethyl acetate.
Midazolam and lorazepam were separated by high-performance
liquid chromatography on a column (150 ? 4.6 mm, internal
diameter) with a matrix of 5-?m Zorbax Eclipse XDB-C8and
with a mobile phase composed of methanol and 10 mM aqueous
ammonium acetate (60:40, vol/vol). Column effluents were
analyzed by mass spectrometry with an atmospheric pressure
chemical ionization interface (29). Calibration curves for mida-
zolam were linear from 1.00 to 200 ng/mL. The accuracy and
precision of measurements ranged from 92.8% to 112% and
from 0.056% to 13.4%, respectively, for four concentrations of
quality control samples analyzed in triplicate on eight separate
Blood samples of 5 mL were collected in heparin-containing
tubes during the first and second irinotecan treatments at the
following times: immediately before infusion; 30 minutes after
the start of infusion; immediately before the end of infusion; and
10, 20, and 30 minutes and 1, 1.5, 2, 3.5, 5, 6.5, 23, 31, 47, and
55 hours after the end of infusion. In addition, patients were
asked to provide a blood sample during their weekly outpatient
visit on days 7, 14, and 20 after infusion. Blood samples were
handled as described previously (30), and concentrations of
irinotecan, SN-38, and SN-38 glucuronide (SN-38G) were de-
termined by reversed-phase high-performance liquid chromatog-
raphy with fluorescence detection as described previously
DNA was isolated from 0.2 mL of whole blood or plasma
with a Total Nucleic Acid Extraction kit on a MagNA Pure
LC (Roche Molecular Biochemicals, Mannheim, Germany)
and amplified by polymerase chain reaction. Restriction
fragment length polymorphism analysis was used to identify
specific variations in the genes ABCB1 (i.e., ABCB1
1236C3T [ABCB1*8], ABCB1 2677G3A/T [ABCB1*7],
and ABCB1 3435C3T [ABCB1*6]), CYP3A4 (CYP3A4*1B,
CYP3A4*2, CYP3A4*3, CYP3A4*17, and CYP3A4*18), and
CYP3A5 (CYP3A5*3 and CYP3A5*6) (33,34). The number of
TA repeats in the promoter of the UGT1A1 gene was determined
by sizing of products from the polymerase chain reaction ob-
tained with the UGT1A1-specific primers 5?-6-carboxyfluorescein-
AAGTGAACTCCCTGCTACCT-3? and 5?-AAAGTGAACT
CCCTGCTACC-3?, followed by fragment analysis carried out
with the automated capillary electrophoresis instrument ABI310
(Applied BioSystems, Foster City, CA).
For pharmacokinetic modeling, we used a previously devel-
oped population model (35) to estimate individual pharmacoki-
netic parameters of irinotecan, SN-38, and SN-38G, which in-
cluded the accumulated area under the plasma concentration
versus time curve (AUC) and clearance. The AUC was deter-
mined for irinotecan and its metabolites in all patients from 0 to
100 hours after start of infusion for a dose corrected to 600 mg.
In this analysis, interoccasion variability in the parameters was
also considered. The analysis was performed with NONMEM
version VI (S. L. Beal and L. B. Sheiner, San Francisco, CA).
The relative extent of conversion (irinotecan to SN-38) was
calculated as the ratio of the AUC of SN-38 and the AUC of
irinotecan, and the relative extent of glucuronidation (SN-38 to
SN-38G) was calculated as the ratio of the AUC of SN-38G and
the AUC of SN-38. WinNonlin version 4.0 (Pharsight, Mountain
View, CA) was used to calculate pharmacokinetics parameters
for the erythromycin breath test and midazolam clearance test
data. Uniform weighted percentages of the administered dose
per minute (CERtx) as input for a one-compartment model
yielded the following parameters for the erythromycin breath
test: the maximal CER (CERmax), tmax, and its reciprocal 1/tmax.
The area under the CER curve from 0 to 40 minutes (CER0–40)
was calculated by use of noncompartmental analysis, and the
percentage of the administered dose per minute in the sample
obtained at the 20-minute point was noted. The clearance was
calculated as the ratio of dose and AUC extrapolated to infinity
obtained from a linear one-compartment model. For the mida-
zolam clearance test, the clearance and the midazolam concen-
tration obtained at the 4-hour sampling point (t4) were evaluated
as potential predictors of irinotecan pharmacokinetics (36). The
midazolam concentration at the 4-hour sampling point is a
commonly used parameter that has been extensively evaluated in
limited sampling schemes to determine the reproducibility of
estimating midazolam AUC by use of only one time point (37).
Complete blood cell counts and blood chemistry data were
obtained for each patient before study entry and before each
chemotherapy course, and these tests were repeated once a week
during the patients’ outpatient visits. If severe hematologic tox-
icity was detected, blood cell counts were measured daily or as
clinically indicated. Diarrhea was scored by use of the National
Cancer Institute Common Toxicity Criteria (NCI-CTC) version
2.0 (available at: http://ctep.info.nih.gov/reporting/ctc.html [last
accessed September 28, 2004]).
Pharmacokinetic data are presented as mean values and 95%
confidence intervals (CIs), unless stated otherwise. Before ge-
notype and phenotype analysis, AUC values were logarithmi-
cally transformed. Associations between irinotecan pharmaco-
kinetics obtained during the first irinotecan treatment and the
CYP3A phenotype as determined by the erythromycin breath
test or the midazolam clearance test were evaluated by use of
Pearson’s correlation coefficient. The influence of the various
genetic variants on irinotecan pharmacokinetics and pharmaco-
Journal of the National Cancer Institute, Vol. 96, No. 21, November 3, 2004 ARTICLES 1587
by guest on June 4, 2013
dynamics during the first irinotecan treatment was assessed by
use of a Kruskal–Wallis one-way analysis of variance or a
nonparametric trend analysis. Although this study was mainly
exploratory in intent, a Hochberg adjustment was used to eval-
uate the statistical significance of the multiple comparisons (38).
All statistical tests were two-sided. With both the Kruskal–
Wallis and the trend analysis tests, P values of less than .01 were
regarded as statistically significant, and those less than .05 were
considered a non–statistically significant trend (i.e., a P?.05 and
?.01 corresponds with some evidence of difference, but the
evidence is not strong enough to declare it to be statistically
significant). These levels were chosen to reduce the risk of
finding purely coincidental associations in view of the number of
parameters analyzed. Statistical calculations were performed
with SPSS version 10.1 (Paris, France) or Stata version 8.2
(Stata, College Station, TX).
Patients and Treatment
A total of 30 eligible adult white patients with cancer (16
males and 14 females) with a median age of 55 years (range ?
38–73 years) were recruited to this study between January 1,
2002, and July 31, 2003 (Table 1). All but three of them received
at least two courses of chemotherapy. The most frequent primary
tumor types were lung cancer (n ? 10) and colorectal cancer
(n ? 12). All patients received the planned fixed irinotecan dose
of 600 mg during the first irinotecan treatment, but four of them
received a 25% dose reduction during the second irinotecan
treatment because of severe side effects experienced with the
first administration, as required by the protocol.
Plasma concentrations of irinotecan, SN-38, and SN-38G as
a function of time were accurately predicted by a modified
version of a previous population model (35), as determined by
goodness-of-fit plots (data not shown). The typical irinotecan
clearance was 31.8 L/h (95% CI ? 28.4 L/h to 35.1 L/h), the
mean relative extent of irinotecan to SN-38 conversion was
0.0263 (95% CI ? 0.0218 to 0.0307), and the mean relative
extent of SN-38 to SN-38G glucuronidation was 6.95 (95% CI
? 5.23 to 8.66), consistent with earlier data (Table 2) (35). The
interoccasion variability in irinotecan clearance was estimated to
be 11.0%. The relative intrapatient variation in parameter esti-
mates was minimal, with mean values for the ratio of the AUC
in the second irinotecan treatment to the AUC in the first
irinotecan treatment of 0.90, 0.87, and 0.86 for irinotecan, SN-
38, and SN-38G, respectively.
Association of CYP3A Phenotype With Irinotecan
CYP3A phenotypic parameters—erythromycin metabolism
(erythromycin breath test parameters of the percentage of the
administered dose in the sample obtained at the 20-minute point
[r ? .846 and P?.001] and the area under the CER curve from
0 to 40 minutes [r ? .840 and P?.001]) and midazolam metab-
olism (midazolam clearance [r ? .661 and P ? .010])—during
the first and second irinotecan treatments were highly correlated,
suggesting limited intraindividual variability in CYP3A expres-
sion and function within the study period. CYP3A activity as
determined from the erythromycin breath test data from the first
irinotecan treatment varied about sevenfold (area under the CER
curve from 0 to 40 minutes range ? 0.223%–1.53% of dose)
among the patients and as determined from the midazolam
clearance test data varied about fourfold (midazolam clearance
range ? 262–1012 mL/min).
Conventional parameters for the erythromycin breath test
(including the percentage of the administered dose in the sample
obtained at the 20-minute point, the area under the CER curve
from 0 to 40 minutes, and 1/tmax) were not statistically signifi-
cantly associated with irinotecan clearance, the AUC of SN-38,
or the relative extent of conversion (Table 3). In contrast, mi-
dazolam clearance (r ? .745, P?.001) and the midazolam
concentration at the 4-hour sampling point (r ? ?.416, P ?
.022) were correlated with irinotecan clearance (Fig. 2 and Table
3). A sex difference in midazolam clearance was not observed
(P ? .260). A weak correlation was found between the dose-
normalized AUC values for midazolam and SN-38G (r ? .368,
P ? .046). In contrast to earlier findings (39), some of the
erythromycin breath test parameters, including the percentage of
the administered dose in the sample obtained at the 20-minute
point and the area under the CER curve from 0 to 40 minutes,
were statistically significantly correlated with midazolam clear-
ance (r ? .529 and P ? .003 for the percentage of the admin-
istered dose in the sample obtained at the 20-minute point; and
r ? .556 and P ? .001 for the area under the CER curve from
0 to 40 minutes) and the midazolam concentration at the 4-hour
sampling point (r ? ?.503 and P ? .005 for the percentage of
Table 1. Patient demographics?
No. of patients eligible (M/F)
Median age, y (range)
Median height, m (range)
Median weight, kg (range)
Median body surface area, m2(range)
Median performance score (range)
Tumor type, No. of patients (%)
Prior chemotherapy, No. of patients (%)
?M, male; F, female; (N)SCLC, (non) small-cell lung cancer.
†Miscellaneous include cervix cancer (n ? 2), breast cancer (n ? 1), cancer
of head and neck (n ? 1), and adenocarcinoma of unknown primary (n ? 4).
Table 2. Summary of irinotecan pharmacokinetic parameters?
Parameter (range)Course 1 Course 2
No. of patients
Irinotecan CL, L/h†
Irinotecan AUC, ?g?h/mL
SN-38 AUC, ?g?h/mL
SN-38G AUC, ?g?h/mL
?CL, clearance; AUC, area under the plasma concentration versus time curve
simulated from time zero to 100 hours after start of infusion; REC, relative
extent of conversion; REG, relative extent of glucuronidation.
†Indicates typical values.
1588 ARTICLES Journal of the National Cancer Institute, Vol. 96, No. 21, November 3, 2004
by guest on June 4, 2013
the administered dose in the sample obtained at the 20-minute
point; and r ? ?.653 and P?.001 for the area under the CER
curve from 0 to 40 minutes).
Influence of Enzyme and Transporter Genotypes on
Among the five genetic variants CYP3A4*2, CYP3A4*3,
CYP3A4*17, CYP3A4*18, and CYP3A5*6, only the wild-type
sequence was found, indicating that for these variants none of
the patients in the studied cohort carried a mutant allele. The
absence of these variants is consistent with previously published
data obtained in the general European white population (40,41).
For the other six variants (ABCB1*8, ABCB1*7, ABCB1*6,
CYP3A4*1B, CYP3A5*3, and UGT1A1*28) studied, the fre-
quency of the variant allele (q) was highly variable, with only
four patients carrying at least one variant allele for CYP3A4*1B
(allele frequency, 0.09) and as many as 29 patients carrying at
least one variant allele for CYP3A5*3 (allele frequency, 0.87)
The presence of two copies of the UGT1A1*28 allele was
associated with statistically significantly altered pharmacokinet-
ics of irinotecan. The UGT1A1*28 allele carries a seven-copy
TA repeat (TA7) in the promoter region instead of a six-copy
repeat (TA6), as in the wild-type UGT1A1 allele. The AUC of
SN-38 was 435 ng · h/mL (95% CI ? 339 to 531 ng · h/mL) in
wild-type patients (TA6/TA6; n ? 12), 631 ng · h/mL (95% CI
? 499 to 762 ng · h/mL) in heterozygous patients (TA6/TA7;
n ? 15), and 1343 ng · h/mL (95% CI ? 0 to 4181 ng · h/mL)
in homozygous variant patients (TA7/TA7; n ? 3) (P ? .006).
Likewise, a UGT1A1*28 genotype-dependent relative extent of
conversion was observed, with a value of 0.018 (95% CI ?
0.013 to 0.022) in wild-type patients, 0.030 (95% CI ? 0.025 to
0.034) in heterozygous patients, and 0.042 (95% CI ? 0 to
0.092) in homozygous variant patients (P?.001). Furthermore,
the relative extent of glucuronidation was 9.27 (95% CI ? 6.00
to 12.53) in wild-type patients, 5.79 (95% CI ? 3.74 to 7.83;
TA6/TA7) in heterozygous patients, and 3.48 (95% CI ? 0 to
9.47) in homozygous variant patients (P ? .010) (Fig. 3).
National Cancer Institute Common Toxicity Criteria scores of
neutropenia (P ? .020), the observed nadir value of the absolute
neutrophil count (P ? .026), and the percent decrease in absolute
neutrophil count at nadir (P ? .024) were also associated with
the UGT1A1*28 genotype. Statistically significant associations
between irinotecan pharmacokinetics and the other variant ge-
notypes studied were not observed, although the ABCB1*8
genotype was associated with a non–statistically significant
trend in a decrease in the AUC of SN-38G (P ? .042, Kruskal–
In this study, we tested the hypothesis that the pharmacoki-
netics of irinotecan can be predicted by the putative CYP3A
phenotyping probes erythromycin and midazolam. We based the
rationale for this investigation on the observations that irinote-
can is a substrate for CYP3A4 and CYP3A5 (42) (Fig. 1) and
that its pharmacokinetic profile is influenced by potent inhibitors
(43) or inducers (44) of these isozymes. We hypothesized that by
administering a probe drug substrate for the enzymes in ques-
tion, enzymatic activity could be evaluated by comparing (met-
abolic) clearance and metabolic ratios of parent compounds
and/or of their metabolites. Like irinotecan, erythromycin is a
substrate for both CYP3A4 and ABCB1, but it is not metabo-
Fig. 2. Correlation between
midazolam clearance and
irinotecan clearance during
the first treatment in all 30
patients. Solid line repre-
sents a linear regression line
(r ?.745, P?.001). All sta-
tistical tests were two-sided.
Table 3. Correlations between CYP3A phenotype and irinotecan parameters?
Parameter 1Parameter 2
Erythromycin breath test
Area under CER0–40curve
Areas under CER0–40curve
Area under CER0–40curve
Midazolam clearance test
?CYP3A, cytochrome P450 3A; CER20, flux of exhaled14CO2at t ? 20
minutes; CL, clearance; CER0–40, flux of exhaled14CO2from 0 to 40 minutes;
AUC, area under the plasma concentration versus time curve simulated from
time zero to 100 hours after start of infusion; REC, relative extent of conversion;
REG, relative extent of glucuronidation; 1/tmax, the reciprocal of time to peak
concentration; t4, midazolam concentration in the 4-hour sample.
†A Pearson’s correlation coefficient was used to evaluate associations be-
tween irinotecan pharmacokinetics and the erythromycin breath test or midazo-
lam clearance test, respectively. All statistical tests were two-sided.
Table 4. Allele frequencies for the studied genes
A893T or S
?Number represents position in nucleotide sequence.
†Number represents amino acid codon number.
‡q ? frequency of the seven-dinucleotide (TA)7repeats.
Journal of the National Cancer Institute, Vol. 96, No. 21, November 3, 2004ARTICLES 1589
by guest on June 4, 2013
lized by CYP3A5 (39). In contrast, midazolam is metabolized by
both CYP3A4 and CYP3A5, but it is a very poor substrate for
ABCB1 (45). Because irinotecan is also partially metabolized by
CYP3A5 (42), we expected the pharmacokinetic correlations of
irinotecan with erythromycin and midazolam to differ. Because
the metabolic pathways of several other commonly used probes,
such as the CYP3A phenotyping probes dextromethorphan and
cortisol (16), differ substantially from that of irinotecan, we did
not evaluate them in this study.
We did not observe statistically significant correlations
between common parameters (CER20, area under the CER0–
40, and 1/tmax) for the erythromycin breath test and irinotecan
pharmacokinetics, but we did obtain a highly statistically
significant correlation between midazolam clearance and iri-
notecan clearance. From this correlation analysis, we predict
that, in the patient cohort in our study, 56% of total interin-
dividual variability in irinotecan clearance can be explained
by variation in CYP3A function. The discrepant findings
between the erythromycin breath test and the midazolam
clearance test may be related to the relatively slow and
inefficient metabolism of midazolam, which more closely
resembles the CYP3A-mediated metabolism of irinotecan
than the fast and extensive CYP3A-mediated metabolism of
erythromycin (39). We also cannot exclude the possibility
that ABCB1 plays a minor role in clearance of irinotecan and
a larger role in the clearance of erythromycin and that this
difference confounds pharmacokinetic interrelationships. We
also observed a trend for a relationship between the AUC for
midazolam and exposure to SN-38G. This finding was unex-
pected because SN-38G is formed from SN-38 through
UGT1A-mediated conjugation that is independent of CYP3A
activity. A possible explanation may be found in the indirect
formation of SN-38 and SN-38G from 7-ethyl-10-[4-amino-
1-piperidino]-carbonyloxycamptothecin (NPC), which results
from a ring-opening oxidation of the terminal piperidine ring
of irinotecan that is mediated by CYP3A4 (46) (Fig. 1).
Although irinotecan is considered a prodrug with little
inherent antitumor activity, the ability to accurately predict its
clearance may be clinically relevant. First, the etiology of
irinotecan-mediated side effects is still not completely under-
stood, and circulating concentrations of the parent drug may
predict hematologic toxicity associated with irinotecan treat-
ment (47). Furthermore, the ratio of concentration of total
unbound irinotecan to the concentration of total unbound
SN-38 is similar to their potency ratio in vitro, and so both
irinotecan and SN-38 are likely to be effective in vivo (5).
Second, SN-38 may be formed both peripherally in the intes-
tines, liver, and blood, as well as in the tumor from the
conversion of irinotecan by carboxylesterase 2 (hCE2), which
may be more important than previously thought and may
contribute to the variable response to irinotecan chemother-
apy for solid tumors (48,49). Consequently, knowledge of
phenotypic CYP3A activity, as a predictor of irinotecan clear-
ance in individual patients, may help to reduce the pharma-
cokinetic and subsequent pharmacodynamic variability asso-
ciated with irinotecan treatment.
We found that certain genetic variants in polymorphic pro-
teins are associated with differences in the elimination of irino-
tecan or its metabolites, as predicted previously (50). We also
confirmed earlier preliminary observations that the UGT1A1*28
genotype is independently associated with the pharmacokinetic
profile of irinotecan (51–53) and that ABCB1*8 has a smaller
association (54). Genotyping for CYP3A4 and CYP3A5 did not
result in statistically significant correlations with irinotecan
pharmacokinetics, perhaps because of the low allele frequency
of mostCYP3A variant genotypes
CYP3A4*18, and CYP3A5*1) in the white population (40,41)
or because of the absence of a clinically important effect on
enzyme activity in vivo (e.g., CYP3A4*1B) (55,56). Because
CYP3A is a complex enzyme system that is easily influenced by
environmental (i.e., comedication, herbal preparations, and/or
food substances) and physiologic (i.e., aging, disease state, and
altered liver and renal function) factors (5), the role of CYP3A
genotyping in the chemotherapeutic treatment of cancer remains
In conclusion, CYP3A phenotype (as determined by midazo-
lam clearance) and UGT1A1*28 genotype appear to be statisti-
cally significant predictors of irinotecan and SN-38 pharmaco-
kinetics, respectively. A prospective study to validate the
usefulness of these phenotyping and genotyping strategies to
optimize chemotherapeutic treatment with irinotecan for indi-
vidual patients is currently under way. In addition, limited
sampling strategies are being developed by use of data obtained
in a larger cohort of patients to further optimize the clinical
applicability of the midazolam clearance test.
Fig. 3. UGT1A1*28 genotype and the metabolism
of irinotecan to 7-ethyl-10-hydroxycamptothecin
(SN-38). A) Area under the plasma concentration
versus time curve (AUC) of SN-38. B) AUC of
SN-38 divided by the AUC of irinotecan. C) AUC
of SN-38 glucuronide (SN-38G) divided by the
AUC of SN-38. The wild-type allele of UGT1A1
is designated TA6, and the homozygous variant
allele is TA7. Solid horizontal lines ? mean val-
ues; dashed horizontal lines ? interquartile range
(25th percentile values to 75th percentile values).
1590 ARTICLESJournal of the National Cancer Institute, Vol. 96, No. 21, November 3, 2004
by guest on June 4, 2013
(1) Grothey A, Sargent D, Goldberg RM, Schmoll HJ. Survival of patients with
advanced colorectal cancer improves with the availability of fluorouracil-
leucovorin, irinotecan, and oxaliplatin in the course of treatment. J Clin
(2) Perez EA, Hillman DW, Mailliard JA, Ingle JN, Ryan JM, Fitch TR, et al.
Randomized phase II study of two irinotecan schedules for patients with
metastatic breast cancer refractory to an anthracycline, a taxane, or both.
J Clin Oncol 2004;22:2849–55.
(3) Ribrag V, Koscielny S, Vantelon JM, Ferme C, Rideller K, Carde P, et al.
Phase II trial of irinotecan (CPT-11) in relapsed or refractory non-
Hodgkin’s lymphomas. Leuk Lymphoma 2003;44:1529–33.
(4) Langer CJ. The global role of irinotecan in the treatment of lung cancer:
2003 update. Oncology (Huntingt) 2003;17:30–40.
(5) Mathijssen RH, van Alphen RJ, Verweij J, Loos WJ, Nooter K, Stoter G,
et al. Clinical pharmacokinetics and metabolism of irinotecan (CPT-11).
Clin Cancer Res 2001;7:2182–94.
(6) Saltz LB, Cox JV, Blanke C, Rosen LS, Fehrenbacher L, Moore MJ, et al.
Irinotecan plus fluorouracil and leucovorin for metastatic colorectal cancer.
Irinotecan Study Group. N Engl J Med 2000;343:905–14.
(7) Vanhoefer U, Harstrick A, Achterrath W, Cao S, Seeber S, Rustum YM.
Irinotecan in the treatment of colorectal cancer: clinical overview. J Clin
(8) Evans WE, Relling MV. Moving towards individualized medicine with
pharmacogenomics. Nature 2004;429:464–8.
(9) Sparreboom A, Cox MC, Acharya MR, Figg WD. Herbal remedies in the
United States: potential adverse interactions with anticancer agents. J Clin
(10) Rivory LP. Drug interactions. In: Figg WD, McLeod HL, editors. Hand-
book of anticancer pharmacokinetics and pharmacodynamics. Totawa (NJ):
Humana Press; 2004. p. 245–66.
(11) Baker SD, Verweij J, Rowinsky EK, Donehower RC, Schellens JH, Gro-
chow LB, et al. Role of body surface area in dosing of investigational
anticancer agents in adults, 1991-2001. J Natl Cancer Inst 2002;94:1883–8.
(12) Mathijssen RH, Verweij J, de Jonge MJ, Nooter K, Stoter G, Sparreboom
A. Impact of body-size measures on irinotecan clearance: alternative dos-
ing recommendations. J Clin Oncol 2002;20:81–7.
(13) de Jong FA, Mathijssen RH, Xie R, Verweij J, Sparreboom A. Flat-fixed
dosing of irinotecan: influence on pharmacokinetic and pharmacodynamic
variability. Clin Cancer Res 2004;10:4068–71.
(14) Guengerich FP. Cytochrome P-450 3A4: regulation and role in drug
metabolism. Annu Rev Pharmacol Toxicol 1999;39:1–17.
(15) Agundez JA. Cytochrome p450 gene polymorphism and cancer. Curr Drug
(16) Kivisto KT, Kroemer HK. Use of probe drugs as predictors of drug
metabolism in humans. J Clin Pharmacol 1997;37:40S–8S.
(17) Tanaka E, Kurata N, Yasuhara H. How useful is the “cocktail approach” for
evaluating human hepatic drug metabolizing capacity using cytochrome
P450 phenotyping probes in vivo? J Clin Pharm Ther 2003;28:157–65.
(18) Streetman DS, Bertino JS Jr, Nafziger AN. Phenotyping of drug-
metabolizing enzymes in adults: a review of in-vivo cytochrome P450
phenotyping probes. Pharmacogenetics 2000;10:187–216.
(19) Rivory LP, Watkins PB. Erythromycin breath test. Clin Pharmacol Ther
(20) Rivory LP, Slaviero KA, Hoskins JM, Clarke SJ. The erythromycin breath
test for the prediction of drug clearance. Clin Pharmacokinet 2001;40:
(21) Hirth J, Watkins PB, Strawderman M, Schott A, Bruno R, Baker LH. The
effect of an individual’s cytochrome CYP3A4 activity on docetaxel clear-
ance. Clin Cancer Res 2000;6:1255–8.
(22) Yamamoto N, Tamura T, Kamiya Y, Sekine I, Kunitoh H, Saijo N.
Correlation between docetaxel clearance and estimated cytochrome P450
activity by urinary metabolite of exogenous cortisol. J Clin Oncol 2000;
(23) Goh BC, Lee SC, Wang LZ, Fan L, Guo JY, Lamba J, et al. Explaining
interindividual variability of docetaxel pharmacokinetics and pharmacody-
namics in Asians through phenotyping and genotyping strategies. J Clin
(24) Puisset F, Chatelut E, Dalenc F, Busi F, Cresteil T, Azema J, et al.
Dexamethasone as a probe for docetaxel clearance. Cancer Chemother
(25) de Jonge MJ, Sparreboom A, Planting AS, van der Burg ME, de Boer-
Dennert MM, ter Steeg J, et al. Phase I study of 3-week schedule of
irinotecan combined with cisplatin in patients with advanced solid tumors.
J Clin Oncol 2000;18:187–94.
(26) Lane EA, Parashos I. Drug pharmacokinetics and the carbon dioxide breath
test. J Pharmacokinet Biopharm 1986;14:29–49.
(27) Ghoos YF, Maes BD, Geypens BJ, Mys G, Hiele MI, Rutgeerts PJ, et al.
Measurement of gastric emptying rate of solids by means of a carbon-
labeled octanoic acid breath test. Gastroenterology 1993;104:1640–7.
(28) Watkins PB, Murray SA, Winkelman LG, Heuman DM, Wrighton SA,
Guzelian PS. Erythromycin breath test as an assay of glucocorticoid-
inducible liver cytochromes P-450. Studies in rats and patients. J Clin
(29) Lepper ER, Hicks JK, Verweij J, Zhai S, Figg WD, Sparreboom A.
Determination of midazolam in human plasma by liquid chromatography
with mass-spectrometric detection. J Chromatogr B Analyt Technol
Biomed Life Sci 2004;806:305–10.
(30) Sparreboom A, de Jonge MJ, de Bruijn P, Brouwer E, Nooter K, Loos WJ,
et al. Irinotecan (CPT-11) metabolism and disposition in cancer patients.
Clin Cancer Res 1998;4:2747–54.
(31) de Bruijn P, Verweij J, Loos WJ, Nooter K, Stoter G, Sparreboom A.
Determination of irinotecan (CPT-11) and its active metabolite SN-38 in
human plasma by reversed-phase high-performance liquid chromatography
with fluorescence detection. J Chromatogr B Biomed Sci Appl 1997;698:
(32) de Bruijn P, de Jonge MJ, Verweij J, Loos WJ, Nooter K, Stoter G, et al.
Femtomole quantitation of 7-ethyl-10-hydroxycamptothecine (SN-38) in
plasma samples by reversed-phase high-performance liquid chromatogra-
phy. Anal Biochem 1999;269:174–8.
(33) Hesselink DA, van Schaik RH, van der Heiden IP, van der Werf M, Gregoor
PJ, Lindemans J, et al. Genetic polymorphisms of the CYP3A4, CYP3A5, and
MDR-1 genes and pharmacokinetics of the calcineurin inhibitors cyclosporine
and tacrolimus. Clin Pharmacol Ther 2003;74:245–54.
(34) Dai D, Tang J, Rose R, Hodgson E, Bienstock RJ, Mohrenweiser HW, et
al. Identification of variants of CYP3A4 and characterization of their
abilities to metabolize testosterone and chlorpyrifos. J Pharmacol Exp Ther
(35) Xie R, Mathijssen RH, Sparreboom A, Verweij J, Karlsson MO. Clinical
pharmacokinetics of irinotecan and its metabolites in relation with diarrhea.
Clin Pharmacol Ther 2002;72:265–75.
(36) Kanto JH. Midazolam: the first water-soluble benzodiazepine. Pharmacol-
ogy, pharmacokinetics and efficacy in insomnia and anesthesia. Pharma-
(37) Kim JS, Nafziger AN, Tsunoda SM, Choo EE, Streetman DS, Kashuba
AD, et al. Limited sampling strategy to predict AUC of the CYP3A
phenotyping probe midazolam in adults: application to various assay tech-
niques. J Clin Pharmacol 2002;42:376–82.
(38) Hochberg Y. A sharper Bonferroni procedure for multiple tests of signif-
icance. Biometrika 1988;75:800–2.
(39) Kinirons MT, O’Shea D, Kim RB, Groopman JD, Thummel KE, Wood AJ,
et al. Failure of erythromycin breath test to correlate with midazolam
clearance as a probe of cytochrome P4503A. Clin Pharmacol Ther 1999;
(40) van Schaik RH, de Wildt SN, Brosens R, van Fessem M, van den Anker
JN, Lindemans J. The CYP3A4*3 allele: is it really rare? Clin Chem
(41) van Schaik RH, van der Heiden IP, van den Anker JN, Lindemans J.
CYP3A5 variant allele frequencies in Dutch Caucasians. Clin Chem 2002;
(42) Santos A, Zanetta S, Cresteil T, Deroussent A, Pein F, Raymond E, et al.
Metabolism of irinotecan (CPT-11) by CYP3A4 and CYP3A5 in humans.
Clin Cancer Res 2000;6:2012–20.
of irinotecan metabolism by ketoconazole. J Clin Oncol 2002;20:3122–9.
(44) Mathijssen RH, Verweij J, de Bruijn P, Loos WJ, Sparreboom A. Effects of St.
John’s wort on irinotecan metabolism. J Natl Cancer Inst 2002;94:1247–9.
Journal of the National Cancer Institute, Vol. 96, No. 21, November 3, 2004ARTICLES 1591
by guest on June 4, 2013
(45) Kim RB, Wandel C, Leake B, Cvetkovic M, Fromm MF, Dempsey PJ, et Download full-text
al. Interrelationship between substrates and inhibitors of human CYP3A
and P-glycoprotein. Pharm Res 1999;16:408–14.
(46) Haaz MC, Riche C, Rivory LP, Robert J. Biosynthesis of an aminopi-
peridino metaboliteof irinotecan
piperidino]carbonyloxycamptothecine] by human hepatic microsomes.
Drug Metab Dispos 1998;26:769–74.
(47) de Jonge MJ, Verweij J, de Bruijn P, Brouwer E, Mathijssen RH, van
Alphen RJ, et al. Pharmacokinetic, metabolic, and pharmacodynamic pro-
files in a dose- escalating study of irinotecan and cisplatin. J Clin Oncol
(48) Xu G, Zhang W, Ma MK, McLeod HL. Human carboxylesterase 2 is
commonly expressed in tumor tissue and is correlated with activation of
irinotecan. Clin Cancer Res 2002;8:2605–11.
(49) ZhangW,Xu G, McLeod
carboxylesterase-2 expression in normal human tissues using tissue array
analysis. Appl Immunohistochem Mol Morphol 2002;10:374–80.
(50) Mathijssen RH, Marsh S, Karlsson MO, Xie R, Baker SD, Verweij J, et al.
Irinotecan pathway genotype analysis to predict pharmacokinetics. Clin
Cancer Res 2003;9:3251–8.
(51) Iyer L, Das S, Janisch L, Wen M, Ramirez J, Karrison T, et al.
UGT1A1*28 polymorphism as a determinant of irinotecan disposition and
toxicity. Pharmacogenomics J 2002;2:43–7.
(52) Innocenti F, Undevia SD, Iyer L, Chen PX, Das S, Kocherginsky M, et
al. Genetic variants in the UDP-glucuronosyltransferase 1A1 gene pre-
dict the risk of severe neutropenia of irinotecan. J Clin Oncol 2004;22:
(53) Sai K, Saeki M, Saito Y, Ozawa S, Katori N, Jinno H, et al. UGT1A1
haplotypes associated with reduced glucuronidation and increased serum
bilirubin in irinotecan-administered Japanese patients with cancer. Clin
Pharmacol Ther 2004;75:501–15.
(54) Sai K, Kaniwa N, Itoda M, Saito Y, Hasegawa R, Komamura K, et al.
Haplotype analysis of ABCB1/MDR1 blocks in a Japanese population
reveals genotype-dependent renal clearance of irinotecan. Pharmacogenet-
(55) Xie HG, Wood AJ, Kim RB, Stein CM, Wilkinson GR. Genetic vari-
ability in CYP3A5 and its possible consequences. Pharmacogenomics
(56) Floyd MD, Gervasini G, Masica AL, Mayo G, George AL Jr, Bhat K, et al.
Genotype-phenotype associations for common CYP3A4 and CYP3A5 vari-
ants in the basal and induced metabolism of midazolam in European- and
African-American men and women. Pharmacogenetics 2003;13:595–606.
Supported in part by the Cornelis Vrolijk Development Fund (IJmuiden, The
We thank Dr. Wim van den Berg, Inge van den Bos, Dirk Buijs, Dr Sjaak
Burgers, Jan Francke, Ilse van der Heiden, Hans van der Meulen, Tatjana Pronk,
Martin van Vliet, and Marloes van der Werf (all from Rotterdam) and Dr. Nicola
F. Smith (from Bethesda, MD) for their contribution to this work.
Presented, in part, at the 22nd and 23rd Annual Meeting of the American
Society of Clinical Oncology, Chicago, IL, May 31, 2003, and New Orleans, LA,
June 6, 2004.
Manuscript received June 22, 2004; revised August 26, 2004; accepted
September 20, 2004.
1592 ARTICLES Journal of the National Cancer Institute, Vol. 96, No. 21, November 3, 2004
by guest on June 4, 2013