Role of Metabolites for Drugs That Undergo Nonlinear
First-Pass Effect: Impact on Bioequivalency Assessment
Using Single-Dose Simulations
APRIL C. BRADDY,1 ANDRE J. JACKSON2
1
Division of Bioequivalence, Office of Generic Drugs, Center for Drug Evaluation and Research,
Food and Drug Administration, Rockville, Maryland
2
Office of Clinical Pharmacology, Office of Translational Sciences, Center for Drug Evaluation and Research,
Food and Drug Administration, Silver Spring, Maryland
Received 22 January 2009; revised 3 April 2009; accepted 6 May 2009
Published online 10 August 2009 in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/jps.21838
ABSTRACT: We investigated the effects of dose and intrasubject variability (ISV) on
bioequivalence (BE) of a parent drug with a single metabolite formed by nonlinear first-
pass. A BE simulation was done using a four-compartment model at doses of 17.5, 35.0,
and 70.0 mg. ISV was set at either 10% or 20% for clearance and either 20% or 50% for
the absorption rate constant, Ka. The ratio of Katest=Kareference was fixed at 1.00 while
fraction available ratios, Ftest/Freference, were varied from 1.00 to 1.25. Results showed
the probability of passing the 90% confidence interval (CI) BE requirement for AUCI,
area-under-the-concentration curve to time infinity, and Cmax, concentration maximum,
were greater for the metabolite than the parent at all Ftest/Freference ratios. For the
parent, the probability of meeting BE criteria for AUCI and Cmax declined from 100% to
60% at the 70 mg dose as the ISV for Ka increased from 20% to 50% with an increased
Ftest/Freference ratio. For the metabolite, the probability of meeting BE criteria was above
80% for all doses and ISV values and Ftest/Freference ratios less than 1.15. Results show
that the parent, reflected absorption, is more informative for determining BE than the
metabolite. Clinical data gave a similar result. ß 2009 Wiley-Liss, Inc. and the American
Pharmacists Association J Pharm Sci 99:515 523, 2010
Keywords: absorption; bioequivalence; first-pass metabolism; nonlinear pharmaco-
kinetics; simulations
INTRODUCTION Considerations in March 2003.3 The guidance
proposed that metabolites should only be used
The role of metabolites in bioequivalence (BE) to assess BE when the parent drug cannot be
determination has been extensively discussed1,2 analyzed or if the metabolite is formed as a result
and continues to be a topic of interest despite of gut wall or presystemic metabolism and contri-
the issuance of the FDA Guidance for Industry: butes meaningfully to safety and/or efficacy of
Bioavailability and Bioequivalence Studies for the drug product. Overall, data for the metabolite
Orally Administered Drug Products General are often considered to be supportive evidence
of the therapeutic outcome when it is active.
However, the metabolite will generally not be
Correspondence to: Andre J. Jackson (Telephone: 301-796-
used to meet the BE criteria.
1545; Fax: [phone]; E-mail: [e-mail])
Although the Guidance for Industry referred
Journal of Pharmaceutical Sciences, Vol. 99, 515 523 (2010)
ß 2009 Wiley-Liss, Inc. and the American Pharmacists Association only to the metabolite in general, the document
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 99, NO. 1, JANUARY 2010 515
516 BRADDY AND JACKSON
makes no distinction between drugs that exhibit The model was described by the following
linear versus nonlinear first pass effects. There- differential equations:
fore, a question remains if the metabolite role in
GUT ź A1
BE determination is indeed the same when the
metabolite is formed by nonlinear first-pass
d Dose
ź KaA1 (1)
pharmacokinetics (PK) as when formed by linear
dt
PK.4 6 There have been few rigorous studies that
dAliver
have investigated the role of the parent drugs and
ź KaA1
the metabolite formed by a nonlinear first-pass dt
effect in the determination of BE. Also, for linear
ððE CLintÞ þðð1 EÞ CLintÞÞ
drugs, it has been definitively shown via simula-
Cliver (2)
tion and experimental data that single-dose (SD)
studies are the most sensitive to changes in BE.1,7
dAparent
The objective of this article is to determine the
źðð1 EÞ CLint CliverÞ CLPU
effect of dose following SD on BE determination dt
of the parent and its single metabolite formed by
Cparent CLPM Cparent (3)
the liver nonlinear first-pass during absorption.
In addition, the influence of intrasubject varia-
dAmetabolite
bility (ISV) in Ka on the P and M 90% CIs for a
dt
drug with a metabolite formed by a nonlinear
ź E CLint Cliver þ CLPM Cparent
first-pass effect was investigated. Based upon
study results, it will be determined if the parent or
CLMU Cmetabolite (4)
the metabolite is the better indicator of BE
following a SD. where A1 is the amount remaining to be absorbed
(dose at time zero), Cliver the concentration of
parent drug in the liver, Cparent the concentration
of parent drug in plasma, Cmetabolite the concen-
MATERIALS AND METHODS tration of metabolite in plasma, Aliver the amount
of drug in the liver, Aparent the amount of parent in
Monte Carlo Simulations plasma, Ametabolite the amount of metabolite in
plasma, Vmax the maximum rate of elimination,
Model
Km the concentration at one-half the maximum
A four-compartment nonlinear first-pass model rate of elimination, CLint the intrinsic clearance.
(Fig. 1) was used to simulate parent drug and
Vmax
metabolite concentrations.
CLint ź
Km þ Cliver
CLint
1 E ź 1
QH þ CLint
CLint
E ź
QH þ CLint
CLPU is the renal clearance for parent, CLMU the
renal clearance for metabolite.
Vmax
CLPM ź
Km þ Cparent
Ka is the absorption rate constant, QH the Hepatic
blood flow-1500 mL/min.
Parameter values from a recent referenced
Figure 1. First-pass model used to simulate propra-
study8 were used for the simulations for propa-
nolol. The model is characterized by drug absorbed into
nolol and presented in Table 1.
the liver via Ka with E being extracted drug undergoing
Ka values were generated using a bivariate log
first-pass, with the 1 E fraction being available. The
normal distribution based upon the reference
formed metabolite, Fm, and absorbed parent drug are
further eliminated via renal clearance. values in Table 1. The Katest=Kareference ratio was
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 99, NO. 1, JANUARY 2010 DOI 10.1002/jps
METABOLITES AND BIOEQUIVALENCE 517
Table 1. Simulation Parameter Valuesa while the Ftest/Freference ratios were changed to the
footnoted ratio values in Table 1.
Intrasubject
For the parent, plasma concentrations were
Parameters Mean Variability
generated at from 0.0 to 60.00 h post-dosing. For
the metabolite, plasma concentrations were gene-
Single doses 17.5, 35, and 70 mg 
Km 46 ng/mL 10% rated at from 0.0 to 84.0 h postdosing. The AUC to
Vmax 458,184.1 ng/min 10%
the last quantifiable concentration after a single
CLPU 115.8 mL/min 20%
dose, (AUCT) was calculated using the trapezoidal
CLMU 204.3 mL/min 20%
rule. The terminal elimination rate constant (kel)
Aliver 5600 mL 10%
was estimated for each simulation. The rate
Aparent 2806 mL 10%
constant was determined from the least-squares
Ametabolite 1441 mL 10%
slope of the log-transformed (ln) concentration
Ka (reference) 0.34 20% and 50%
versus time curve of the log-linear phase. AUCI,
F (reference) 0.79b 
area under the concentration curve to time
a
The simulation parameter values were obtained from Ref.
infinity, was estimated by adding to AUCT the
8 used for simulations for propanolol.
b calculated remaining area Cp (t)/kel. Cmax was
F(test)/F(reference) ratios of 1, 1.05, 1.10, 1.15, 1.20, and
the observed highest plasma concentration. The
1.25 were used in the simulations. The range for the uniform
distribution was þ0.02 units for each mean F(test) value.
doses were 17.5, 35, and 70 mg.
Analysis of the Simulated Study Data
fixed at 1.00. F values were calculated using
a uniform distribution centered at 0.79 (range Log-transformed parameters from each trial were
þ0.02 U) for the reference and test/reference analyzed using SAS. The number of times AUCI
ratios of 1, 1.05, 1.10, 1.15, 1.20, and 1.25. All and Cmax for the parent and metabolite were
other parameters were generated based upon a within the acceptable 90% CI BE regulatory
bivariate log normal distribution. Simulations criterion of 80 125% were recorded. In addition,
were done at 10% ISV for Vmax and Km. Stochastic the number of times the parent and metabolite
variation was introduced in the parameters by a 90% CI s for AUCI contained the true ratio of
random number generator, Rannor (0) in the SAS the means for the simulation, (i.e., accuracy),
system. The parameter distributions were then were also recorded. The simulations focused on
used to construct a MACRO in WINSAAM9 for the effects of ISV variation in Vmax, Km, and in Ka
the simulation of the parent concentration, (Cp), (at 20% low error, LE, and 50% high error, HE)
values for the BE studies. Random assay error and additionally the role of dose and the Ftest/
(i.e., standard deviation) both proportional and Freference ratio on the probability of the parent
additive were added to each concentration as and metabolite to declare BE and the resulting
dc ź (0.2 Cp þ 1)10 where Cp is the generated CI accuracy.
concentration value. Sequence and period were
randomized to mimic the usual two-period cross-
over BE study.
Experimental Data
The simulations are based upon the assumption
that first-pass metabolism is restricted to the Additional supportive evidence was obtained from
liver and that there is no gut wall metabolism or experimental data. Clinical data submitted to the
P-glycoprotein (P-gp) involvement in the absorp- FDA for two drugs (Drug A and Drug B) that
tion of propranolol.11,12 There were no assump- formed a metabolite by a nonlinear first-pass
tions concerning protein binding made during this effect were investigated. The pharmacokinetic
simulation. study for Drug A was done in 24 subjects at single
doses of 1, 4, and 16 mg to investigate linearity.
BE studies comparing the test and reference
formulations were done on the marketed doses of 1
Simulated Single Dose Study
and 4 mg. Drug B linearity was studied in 89
A 36-subject study was simulated 1000 times at subjects at single doses of 1, 2, and 30 mg. The BE
each ( Ftest/Freference), fraction available, ratio in study was done at the 30 mg dose in 30 subjects.
Table 1. The Katest=Kareference ratio was set to 1.0 The data for both drugs were analyzed by ANOVA
DOI 10.1002/jps JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 99, NO. 1, JANUARY 2010
518 BRADDY AND JACKSON
to determine the 90% CI for the log-transformed Table 2. Slope Parameters and 90% CI for Power
Model Analysis of the Mean Simulated Parent Drug P
AUCI and Cmax.
and Metabolite M Data at the Higher Error Level
Slope 90% CI
Determination of Nonlinearity
Parameter Parent Metabolite Parent Metabolite
A power model, ln Y ź a þ b ln dose13 was used
to determine the linearity for the simulated and
AUCI 2.07 0.74 1.95 2.18 0.70 0.78
clinical single-dose plasma concentrations of the
Cmax 1.38 0.27 1.31 1.45 0.23 0.31
parent and metabolite.
Ka ź 50%, simulated run n ź 100.
RESULTS
was there a decrease in the probability of the
Simulated Study
metabolite passing the 90% CI as the Ftest/
The relationship between Cmax and dose in Freference ratio increased. The decrease in the
Figure 2 was similar to that observed in the probability of passing the 90% CI for AUCI for the
reference study8 for the propranolol Cmax. The parent was significantly greater when compared
power curve parameters are presented in Table 2. to the metabolite as the Ftest/Freference ratios were
The AUCI for the metabolite, though less than increased. This was true at all doses and levels of
dose-proportional had a slope of 0.74, while the ISV for Ftest/Freference ratios. There was a drastic
metabolite Cmax had a slope of 0.272 and was more and steep decrease in the probability of passing
impacted by the nonlinearity than the AUCI. the 90% CI for AUCI for the parent as the ratio
The power curves in Figure 3 for AUCI show the was increased from 1.00 to 1.25.
effect of dose and intrasubject variability in Ka on Figure 4 illustrated the effects of dose, ISV in Ka
the probability of passing the 90% AUCI CI for the and Ftest/Freference on the probability of the parent
parent and metabolite at different Ftest/Freference and metabolite passing the 90% CI for Cmax. The
ratios. The probability of passing the 90% AUCI CI Cmax for the parent was similar to AUCI for the
for the parent was directly related to the increase parent, Figure 3, with the higher dose having
in dose and inversely related to intrasubject exhibited the greater probability of passing the
variability in Ka. The metabolite had a 100% 90% CI for a given level of ISV in Ka as the Ftest/
probability of passing the 90% CI with 20% ISV in Freference ratios were increased. For the parent, the
Ka. Only when the Ka ISV was increased to 50% decrease in power for Cmax was much less than the
decrease in power for AUCI when the Ftest/
Freference ratio was increased from 1.00 to 1.25.
On the other hand, for the metabolite, it is Cmax
showed no response to changes in dose, the Ftest/
Freference ratios or intrasubject variation in Ka.
The results of the simulation studies showed
that AUCI for the parent and metabolite were
more sensitive to the change in dose, ISV, and
Ftest/Freference ratios than was Cmax. For AUCI, the
accuracy of the 90% CI rose as the Ftest/Freference
ratio increased. This was related to dose, the Ftest/
Freference ratio and the ISV in Ka (Fig. 5). At 50%
ISV in Ka, the accuracy was near 80% for the
parent and metbaolite until the Ftest/Freference
ratio exceeded 1.05 then decreased independent
of dose. For the low error simulations, a steep and
Figure 2. Mean AUCI and mean Cmax values
immediate decrease in accuracy was observed as
( standard deviations) versus dose showing a greater
the Ftest/Freference ratio was increased. The meta-
than dose proportional increase for parent and a less
bolite also had a similar decrease in accuracy with
than proportional increase in both parameters with
the highest accuracy observed at the higher level
dose for the metabolite. Predicted lines are based upon
the power curve fit of the data. of intrasubject variability in Ka.
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 99, NO. 1, JANUARY 2010 DOI 10.1002/jps
METABOLITES AND BIOEQUIVALENCE 519
Figure 3. Power curves versus the ( Ft/Fr) ratios, fraction absorbed ratios and
standard deviations at doses 17.5, 35, and 70 mg for AUC of the (a) parent drug and
(b) metabolite. The individual curves represent the effect of intrasubject variability in Ka
of 20% and 50%. The (Kat/Kar) ratio was fixed at 1.0 and the ( Ft/Fr) ratios were changed
from 1.0 to 1.25 but not labeled on the abscissa. Symbols are defined in the graph legend.
Experimental Data was observed that at the higher dose the prob-
ability of passing the CI for AUCI of the parent
In Table 3, the experimental data for Drug A
increased compared to lower doses (Fig. 3). A
showed a greater than dose-proportional increase
comparable decrease in intrasubject variability,
in Cmax for the parent and metabolite and a less
with increasing dose, had also been reported in
than dose-proportional increase for AUCI. Drug B
a study of the nonlinear PK of nefazadone.14
showed a greater than dose-proportional increase
This was related to the decreasing influence
in Cmax for the parent while the metabolite had a
of the proportional error as the dose increased
greater than proportional increase for Cmax and a
(i.e., the role of the increasing denominator as
less than proportional increase for AUCI. In
the dose increased) in decreasing the %CV [i.e.,
Table 4, the resulting 90% CI for drug A were
%CV ź ((Cp error) þ error)/Cp]. This did not
similar to the simulations with the metabolite
occur when the error model was only proportional.
passing at both doses while the parent only passed
The parent had a concave or greater than unity
at the lower 1 mg dose. For Drug B the parent
response to dose for AUCI and Cmax. However,
passed AUCI but failed Cmax while the metabolite
AUCI was more responsive to dose than was Cmax.
passed both parameters. The nonlinear first-pass
This was reflected in its larger changes in the
effect was also supported by the increase in the
AUCI 90% CI as a function of dose compared to
parent/metabolite ratio with dose for both Drug A
Cmax. When the Ftest/Freference ratio for AUCI was
and Drug B (data not presented).
equal to 1, the probabilities for passing the 90%
AUCI CI were 60%, 30% and 10% for doses of 70,
35, and 17.5 mg, respectively. In contrast, Cmax
DISCUSSION exhibited probabilities of passing the 90% CI of
95%, 90%, and 70% at the same respective doses.
Several known drug products have a nonlinear The relationship between the metabolite and
first-pass effect, such as propranolol, diltiazem, dose showed conventional convex curvature capa-
rufinamide, and nefazadone. In our simulations it city limited formation of the metabolite. The AUCI
DOI 10.1002/jps JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 99, NO. 1, JANUARY 2010
520 BRADDY AND JACKSON
Figure 4. Power curves versus the ( Ft/Fr) ratios, fraction absorbed ratios and
standard deviations at doses17.5, 35, and 70 mg for Cmax of the (a) parent drug and
(b) metabolite. The individual curves represent the effect of intrasubject variability in Ka
of 20% and 50%. The (Kat/Kar) ratio was fixed at 1.0 and for the ( Ft/Fr) ratios were
changed from 1.0 to 1.25 but were not labeled on the abscissa. Symbols are defined in the
graph legend.
for the metabolite had a somewhat linear rise but this observation, it is postulated that BE studies
does decrease as dose increased. This was should be done on the highest dose since the
reflected in the 90% CI for the metabolite AUCI highest dose had the best accuracy. The AUCI and
with more response to variability in Ka and Ftest/ Cmax for the metabolite were responsive to the
Freference ratio as the dose was decreased. On the increase in nonlinearity with dose for the parent
other hand, Cmax for the metabolite was far less since it was more sensitive to changes in Ftest/
responsive to dose than was the parent and seems Freference ratios than was the metabolite. The
to plateau. This was clearly reflected by the lack of AUCI for the metabolite was far less responsive
response of the metabolite 90%CI to the changes to changes in dose, Ftest/Freference ratio and Ka
in the Ftest/Freference ratio, dose, and variability in variability than was the parent. The Cmax for the
Ka. metabolite was not responsive to any of these
The accuracy of the 90% CI for AUCI reflected factors. Differences in slopes Figure 2 for AUCI for
the impact of nonlinear kinetics. As the Ftest/ the parent and Cmax for the metabolite between
Freference ratio and dose increased the degree of the simulated and experimental data indicated
nonlinearity increased. This resulted in the true that PK associated with the nonlinearity may
Ftest/Freference ratio being within the AUCI CI only differ.15 The experimental compounds A and B
90% of the time for ratios of Ftest/Freference at or had more than one metabolite affecting their PKs,
near 1.05 or less. The complicating issue for these whereas the simulated data had one formed
simulations was dose. As the dose was increased, metabolite. Although the PK expression of non-
the probability of passing the 90% CI increased for linearity may differ between simulated and
Cmax and AUCI for equivalent drugs. However, experimental data, both showed that 90% CI of
dose had no effect for drugs with Ftest/Freference the parent was sensitive to changes in absorption.
ratios near 1.25 (i.e., not bioequivalent). Based on This was supported by the experimental data for
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 99, NO. 1, JANUARY 2010 DOI 10.1002/jps
METABOLITES AND BIOEQUIVALENCE 521
Figure 5. The accuracy of the 90% CI for AUCI as the ( Ft/Fr) ratio is varied from 1.0 to
1.25 and the (Kat/Kar) ź 1.0 for (a) parent drug at 20% intrasubject variability in Ka
(b) parent drug at 50% intrasubject variability in Ka (c) for the metabolite at 20%
intrasubject variability in Ka and (d) for the metabolite at 50% intrasubject variability
in Ka.
Drug A that showed the metabolite passing the lishing BE under certain conditions, such as those
90% CI while the parent did not pass. There was drugs that may form a metabolite by a nonlinear
a similar result for Drug B with the 90% CI first-pass effect. Preferred use of the parent for
for the higher dose for the parent falling outside nonlinear drugs for BE determination was similar
the acceptable range of 80 125% of the reference. to what had been recommended for drugs with
This may be related to the mechanism/degree of linear PK (i.e., that the most sensitive metric be
nonlinearity of the product.16 used to assess BE). Therefore, the metric which
The simulations provided some insight into the most clearly reflected absorption, that is, parent
possible usefulness of the parent data for estab- drug, should be chosen to determine BE.
Table 3. Slope Parameters and 90% CI for Power Model Analysis of the Mean
Experimental Parent Drug P and Metabolite M Data
Drug A Drug B
Parameter Slope 90% CI Slope 90% CI
AUCI P (ng h/mL) 0.81 0.57 1.05 1.01 0.99 1.20
Cmax P (ng/mL) 1.17 0.91 1.42 1.42 1.30 1.54
AUCI M (ng h/mL) 0.24 0.13 0.36 0.97 0.89 1.05
Cmax M (ng/mL) 1.09 0.95 1.23 1.33 1.24 1.41
DOI 10.1002/jps JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 99, NO. 1, JANUARY 2010
522 BRADDY AND JACKSON
Table 4. 90% CI for the Test/Reference AUCI and Cmax Geometric Means for Drug A and Drug B With a Metabolite
Formed by a Nonlinear First-Pass Effect
1 mg Dose 4 mg Dose
Parameter Parent Metabolite Parent Metabolite
Drug A
AUCI (ng h/mL) 94.0 124.0 80.0 113.0 79.0 104.0 87.0 104.0
Cmax (ng/mL) 87.0 129.0 86.0 104.0 76.0 108.0 87.0 97.0
3 mg Dose
Parameter Parent Metabolite
Drug B
AUCI (ng h/mL) 89.0 107.0 91.0 105.0
Cmax (ng/mL) 69.0 117.0 81.0 102.0
ACKNOWLEDGMENTS 7. El-Tahtawy AA, Tozer TN, Harrison F, Lesko L,
Williams RL. 1998. Evaluation of bioequivalence of
highly variable drugs using clinical trial simula-
The authors would like to thank Dale Conner for
tions II: Comparison of single and multiple-dose
helpful comments and Larry Ouderkirk for his
trials using AUC and Cmax. Pharm Res 15:98
help in editing this manuscript. The views
104.
expressed in this article are those of the authors
8. Semple HA, Tam YK, Coutts RT. 1990. A computer
and do not reflect the official policy of the FDA.
simulation of the food effect: Transient changes in
No official support or endorsement by the FDA is
hepatic blood flow and Michaelis-Menten para-
intended or should be inferred.
meters as mediators of hepatic first-pass metabo-
lism and bioavailability of propranolol. Biopharm
Drug Dispos 11:61 76.
9. Stefanovski D, Moate PJ, Boston RC. 2003. Win-
REFERENCES SAAM: A windows-based compartmental modeling
system. Metabolism 52:1153 1166.
10. Fullerton T, Forest A, Levy G. 1996. Pharmaco-
1. Midha KK, Rawson MJ, Hubbard JW. 2004. The
dyanmic analysis of sparse data from concentra-
role of metabolites in bioequivalence. Pharm Res
tion-and effect-controlled clinical trials guided by a
21:1331 1344.
pilot study. An investigation by simulations.
2. Jackson AJ, Robbie G, Marroum P. 2004. Metabo-
J Pharm Sci 85:600 607.
lites and bioequivalence Past and present. Clin
11. Elsby R, Surry DD, Smith VN, Gray AJ. 2008.
Pharmacokinet 43:655 672.
Validation and application of caco-2 assays for
3. Guidance for Industry: Bioavailability and bioequi-
the in vitro evaluation of development candidates
valence studies for orally administered drug pro-
drugs as substrates or inhibitors of P-glycoprotein
ducts-General considerations. U.S. Department of
to support regulatory submissions. Xenobiotica
Health and Human Services, Food and Drug
10:1140 1164.
Administration, Center for Drug Evaluation and
12. Ludden TM. 1991. Nonlinear pharmacokinetics:
Research [March 2003].
Clinical implications. Clin Pharmacokin 20:429
4. Chen ML, Jackson AJ. 1991. The role of metabolites
446.
in bioequivalency assessment. I. Linear pharmaco-
13. Gough K, Hutchison M, Keene O, Byrom B, Ellis S,
kinetics without first pass effect. Pharm Res 8:25
Lacey L, McKellar J. 1995. Assessment of dose
32.
proportionality: Report from the statisticians in
5. Chen ML, Jackson AJ. 1995. The role of metabolites
the Pharmaceutical industry/Pharmacokinetics
in bioequivalency assessment. II. Drugs with linear
UK joint working party. Drug Inform J 29:1039
pharmacokinetics and first-pass effect. Pharm Res
1048.
12:700 708.
14. Barbhaiya RH, Shukla UA, Chaikin P, Greene DS,
6. Jackson AJ. 2000. The role of metabolites in bioe-
Marathe PH. 1996. Nefazadone pharmacokinetics:
quivalency assessment. III. Highly variable drugs
Assessment of nonlinearity, intra-subject vari-
with linear kinetics and first-pass effect. Pharm Res
ability and time to attain steady-state plasma con-
17:1432 1436.
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 99, NO. 1, JANUARY 2010 DOI 10.1002/jps
METABOLITES AND BIOEQUIVALENCE 523
centrations after dose escalation and de-escalation. 16. Shepard TA, Lordi N, Sparrow PE. 1993. Influence
Eur J Clin Pharmcol 50:101 107. of dose range on degree of nonlinearity detected in
15. Gabrielsson J, Weiner D. 2000. Pharmacokinetic dose-proportionality studies for drugs with satur-
and pharmacodyanmic data analysis: Concepts and able elimination: Single-dose and steady-state stu-
applications. Sweden: Swedish Pharmaceutical dies. Pharm Res 10:289 293.
Press. pp. 122 123.
DOI 10.1002/jps JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 99, NO. 1, JANUARY 2010