Biowaiver Approach for Biopharmaceutics Classification

System Class 3 Compound Metformin Hydrochloride

Using In Silico Modeling
JOHN R. CRISON,1 PETER TIMMINS,2 ANTHER KEUNG,3 VIJAY V. UPRETI,3 DAVID W. BOULTON,3 BARRY J. SCHEER4
1
Drug Product Science and Technology, Bristol–Myers Squibb, New Brunswick, New Jersey 08901
2
Drug Product Science and Technology, Bristol–Myers Squibb, Moreton, Merseyside CH46 1QW, United Kingdom
3
Discovery Medicine and Clinical Pharmacology, Bristol–Myers Squibb, Princeton, New Jersey 08540
4
Analytical and Bioanalytical Development, Bristol–Myers Squibb, New Brunswick, New Jersey 08901

Received 30 June 2011; revised 21 December 2011; accepted 4 January 2012

Published online 14 February 2012 in Wiley Online Library (wileyonlinelibrary.com). DOI 10.1002/jps.23063

ABSTRACT: The dependency of metformin in vivo disposition on the rate and extent of
dissolution was studied. The analysis includes the use of fundamental principles of drug input,
permeability, and intestinal transit time within the framework of a compartmental absorption
transit model to predict key pharmacokinetic (PK) parameters and then compare the results to
clinical data. The simulations show that the maximum plasma concentration (Cmax ) and area
under the curve (AUC) are not significantly affected when 100% of drug is released within
2 h of oral dosing, which was confirmed with corresponding human PK data. Furthermore,
in vitro dissolution profiles measured in aqueous buffers at pH values of 1.2, 4.5, and 6.8 were
slower than in vivo release profiles generated by deconvolution of metformin products that were
bioequivalent. On the basis of this work, formulations of metformin that release 100% in vitro
in a time period equal to or less than two hours are indicated to be bioequivalent. The use of
modeling offers a mechanistic-based approach for demonstrating acceptable bioperformance for
metformin formulations without having to resort to in vivo bioequivalence studies and may be
more robust than statistical comparison of in vitro release profiles. This work further provides
a strategy for considering Biopharmaceutics Classification System (BCS) Class 3 compounds to
be included under biowaiver guidelines as for BCS Class 1 compounds. © 2012 Wiley Periodicals,
Inc. and the American Pharmacists Association J Pharm Sci 101:1773–1782, 2012
Keywords: oral absorption; Biopharmaceutics Classification System (BCS); mathematical
model; dissolution; permeability

INTRODUCTION As such it might be expected to have limited, pas-

sive diffusion through cell membranes and, based on
Metformin is a biguanide with antihyperglycemic
Caco-2 studies, the proposed mechanism of absorption
properties that is widely used in the treatment of
is passive paracellular, that is, 91%–95% paracellular
type II diabetes, being recommended as first-line ther-
and 5%–9% transcellular.3,4
apy in all newly diagnosed patients regardless of
The major site of absorption for metformin is the
age.1 It is available as its hydrochloride salt in both
proximal small intestine and the primary route of
immediate and modified release dosage forms. It is
elimination is via the kidneys.5 Studies have sug-
a relatively low-molecular-weight (MW) hydrophilic
gested that metformin is a substrate for organic cation
base (MW = 129.17 Da, pKa 11.5), has a logD of
transporters (OCTs) in enterocytes, small intestine,
−3.37 at pH 4, and an aqueous solubility greater than
and hepatocytes, although no specific transporter has
100 mg/mL throughout the physiological pH range.2
been identified.6 The relevance of these transporters
to metformin absorption is unclear as volunteers with
Correspondence to: John R. Crison (Telephone: +732-227-5890;
Fax: +732-227-3986; E-mail: john.crison@bms.com) genetic variation in expression of OCT did not show
Journal of Pharmaceutical Sciences, Vol. 101, 1773–1782 (2012) marked differences in oral absorption of metformin.6,7
© 2012 Wiley Periodicals, Inc. and the American Pharmacists Association The permeability of metformin in rat duodenum

JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 101, NO. 5, MAY 2012 1773

1774 CRISON ET AL.
decreased by 60% when perfused with a 200 :g/mL
versus 50 :g/mL solution of metformin based on
single-pass in situ perfusion experiments.8 Met-
formin is a Biopharmaceutics Classification System
Class 3 (BCS 3) compound (high solubility, poor
permeability).9,10
During the development of fixed dose combina-
tion (FDC) products containing metformin for global
regulatory submission, there was a need to assure
equivalent bioperformance of metformin from the
fixed combination product to metformin marketed
in different territories. Drug release studies on met-
formin immediate-release tablets sourced from differ-
ent markets showed diverse drug release properties.
During this investigation, one of the FDC products
had dissimilar in vitro dissolution profiles to the refer-
ence product, yet was demonstrated as bioequivalent
in a subsequent clinical study. It was desired not to
undertake multiple bioequivalence studies for a well
established, widely studied drug-like metformin, but
to investigate whether a biowaiver approach might be
feasible.
A biowaiver approach allows, on the basis of com-
parative in vitro release studies, the approval of
changes to a drug product that are predicted not to
affect in vivo performance, minimizing review burden
on regulators and avoiding unnecessary clinical stud-
ies, expense, and delay for the sponsor. Several groups
have argued that BCS 3 compounds should be eligi-
ble for biowaiver status in a similar fashion to BCS
1 compounds because the controlling factors in the
absorption process from the drug product are not the
drug substance solubility.9,11–16 Because the absorp-
tion profile of BCS 3 compounds is also affected by the
gastrointestinal (GI) transit time, release times must
be within the boundaries of the absorption window in
the intestine for two products to be bioequivalent, and
any biowaiver approach accommodating BCS 3 drugs
has to consider this.
To date, however, biowaivers are only broadly
recognized as applicable to BCS 1, highly soluble,
highly permeable, compounds, although European
Medicines Agency guidelines allow application of
biowaiver in the case of BCS 3 compounds under very
specific conditions of excipient composition of the com-
pared materials and where dissolution is very rapid,
not less than 85% dissolved in 15 min across the phys-
iological pH range.17
This present work represents a thorough approach
using validated computer simulations from a mecha-
nistically based in silico absorption model to predict
performance for use as a biowaiver for this BCS 3 com-
pound by modeling within the range of gastric transit
times expected in human subjects in order to show
the broad range of release rates that are expected to
have no impact on AUC and Cmax and therefore re-
sult in bioequivalence. The model used accounts for
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 101, NO. 5, MAY 2012
the limited permeability through the intestinal mem-
brane and the change in that profile throughout the
GI tract and has shown to be predictive based on clin-
ical data for a wide range of in vitro release times.
This more detailed biopharmaceutics-based approach
to a biowaiver proposal avoids the need for very rapid
dissolution times and challenges of current similar-
ity factor assessments for in vitro dissolution profiles,
which may show nonsimilarity for bioequivalent drug
products.
MATERIALS AND METHODS
Software
Commercially available software, GastroPlusTM , ver-
sion 7.0 (Simulations Plus, Inc., Lancaster, Califor-
nia) was used to model the absorption, distribution,
and elimination of metformin in humans. WinNonlin
version 5.2.1 (Pharsight Corporation, Cary, North
Carolina) was used to perform the statistical analy-
sis for establishing bioequivalence between the model
simulations and clinical data and between study
groups.
In Vitro Drug Release Studies
Dissolution was determined using United States
Pharmacopeia Apparatus 1 (basket), with a rotation
speed of 100 rpm in 1000 mL of dissolution medium
for pH 1.2, 4.5, and 6.8. Dissolution media (1000 mL,
37◦ C) used were 50 mM potassium phosphate (pH
6.8), 50 mM sodium acetate buffer (pH 4.5), and
0.1 N HCl (pH 1.2). Samples were withdrawn at 15,
30, 45, and 60 min and filtered, and quantitation of
metformin in the samples was achieved using high-
performance liquid chromatography with ultraviolet
detection at 232 nm. In vitro dissolution data are
presented in Figures 1–4.
In Vivo Clinical Studies
Descriptions of the clinical studies presented in this
paper, that is, study size, test conditions, pharma-
cokinetics (PK) results, and so on, are provided in
Table 1. The clinical data for the immediate-release
formulations that are presented were dosed under fed
conditions per approved dosage and administration
requirements noted in the product package insert.18
The extended-release clinical data used were gener-
ated under fasted conditions.
In Silico Model Development
Parameter Input
The data used in developing the metformin model for
a 500 mg dose were taken from the literature and the
parameter input values are listed in Table 2. The
DOI 10.1002/jps
 BIOWAIVER APPROACH FOR BCS 3 COMPOUND METFORMIN HYDROCHLORIDE 1775

Figure 1. (a) In vitro dissolution profile for test and reference products used in clinical studies
a. The f2 test values for this plot is 52. (b) In vitro dissolution profile for test and reference
products used in clinical studies a. The f2 test value for this plot is 42. (c) In vitro dissolution
profile for test and reference products used in clinical studies a. The f2 test values for this plot
could not be calculated. (d) In vitro dissolution profile for test and reference products used in
clinical studies b. The f2 test value for this plot is 57. (e) In vitro dissolution profile for test
and reference products used in clinical studies b. The f2 test value for this plot is 43. (f) In vitro
dissolution profile for test and reference products used in clinical studies b. The f2 test values for
this plot could not be calculated. (g) In vitro dissolution profile for test and reference products
used in clinical studies c. The f2 test value for this plot is 31. (h) In vitro dissolution profile
for test and reference products used in clinical studies c. The f2 test value for this plot is 32.
(i) In vitro dissolution profile for test and reference products used in clinical studies C. The f2
test values for this plot could not be calculated.

absorption and clearance parameters used in the a 500 mg solution of metformin.19 This study also
model are based on literature values for intravenous showed that the drug was primarily absorbed in the
dosing and data generated from a clinical study de- proximal small intestine and the primary route for
signed to determine the site-specific absorption for elimination is the kidneys, which is in agreement with

DOI 10.1002/jps JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 101, NO. 5, MAY 2012
1776 CRISON ET AL.

Figure 4. In vitro dissolution profiles at pH 1.2, 4.5, and

Figure 2. In vitro dissolution profiles at pH 1.2, 4.5, and 6.8 for Study c, metformin reference.
6.8 for Study a, metformin reference.

Metformin PK is reported to be nonlinear with re-

Figure 3. In vitro dissolution profiles at pH 1.2, 4.5, and
6.8 for Study b, metformin reference.
other published results.5 The permeability parame-
ter in the model was scaled to fit the clinical data
and agrees well with jejunal permeability measure-
ments in rats based on an established rat to human
correlation.8,20 Remaining model parameters used in
the model are based on the default physiology param-
eters in GastroPlusTM (Simulations Plus, Inc.), which
are established values that have been reported in the
literature.
spect to dose and studies have suggested that met-
formin is a substrate for OCTs in enterocytes, small
intestine, and hepatocytes, although no specific trans-
porter has been identified.6 The relevance of these
transporters to metformin absorption is unclear as
volunteers with genetic variation in expression of
OCT did not show marked differences in oral absorp-
tion of metformin.6,7 In the present work, however,
the model developed is based on 500 mg data and
therefore the simulations are limited to this dose.
Data from the human bioavailability clinical stud-
ies were used in the model development and to vali-
date the model’s capability to predict human plasma
concentrations following administration of a 500 mg
oral dose. The model was deemed predictive of the
clinical data by performing a virtual clinical trial (de-
fined as the test article) and comparing it statistically
to the observed clinical data (defined as the reference
article). Two clinical studies for immediate-release
formulations were used in the model development and
two different clinical studies, one immediate release
and one extended release, were used as validation to
confirm that the model was predictive over a wide
range of drug release times. These formulations pro-
vided a wide range of in vitro release times to test the
predictive capabilities of the model.

Table 1. Clinical Study Details—Metformin Reference Product (Discovery Medicine and Clinical Pharmacology, Bristol–Myers Squibb)

Cmax (ng/mL), AUC(0– T) (ng·h/mL),

Study Number of Subjects Dose (mg) Description Geomean (%CV) Geomean (%CV)
Study a 26 500 Immediate release, single dose, fed 810.1 (22.0) 7497.3 (24.0)
Study b 24 500 Immediate release, single dose, fed 974.8 (20.4) 7288.5 (17.1)
Study c 27 500 Immediate release, single dose, fed 1019.6 (26.1) 7080.8 (27.0)
Study d 15 500 Extended release, single dose, fasted 553.3 (28.8) 4186.0 (27.4)

JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 101, NO. 5, MAY 2012 DOI 10.1002/jps
 BIOWAIVER APPROACH FOR BCS 3 COMPOUND METFORMIN HYDROCHLORIDE 1777

Table 2. Model Parameters Nonlin Bioequivalence Toolkit; Pharsight Corpora-

Parameter Value
tion) was performed. Except where noted, the test
formulation was the virtual trial simulation and the
Physical–chemical Dose 500 mg
reference formulation was the clinical trial data. The
log D −3.37 (pH 4)
Physiology test and reference formulations were considered bioe-
Permeability 2.2 × 10−4 cm/s quivalent if the 90% confidence interval for Cmax and
(jejunum) AUC(0–T) fell within the range of 0.80–1.25.21
Absorption scale factors Once the model was finalized, all parameters were
Duodenum 0.549
kept constant during the simulations comparing the
Jejunum 1 0.540
Jejunum 2 0.536 different release profiles.
Ileum 1 0.530
Generation of In Vitro Release Profiles for Evaluating
Ileum 2 0.500
Ileum 3 0.707 the Effect of Drug Release Rate on Predicted In Vivo
Cecum 0.015 Performance
Ascending colon 0.015
Pharmacokinetic Simulated drug release profiles within the range of
Clearance 0.61 [L/(h kg)] 5 min up to 10 h for 100% of metformin to be released
Volume of distribution 0.578 L/kg (Fig. 5) were used as input for the model. The Higuchi
(central) equation was used to simulate these profiles based on
K12 1.92 1/h
the following two assumptions22 : (1) the dissolution
K21 1.03 1/h
K13 0.137 1/h of metformin is assumed to be rapid if there are no
K31 0.237 1/h formulation effects due to its high solubility (100%
within 15 min) and (2) release of the metformin be-
yond 15 min is diffusion limited due to the formu-
Deconvolution lation, as was evidenced from the slower releasing
in vitro profiles, that is, 85% greater than 15 min, be-
Deconvolution was performed on each individual ing proportional to the square root of time. The Hix-
subject plasma concentration versus time data to son–Crowell Cube Root Law was not used for these
determine the in vivo release for the three immediate- simulations as this equation is written to assume that
release formulation studies using the mechanistic ab- dissolution is primarily a function of the particle size
sorption model developed by Simulations Plus, Inc. and solubility of the drug.23 As metformin is highly
(IVIVCPlusTM module, Simulations Plus, Inc.). This soluble, much greater than 100 mg/mL, particle size
approach does not assume a constant absorption rate does not affect the dissolution rate within the range
throughout the GI tract as in traditional methods. of particle diameters that are used in the formula-
The in vivo release profile was optimized such that tions. The simulations to predict the plasma concen-
when convoluted, it gave the best fit to observed clin- trations of metformin for different release rates were
ical data. performed as virtual clinical trials so that variance
could be introduced into the predictions and the AUC
Virtual Trials—Bioequivalence Testing
and Cmax compared statistically to the clinical data.
In order to include intersubject variability into the
simulations, virtual clinical trials were simulated us-
RESULTS AND DISCUSSION
ing the “virtual trial” feature of the GastroPlusTM
software (Simulations Plus, Inc.). The coefficient of The extension of previous biowaiver considerations
variation (CV) of the model input parameters, such beyond BCS 1 compounds to BCS 3 compounds
as clearance, volume of distribution, subject weight, has been previously studied for both metformin and
and so on, was adjusted to match the variability of the cimetidine.9,12–15 A simulations approach investigat-
clinical study data, thereby resulting in %CV for the ing the range of absorption and elimination rate con-
Cmax and AUC similar to those of the clinical study. stants for typical BCS 1 and BCS 3 compounds sug-
The numbers of subjects used in the virtual trial sim- gested that all BCS 3 compounds might be appro-
ulations were the same as the number of subjects for priate for biowaivers.13 Comparing in vitro dissolu-
the clinical study being compared. tion parameters with in vivo performance for two
Bioequivalence between the test and reference metformin formulations with almost identical drug
products was established for the area under the curve release profiles across physiological pH range indi-
determined from the plasma concentration vs. time cated that a biowaiver approach would be feasible
data for a specified time interval (AUC(0 –T)) and for this compound, although the evaluation for dis-
Cmax obtained from the individual subject plasma con- similar drug release profiles was not undertaken.13
centration–time profiles and an average bioequiva- That issue was pursued by Homsek et al.14 who stud-
lence assessment using the fixed effects model (Win- ied two formulations of metformin with differing drug

DOI 10.1002/jps JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 101, NO. 5, MAY 2012
1778 CRISON ET AL.
Figure 5. Simulated metformin release profiles based on
the Higuchi equation plus in vitro dissolution profiles from
products from Studies a and d. These profiles were used as
input to the PK absorption model.
release properties that were demonstrated to be bioe-
quivalent and a good in vitro–in vivo correlation was
established. Similarity factor calculations of drug re-
lease profiles failed to demonstrate similarity of the
products tested, which would have obviated the pos-
sibility if a biowaiver approach as is done for BCS 1
compounds.
Jantratid et al.15 showed that cimetidine formula-
tions engineered to have different release rates could
be used to define an in vitro–in vivo correlation that
identified the limits of dissolution rate ranges where
an effect on bioperformance might be expected.15 This
allowed definition of tightly defined criteria around
formulation, dissolution method, and a “rapidly dis-
solving” performance in that method that could be
used as a basis for a biowaiver for cimetidine tablets.
These existing studies have not fully considered the
interaction of drug release, permeability, and GI tran-
sit time and so may still not encompass all dosage
forms with quite diverse drug release characteris-
tics that would be bioequivalent, although Tsume
and Amidon16 have undertaken in silico modeling
for three BCS 3 compounds, atenolol, cimetidine, and
amoxicillin, and demonstrated that bioequivalence is
expected when comparing products with drug release
times of up to 60 min.16
Two approaches are commonly used to establish
bioequivalence between two drug products, that is,
statistical comparison of clinical studies and statis-
tical comparison between sets of in vitro dissolu-
tion profiles.21,24,25 For the specific case of metformin,
in silico modeling and simulation is presented to offer
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 101, NO. 5, MAY 2012
a third approach of equal validity as it includes all of
the key parameters that fully define the absorption
profile of this compound, that is, dissolution, perme-
ability, and GI residence time. In comparing an f2 cal-
culation versus a modeling and simulation approach,
as BCS 3 compounds have limited permeability, tran-
sit time becomes a critical factor to the total fraction
dose absorbed, and a model which includes GI tran-
sit as well as dissolution and permeability should be
more mechanistically accurate and a stronger tool for
making bioequivalence comparisons than dissolution
alone. It is based on this argument that the in sil-
ico model was considered to show that for a specific
range of dissolution rates, metformin products should
be bioequivalent regardless of the results of the f2 test.
Confirmation of Model Predictability
Table 3 shows the point estimates and upper and
lower 90% confidence intervals comparing the model
predictions to the observed clinical values for three
immediate-release and one sustained-release met-
formin products. Studies B and C were used to build
the model and then the model was used to predict
the AUC(0–T) and Cmax for Study a (immediate re-
lease) and Study d (extended release). In each case,
the simulations were bioequivalent confirming the ac-
curacy of the model to predict clinical PK outcomes in
multiple ways. First, it shows that model variance
adequately represents the subject variability for each
clinical group studied. Furthermore, as the slow re-
lease dose (extended-release formulation) is also well
predicted, these results show that the model accu-
rately represents the permeability profile and transit
times throughout the entire GI tract and provides an
example where the release rate of metformin begins
to impact the Cmax and AUC.
This model was then used to simulate the plasma
profiles for drug release profiles ranging from 100%
drug release in 5 min up to 100% drug release
in 14 h.
In Vitro Dissolution
The premise for the present work is based on the
in vitro comparison of two 500 mg immediate-release
metformin products (an FDC compared with a single
entity) at pH values of 1.2, 4.5, and 6.8. The aver-
age data for the in vitro dissolution for the metformin
products used in the clinical studies summarized in
Table 1 are shown in Figures 1a–1i. Although there
is considerable variability in the 60 mi dissolution for
these products, all were shown to be bioequivalent to
the metformin reference product. In Figures 2, 3, and
4, the in vitro dissolution profiles at the individual pH
values are plotted together for each clinical study as
a function of pH and time (for the immediate-release
dosage forms only). As is shown in these surface plots,
the pH does not impact the release of drug over 60 min
DOI 10.1002/jps
 BIOWAIVER APPROACH FOR BCS 3 COMPOUND METFORMIN HYDROCHLORIDE
Table 3. Results of Model Simulations and Statistical Comparison of the Cmax and AUC(0– T) to Clinical Data
for Four 500 mg Metformin Products
Formulation Parameter
500 mg IR, fed (Study a)a AUC(0– T) (ng·h/mL)
Cmax (ng/mL)
500 mg IR, fed (Study b) AUC(0– T) (ng·h/mL)
Cmax (ng/mL)
500 mg IR, fed, (Study c) AUC(0– T) (ng·h/mL)
Cmax (ng/mL)
500 mg SR, fasted, (Study d) AUC(0– T) (ng·h/mL)
Cmax (ng/mL)
a Study a are the results of model simulations and statistical comparison of the C
metformin product exhibiting dissimilar in vitro dissolution profiles (f2 = 42) and bioequivalent clinical data.
for the Study a metformin reference, whereas there
is a greater dependence on pH for the metformin
reference products used in Studies B and C based
on the in vitro dissolution even though the products
were bioequivalent.
Figure 1b shows the in vitro dissolution profiles
of the metformin products tested from Study a that
were bioequivalent but did not meet the f2 criteria
for equivalence, that is, the dissolution at pH 4.5 had
an f2 = 42 and therefore did not meet the similarity
criteria.
Comparison of In Vitro Release Profiles to In Vivo
Performance for Metformin
Although generally limited to BCS 1 compounds and
in vitro–in vivo correlations, comparison of in vitro
release profiles has been also proposed as an argu-
ment for waiving bioequivalence studies for BCS 3
compounds. The scientific argument for BCS 3 com-
pounds is the same as for a BCS 1 compound, that is, if
two drug products have the same in vivo dissolution
profile under all luminal conditions, they will have
the same rate and extent of drug absorption and as-
sumes that any difference in bioavailability between
two products will be due to the release of drug from
the product and not permeability.26 However, it must
be noted that for BCS 3 compounds, the phrase “un-
der all luminal conditions” is more limited than for
BCS 1 compounds in that changes in the GI transit
time can have a greater impact on the fraction dose
absorbed due to the drug’s low permeability. As both
BCS 1 and 3 compounds have high solubility through-
out the luminal pH range, in vitro dissolution testing
may be considered adequate to show the similarity in
the dissolution of the two products. The pH values of-
ten used to represent “all” luminal conditions are 1.2,
4.5, and 6.8 and the statistical test used to determine
if two profiles are the same is the f2 similarity test.25
It is important to note that although the f2 similar-
ity test adequately determines when two dissolution
profiles are similar, that is, less than 10% difference
between time points, it does not provide any informa-
tion pertaining to bioequivalence if the profiles are
DOI 10.1002/jps
1779
Point Estimate Lower 90% CI Upper 90% CI
0.92 0.81 1.04
1.09 0.96 1.23
0.95 0.82 1.08
0.91 0.80 1.05
1.00 0.89 1.13
0.91 0.80 1.03
1.06 0.89 1.25
0.97 0.81 1.15
max and AUC(0– T) to clinical data for the
not closely similar. The test infers that if the f2 value
is less than 50, the two products that are being com-
pared are not bioequivalent. In addition to the implicit
nature of this test, there are other restrictions to this
calculation. The f2 calculation is based on the sum of
the square of the errors and therefore is dependent on
the selection of the sample points.27–29 As a result, the
two products under consideration may or may not be
equivalent depending on the reasons described above.
Two metformin products were compared using
in vitro dissolution at pH 1.2, 0.5, and 6.8. When
the dissolution at pH 4.5 did not pass the f2 test, a
clinical study of 24 subjects was conducted for the
FDC and single entity metformin products under fed
conditions. The AUC(0-T) and Cmax for the individual
subjects were compared and the two products were
found to be bioequivalent based on the lower and up-
per 90% confidence intervals being within 0.80 and
1.25, respectively (Study a, Table 4).
Use of Modeling and Simulation to Establish
Bioequivalence
The use of in silico tools has expanded in recent years
to include contributions of the dosage form, release
rate, and GI and physiological properties in addition
to classical compartmental PK to model and predict
the drug absorption process.30,31 The advantage of us-
ing these models to predict PK outcomes is that they
go beyond comparison of the in vitro dissolution data
by adding GI transit time, permeability, and clear-
ance. The model developed for this current analysis
combines the dissolution rate of metformin from the
formulation, GI permeability and transit, and physi-
ological variability to provide an accurate prediction
of the Cmax and AUC.
Additional Considerations Regarding the Relationship
Between the In Vitro Dissolution, In Vivo Release, and
the Fraction Dose Absorbed
To help understand why this range of in vivo release
will result in no significant change to the AUC and
Cmax , the in silico model was used to simulate the
plasma concentration versus time curves using a wide
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 101, NO. 5, MAY 2012
1780 CRISON ET AL.
Table 4. Clinical Trial Results Showing Bioequivalence Statistics of the Test and Reference Metformin Products
Study # Test and Reference Products
Study a Metformin FDC Test versus Metformin Reference
Study b Metformin FDC Test versus Metformin Reference
Study c Metformin Test versus Metformin Reference
range (5 min up to 14 h release) of metformin release
profiles as input. Figure 5 shows the simulated drug
release profiles (100% released in 5 min up to 10 h) as
well as the experimental in vitro dissolution profiles
associated with the metformin products used in the
clinical studies A (83% released in 1 h and 100% re-
leased in approximately 1.5 h) and D (93% released in
12 h and approximately 100% released in 14 h) that
were used as input in the model simulations.
Figures 6 and 7 show the AUC(0–T) and Cmax plot-
ted as geometric means and SD for the different re-
lease rates. The symbols (circles) represent the clini-
cal data (Table 3) corresponding to dissolution times
observed for the products tested shown in Figure 5.
In each case, the simulations were bioequivalent to
the clinical data, that is, the Cmax and AUC(0–T) fell
within the 0.80–1.25 range for the 90% confidence
intervals. On the basis of these simulations, release
times for metformin ranging from 5 min up to 2 h did
not have a statistically significant effect on Cmax and
AUC. Beyond 2 h, the Cmax and AUC(0–T) decreased
and were no longer bioequivalent. These results are
not unexpected and are substantiated by considering
the wide range of in vitro release profiles presented in
Figure 6. Model simulations versus clinical data (circles)
for metformin release profiles ranging from 5 min to 14 h for
100% to be released. Note that the model predictions were
bioequivalent to the clinical data.
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 101, NO. 5, MAY 2012
Point Estimate (90% Confidence Interval)
Cmax (ng/mL) AUC(0– T) (ng·h/mL)
1.00 (0.97–1.04) 1.02 (0.96–1.08)
1.00 (0.93–1.08) 1.02 (0.99–1.06)
1.01 (0.96–1.06) 1.03 (1.00–1.07)
Figure 7. Model simulations versus clinical data (circles)
for metformin release profiles ranging from 5 min to 14 h for
100% to be released. Note that the model predictions were
bioequivalent to the clinical data.
Figures 1a–1i that show bioequivalence and the fact
that the rate-limiting step to absorption of metformin
is permeability and not solubility, as has been shown
by several authors.
To further the understanding of the relationship
between the in vitro drug release and the in vivo drug
release, these values were plotted in Figure 8 as the
mean ± SD for the individual data from the met-
formin products tested in the clinical studies along
with the fraction dose absorbed This plot was con-
structed by determining the mean and SD for the (1)
in vivo release profiles of the individual subjects from
the three immediate-release studies determined via
deconvolution; (2) in vitro release profiles for met-
formin at pH 1.2, 4.5, and 6.8; and (3) the fraction dose
absorbed for the mean value of the input parameters.
Comparing first the in vitro to in vivo release pro-
files, Figure 8 shows that although both profiles have
considerable variability (20%–70% for the in vivo and
7%–35% for the in vitro during the first hour), the
in vitro release is slower than the in vivo release.
When these profiles are compared with the fraction
dose absorbed, both are faster, thereby confirming
DOI 10.1002/jps
 BIOWAIVER APPROACH FOR BCS 3 COMPOUND METFORMIN HYDROCHLORIDE
Figure 8. Average (± SD) in vivo release profiles for clin-
ical studies a, b, and c, average (± SD) in vitro dissolution
for metformin at all pH values, and the metformin fraction
dose absorbed (mean values only).
that the permeability and transit time in the intestine
are rate limiting to absorption and not solubility.32
CONCLUSIONS
The modeling approach presented has defined the
range of in vitro drug release profiles that map to
a bioequivalent in vivo performance for metformin,
accounting for release rate, permeability, and transit
time. Once the initial clinical data are available to
build the model, it will allow the waiving of in vivo
bioequivalence studies for a metformin drug prod-
uct available from multiple sources where diverse
drug release properties are exhibited. This work rep-
resents a more detailed biopharmaceutics-based ap-
proach and avoids the challenges of current similar-
ity factor assessments for in vitro dissolution profiles,
which may show nonsimilarity for bioequivalent BCS
3 drug products.
It was shown through modeling techniques and
clinical data that there is no impact of in vitro drug
release on in vivo performance over a broad, defined
range of metformin release rates. This analysis also
illustrates that it is possible to develop an approach
for requesting a biowaiver for the BCS 3 compound
metformin through the use of combining PK model-
ing tools with in vitro dissolution to set a range of
release rates that are expected to have no impact on
AUC and Cmax and therefore result in bioequivalence.
DOI 10.1002/jps
1781
Although this work specifically applies to met-
formin, this approach may be applicable to other BCS
3 compounds that have well defined dissolution and
known absorption characteristics.
REFERENCES
1. Nathan DM, Buse JB, Davidson MB, Heine RJ, Holman RR,
Sherwin R, Zinmann B. 2006. Management of hyperglycaemia
in type 2 diabetes: A consensus algorithm for the initiation and
adjustment of therapy. Diabetes Care 29:1963–1972.
2. Bretnall AE, Clarke GS. 1998. Metformin hydrochloride. In
Analytical profiles of drug substances and Excipients; Brittain
HG, Ed. Vol. 25. San Diego, California: Academic Press, pp
243–293.
3. Nicklin P, Keates AC, Page T, Bailey CJ. 1996. Transfer of
metformin across monolayes of human intestinal Caco-2 cells
and across rat intestine. Int J Pharm 128:155–162.
4. Proctor WR, Bourdet DL, Thakker DR. 2008. Mechanisms un-
derlying saturable intestinal absorption of metformin. Drug
Metab Dispos 36:1650–1658.
5. Tucker GT, Casey C, Phillips PJ, Conor H, Ward JD, Woods
HF. 1981. Metformin kinetics in healthy subjects and in pa-
tients with diabetes mellitus. Br J Clin Pharmacol 12:235–246.
6. Zhou M, Xia L, Wang J. 2007. Metformin transport by a newly-
cloned proton-stimulated organic cation transporter (plasma
membrane monoamine transporter) expressed in human in-
testine. Drug Metab Dispos 35:1956–1962.
7. Shu Y, Brown C, Castro RA. 2008. Effect of genetic variation
in the organic cation transporter 1, OCT1, on metformin phar-
macokinetics. Clin Pharm Ther 83:273–280.
8. Song, NN, Li QS, Lui CX. 2006. Intestinal permeability of
metformin using single-pass intestinal perfusion in rats. World
J Gastroent 12(25):4064–4070.
9. Cheng CL, Yu LX, Lee HL, Yang CY, Lue CS, Chou CH. 2004.
Biowaiver extension potential to BCS Class III high solubil-
ity–low permeability drugs: Bridging evidence for metformin
immediate-release tablet. Eur J Pharm Sci 22:297–304.
10. Guidance for industry. 2000. Waiver of in vivo bioavailabil-
ity and bioequivalence studies for immediate release solid oral
dosage forms based on a Biopharmaceutics Classification Sys-
tem. US Department of Health and Human Services Food
and Drug Administration Center for Drug Evaluation and Re-
search (CDER), Silver Spring, MD.
11. Blume HH, Schug BH. 1996. The Biopharmaceutics Classifi-
cation System (BCS): Class III drugs—Better candidates for
BA/BE waiver? Eur J Pharm Sci 9:117–121.
12. Jantratid E, Prakongpan S, Dressman JB, Amidon GL,
Junginger HE, Midha KK, Barends DM. 2006. Biowaiver
monographs for immediate release solid oral dosage forms:
Cimetidine. J Pharm Sci 95:974–984.
13. Kortejarvi H, Urtti A, Yliperttula M. 2007. Pharmacokinetic
simulation of biowaiver criteria: The effects of gastric emp-
tying, dissolution, absorption, and elimination rates. Eur J
Pharm Sci 30(2):155–166.
14. Homsek I, Parojcic J, Dacevic M, Petrovic L, Javanovic D.
2010. Justification of metformin hydrochloride biowaiver cri-
teria based on bioequivalence study. Arzneimittelforschung
60(9):553–559.
15. Jantratid E, Prakongpan S, Amidon GL, Dressman JB. 2006.
Feasibility of biowaiver extension to biopharmaceutics classi-
fication system class III drug products. Cimetidine. Clin Phar-
macokinet 45:385–399.
16. Tsume Y, Amidon GL. 2010. The Biowaiver extension for
BCS class III drugs: The effect of dissolution rate on the
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 101, NO. 5, MAY 2012
1782 CRISON ET AL.
bioequivalence of BCS class III immediate-release drugs
predicted by computer simulation. Mol Pharm 7(4):1295–
1243.
17. EMEA Committee for Medicinal Products for Human Use
2010. Guideline on the Investigation of Bioequivalence,
European Medicines Agency, London .
18. Glucophage R
, Glucophage XRR
Package Insert, Bristol-Myers
Squibb Company, Princeton, NJ.
19. Marathe P, Greene D, Rashmi B, Wen Y, Crison J, Timmins P.
Use of a pharmacokinetic absorption model and site of absorp-
tion studies to design modified release formulations for BCS
3 compounds. Third PharmSciFair, Pharmaceutical Sciences
for the Future of Medicines. June 13–17, 2011. Prague, Czech
Republic.
20. Fagerholm U, Johansson M, Lennernas H. 1996. Comparison
between permeability coefficients in rat and human jejunum.
Pharm Res 13(9):1336–1342.
21. Guidance for industry. 2001. Statistical approaches to estab-
lishing bioequivalence. US Department of Health and Human
Services Food and Drug Administration Center for Drug Eval-
uation and Research (CDER), Silver Spring, MD.
22. Higuchi T. 1961. Rate of release of medicaments from an
ointment base containing drugs in suspension. J Pharm Sci
50(10):874–875.
23. Hixson AW, Crowell JH. 1931. Dependence of reaction velocity
upon surface and agitation. Ind Eng Chem 23:923–931.
24. Moore JW, Flanner HH. 1996. Mathematical comparison of
dissolution profiles. Pharm Technol 20(6):64–75.
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 101, NO. 5, MAY 2012
25. Guidance for Industry. 1997. Dissolution testing of immediate
release solid oral dosage forms. US Department of Health and
Human Services Food and Drug Administration Center for
Drug Evaluation and Research (CDER), Silver Spring, MD.
26. Amidon GE, Lennernäs H, Shah VP, Crison JR. 1995. A theo-
retical basis for a Biopharmaceutical Drug Classification Sys-
tem: The correlation of in vitro drug product dissolution and
in vivo bioavailability. Pharm Res 12:413–420.
27. Polli JE, Singh Rekhi G, Augsburger LL, Shah VP. 1997. Meth-
ods to compare dissolution profiles and a rationale for wide dis-
solution specifications for metoprolol tartrate tablets. J Pharm
Sci 86:690–700.
28. Chow SC, Ki FYC. 1997. Statistical comparisons between dis-
solution profiles of drug products. J Biopharm Stat 7:241–
258.
29. Shah VP, Tsong Y, Sathe P, Liu JP. 1998. In vitro dissolution
profile comparison—Statistics and analysis of the similarity
factor, f2 . Pharm Res 15(6):889–896.
30. Rowland M, Peck C, Tucker G. 2011. Physiologically-based
pharmacokinetics in drug development and regulatory science.
Annu Rev Pharmacol Toxicol 51:45–73.
31. Zhang X, Lionberger RA, Davit BM, Yu LX. 2011. Utility
of physiologically based absorption modeling in implement-
ing quality by design in drug development. AAPS J 13(1):59–
71.
32. Sinko PJ, Leesman GD, Amidon GL. 1991. Predicting fraction
dose absorbed in humans using a macroscopic mass balance
approach. Pharm Res 8(8):979–988.
DOI 10.1002/jps
Copyright of Journal of Pharmaceutical Sciences is the property of John Wiley & Sons, Inc.
and its content may not be copied or emailed to multiple sites or posted to a listserv without
the copyright holder's express written permission. However, users may print, download, or
email articles for individual use.