International Journal of Pharmaceutics 473 (2014) 356365

Contents lists available at ScienceDirect

International Journal of Pharmaceutics

journal homepage: www.elsevier.com/locate/ijpharm

Combining in vitro and in silico methods for better prediction of

surfactant effects on the absorption of poorly water soluble drugsa
fenobrate case example
Ragna Berthelsen a , Erik Sjgren b , Jette Jacobsen a , Jakob Kristensen c, Ren Holm d,
Bertil Abrahamsson e , Anette Mllertz f, *
a

Department of Pharmacy, University of Copenhagen, Denmark

Department of Pharmacy, Uppsala University, Sweden
c
Global R&D, Ferring Pharmaceuticals A/S, Copenhagen, Denmark
d
Biologics and Pharmaceutical Sciences, H. Lundbeck A/S, Copenhagen, Denmark
e
AstraZeneca Pharmaceutics, R&D, Mlndal, Sweden
f
Bioneer:FARMA, Department of Pharmacy, Universitetsparken 2, University of Copenhagen, Copenhagen 2100, Denmark
b

A R T I C L E I N F O

A B S T R A C T

Article history:
Received 30 April 2014
Accepted 30 June 2014
Available online 2 July 2014

The aim of this study was to develop a sensitive and discriminative in vitroin silico model able to
simulate the in vivo performance of three fenobrate immediate release formulations containing
different surfactants. In addition, the study was designed to investigate the effect of dissolution volume
when predicting the oral bioavailability of the formulations.
In vitro dissolution studies were carried out using the USP apparatus 2 or a mini paddle assembly,
containing 1000 mL or 100 mL fasted state biorelevant medium, respectively. In silico simulations of small
intestinal absorption were performed using the GI-Sim absorption model. All simulation runs were
performed twice adopting either a total small intestinal volume of 533 mL or 105 mL, in order to examine
the implication of free luminal water volumes for the in silico predictions.
For the tested formulations, the use of a small biorelevant dissolution volume was critical for in vitroin
silico prediction of drug absorption. Good predictions, demonstrating rank order in vivoin vitroin silico
correlations for Cmax, were obtained with in silico predictions utilizing a 105 mL estimate for the human
intestinal water content combined with solubility and dissolution data performed in a mini paddle
apparatus with 100 mL fasted state simulated media.
2014 Published by Elsevier B.V.

Chemical compounds studied in this article:

Fenobrate (PubChem CID: 3339)
Poloxamer 188, poloxamer 407 (PubChem
CID: 24751)
Sodium dodecyl sulfate (PubChem CID
342365)
Cremophor ELP
Keywords:
Immediate release tablets
Surfactants
Clinical data
Mini paddle dissolution
GI-Sim absorption model
In vivoin vitroin silico correlation

1. Introduction
Most newly discovered drug candidates developed for oral
administration belong to class II or class IV of the Biopharmaceutics Classication System (BCS); thus, displaying poor water
solubility (Amidon et al., 1995; Williams et al., 2013). Low
aqueous solubility can be a signicant problem for drugs intended
for oral administration, as it often leads to low and variable
bioavailability (Lipinski et al., 1997). Different formulation
approaches, e.g. particle size reduction, addition of surfactants
and co-solvents, salt-formation and use of amorphous or lipid

* Corresponding author. Tel.: +45 35 33 64 40; fax: +45 35 30 60 31.

E-mail address: Anette.mullertz@sund.ku.dk (A. Mllertz).
http://dx.doi.org/10.1016/j.ijpharm.2014.06.060
0378-5173/ 2014 Published by Elsevier B.V.

based drug delivery systems can be applied to increase the

bioavailability of the drug (van et al., 2011; Williams et al., 2013).
For development of an optimal drug delivery system, predictions
of the drug behaviour in the human gastro-intestinal (GI) lumen
when dosed in different delivery systems will greatly facilitate the
development process and might enable faster entry into clinical
trials (Reppas and Vertzoni, 2012). Several different models can be
used to predict the performance of dosage forms after oral
administration, e.g. in vivo and ex vivo nonclinical models, in vitro
and in silico models. Animal models can, however, display
interspecies variation compared to man. In addition, they may
be expensive and the capacity is limited (Kataoka et al., 2003).
Furthermore, the pharmaceutical industry has made a commitment to reduce the use of preclinical animal models where
possible (European Federation of Pharmaceutical Industries and

R. Berthelsen et al. / International Journal of Pharmaceutics 473 (2014) 356365

Associations, 2013); thus, making in vitro or in silico models the

preferred choice when applicable. Previous studies using either in
vitro and/or in silico models to predict the bioavailability of BCS
class II compounds have shown good in vitroin vivo correlations
(IVIVCs) (Galia et al., 1998; Sunesen et al., 2005; Takano et al.,
2012; Wagner et al., 2012), greatly supporting their use in
formulation development.
For BCS class II compounds dosed in the solid form, the
dissolution in the GI tract is generally expected to be the rate
limiting step in the overall absorption process (Amidon et al.,
1995). Therefore, when working with these compounds, dissolution models often receive more attention compared to permeability models. The conditions in the GI tract are simulated in the
dissolution models using physiologically relevant media, containing bile salts and phospholipids as these components are present in
human intestinal uids and will facilitate drug solubilisation,
which is of special importance for drugs with low aqueous
solubility (Galia et al., 1998). The composition of these biorelevant
dissolution media has received much attention over the last
decades (Bergstrom et al., 2013; Reppas and Vertzoni, 2012), while
the volume of the dissolution medium has been given less
attention. This, despite it being a limiting factor for drugs with
solubility limited drug absorption (dose number >1) (Amidon et al.,
1995). When dissolution testing is used as a prognostic tool for
drug absorption, volumes of 5001000 mL biorelevant medium
have been suggested simulating the fasted and the fed state
(Dressman et al., 1998). However, more recent studies in humans
report the mean volume in the small intestine to be much smaller,
approximately 105 mL in the fasted state (minimum 45 mL, median
83 mL and mean 105 mL; Schiller et al., 2005). There may,
therefore, be an argument to work with smaller volumes when
developing prognostic in vitro methods, due to the volume impact
on enabling formulation, e.g. surfactant concentration after
dilution, degree of supersaturation of amorphous stabilized
formulations etc.
Fenobrate is a neutral, lipophilic compound displaying a very
low aqueous solubility (<0.5 mg/L) (Granero et al., 2005) and a
high log P (5.2; Do et al., 2011), making it a BCS class II compound
and a good model compound for poorly water soluble drugs.
Fenobrate is a lipid-lowering agent used for reduction of
cholesterol levels in patients at risk of cardiovascular disease
(Balfour et al., 1990; Do et al., 2011). Fenobrate is recognized as an
inhibitor of the efux transporter, P-glycoprotein (Pgp), but not a as
substrate (Ehrhardt et al., 2004). After oral administration in
humans, fenobrate is completely hydrolysed by tissue and plasma
esterases to its active major metabolite, fenobric acid (Balfour
et al., 1990; Zhu et al., 2010).
With a clinical dataset, containing relative bioavailability data
for different fenobrate formulations as reference and access to
these formulations for additional in vitro testing, the aim of the
present study was to develop a discriminative in vitroin silico
model able to simulate the in vivo performance of these different
fenobrate formulations. Secondly, the study was designed to
investigate the effect of dissolution volume when predicting the
oral bioavailability of fenobrate formulated with different
surfactants.
2. Materials and methods
2.1. Chemicals and reagents
Fenobrate was purchased from SigmaAldrich Chemie (Deisenhofen, Germany). MeltDose1 formulations AC, and samples
from all the used ingredients, were kindly provided by LifeCycle
Pharma (Hrsholm, Denmark) and were used as delivered. TriCor1
was obtained from the manufacturer (Abbott Laboratories,

357

Chicago, IL). Crude porcine bile extract and maleic acid (analytical
grade) was purchased from SigmaAldrich Chemie (Steinheim,
Germany), and egg-phosphatidylcholine (Lipoid S PC1) was from
Lipoid GmbH (Ludwigshafen, Germany). Sodium chloride and
sodium hydroxide (both analytical grade) were purchased from E.
Merck (Darmstadt, Germany), and methanol (isocratic grade) was
purchased from VWR International (Lutterworth, UK).
2.2. Formulations investigated
Formulations AC tested in present study are all based on the
solid dispersion principle (MeltDose1-technique) (Holm and
Norling, 2006). The MeltDose1 (MD) formulations all contained
145 mg of fenobrate. The fenobrate was dispersed in hydrophilic polymer based granules. The granules were compressed
into tablets following mixing with Avicel PH 200 (binder),
Polyplasdone XL (disintegrant) and magnesium stearate (lubricant). The main difference between the three formulations was
the choice and contents of surfactants. Table 1 lists the
surfactants and contents of the three formulations. In each of
the formulations, the content of surfactants ranged from 6.4 to
8.8% (w/w) of the total content.
TriCor1 is a fenobrate nanoparticulate tablet formulation
containing fenobrate, crospovidone, docusate sodium, hypromellose 2910, lactose monohydrate, magnesium stearate, polyvinyl alcohol, silicied microcrystalline cellulose, SLS, soybean
lecithin, sucrose, talc, titanium dioxide and xanthan gum. It was
used as delivered. TriCor1 was the reference formulation in the
bioavailability study and was only included in the in vitro and in
silico studies to test if the developed method were applicable to
other types of formulations than the MD formulations.
2.3. In vivo clinical data
In vivo data from a study with 20 healthy subjects were kindly
provided by LifeCycle Pharma (Hrsholm, Denmark). Buch et al.
have previously published details on the study design and analysis
of plasma samples originating from the same series of clinical
studies (Buch et al., 2010). In short, the study was conducted as a
four-way crossover, open-label, single-dose, fasting and comparative bioavailability study of three MD fenobrate formulations vs
TriCor1 in normal, healthy, non-smoking male and female
subjects. Following an overnight fast of at least 10 h, each subject
received one tablet administered orally with 240 mL of water
(ambient temperature).
The plasma concentration vs time proles of fenobric acid
following oral administration of the four fenobrate formulations
are shown in Fig. 1. The pharmacokinetic parameters following
noncompartmental analysis are summarized in Table 2. The study
shows that two of the three tested fenobrate MD formulations (B
and C) were bioequivalent with TriCor1. Formulation A, containing
Poloxamer 407, had a peak plasma concentration (Cmax), above the
125% condence interval compared to TriCor1 and with respect to
Cmax, the three MD formulations and TriCor1 were rank ordered
A > B > TriCor1 > C.

Table 1
Surfactant composition of the MD tablet formulations.
Formulation

A
B
C

Surfactant 1

Surfactant 2

Name

Amount
(mg/tablet)

Name

Amount
(mg/tablet)

Poloxamer 407
Poloxamer 188
Poloxamer 188

53.8
16.2
53.8

Cremophor ELP
SLS

37.7
19.9

358

R. Berthelsen et al. / International Journal of Pharmaceutics 473 (2014) 356365

Table 3
USP performance test conditions, dissolution, for fenobrate capsules (The United
States Pharmacopeia and National Formulary, 2012).
USP performance Test 1
tests:
dissolution<711>

Test 2

Test 3

Apparatus

0.05 M sodium
lauryl sulfate
in water;
1000 mL
2; 75 rpm

Phosphate buffer pH
6.8  0.1 containing 0.1%
pancreatin and 2%
polysorbate 80; 900 mL
2; 75 rpm with sinkers

Time

40 min

2h

0.72% Sodium
lauryl sulfate
in water,
1000 mL
2; 75 rpm,
with sinkers
30 min

Medium

Fig. 1. Average (SEM) fenobric acid plasma concentrations in humans (n = 20)

after oral administration of different fenobrate MD immediate release formulations (formulation A (black circles), formulation B (dark grey squares), formulation
C (light grey triangles) and TriCor1 (black triangles)). For a better differentiation
between the formulations, only the rst 8 h of the plasma proles is shown.

2.4. Dissolution tests

2.4.1. USP apparatus 2
Dissolution studies were performed in USP test media 13, see
Table 3, following the monograph for fenobrate capsules (The
United States Pharmacopeia and National Formulary, 2012), using
a USP 2 dissolution apparatus (Erweka DT70, Heusenstamm,
Germany). All three media ensured sink conditions.
To simulate the condition in the upper small intestine, fasted
state simulated medium (CPH fasted) was used. Medium
preparation and composition has previously been described by
Kleberg et al. (2010). In short, the medium was based on a 20 mM
maleic acid buffer supplemented with 2.5 mM bile salts and
0.625 mM phospholipid. The pH of the medium was adjusted to
6.5, and sodium chloride was added to give an osmolality of
270 mOsm/kg.
Dissolution tests of the fenobrate MD formulations (145 mg)
in CPH fasted medium were carried out using standard dissolution volume of 1000 mL per vessel, representing maximum
capacity volume for the apparatus. With this setup sink
conditions was not obtainable for the tested formulations;
however, the maximum capacity volume was chosen to approach
sink conditions. The temperature in each vessel was maintained
at 37
C  0.5
C with a paddle revolution speed of 50 rpm.
Sampling was performed manually using 5 mL plastic syringes
connected to a stainless steel sampling device at sampling times
of 2, 4, 6, 8, 10, 15, 30 and 60 min. At each time point, 5 mL of the
dissolution medium was withdrawn from each vessel, and the
sample solution was ltered through a syringe lter w/0.45 mm
PTFE Membrane (VWR International, Radnor, PA, USA), discarding

Table 2
Mean pharmacokinetic parameters (n = 20) following oral administration of 145 mg
fenobrate formulations in percentage relative to TriCor1 with 90% condence
intervals (CI). Tmax is given as the median (minmax).
Formulation

In vivo data
Tmax
[h]

Cmax [%]
(90% CI)

AUC01 [%]
(90% CI)

TriCor1
A

2.5 (1.05.0)
1.5 (1.02.5)

1.8 (1.05.0)

2.5 (1.55.0)

100
123.4
(115.1132.3)a
95.7
(89.3102.6)
86.3
(80.692.5)

100
106.8
(102.2111.7)
103.2
(98.7108.0)
101.2
(96.9106.0)

a
Indicates values are outside 125% condence interval (not bioequivalent)
compared to TriCor1.

at least the rst 1 mL. Dissolution samples from TriCor1 tablets

were ltered through a 0.20 mm PTFE membrane syringe lter
and processed in the same way. The samples were replaced with
fresh medium at 37
C. The dissolution experiments were
performed in triplicate.
2.4.2. Mini paddle assembly
Disregarding the desire for sink conditions, dissolution tests
were also conducted with a mini paddle assembly with special
inserts and 250 mL vessels (Erweka DT 70, Heusenstamm,
Germany). This assembly is a down scaled version of USP
apparatus 2 (paddle), fabricated to remain hydrodynamics
essentially similar to the standard USP 2 apparatus (Klein and
Shah, 2008). The dissolution tests were performed using 75 mL or
100 mL CPH fasted medium per vessel, corresponding to previous
measurements of human intestinal uid volume (Schiller et al.,
2005). The stirring speed was 75 rpm and the temperature
37
C  0.5
C. 2 mL samples were drawn and processed as
described above. The dissolution experiments were performed
in triplicate.
2.5. Solubility study
The solubility of fenobrate was determined in CPH fasted
medium. Portions of approximately 15 mg of fenobrate were
weighed out into glass tubes and incubated with 10 mL of
medium on an end-to-end rotator at 37
C. Aliquots were taken at
2, 5, 24 and 48 h. Prior to sampling, the tubes were centrifuged at
4500 rpm for 10 min. 1 mL of the supernatant was transferred to
an eppendorph tube and centrifuged for 15 min at 15,000 rpm.
The supernatant was diluted with mobile phase and analysed by
HPLC. The study was performed in triplicate.
2.6. High performance liquid chromatography
Samples from the dissolution tests were analysed using an
isocratic high performance liquid chromatography (HPLC) method. Fenobrate was quantied using a Dionex ASI-100 Automated
Sample injector, P680HPLC Pump and PDA-100 Photodiode Array
Detector (Thermo Fisher Scientic, Waltham, Massachusetts,
USA). The chromatograms were evaluated using Thermo Scientic Dionex Chromeleon 7 Chromatography Data System Software
(Thermo Fisher Scientic, Waltham, Massachusetts, USA).
For the retention of fenobrate, a Phenomenex1 Kinetex1 5 m
XB-C18 100 A (100  4.6 mm) HPLC column protected by a
Phenomenex1 p/n:A/J0-9000 guard-column was used (Phenomenex, Torrance, USA). The mobile phase consisted of methanol
and puried water in the ratio 85:15 (v/v). The ow rate was
1.5 mL/min with an injection volume of 20 mL. The efuent was
monitored at 280 nm for 2.5 min, and the drug retention time was
1.4 min. The HPLC analysis was performed under ambient
conditions.

R. Berthelsen et al. / International Journal of Pharmaceutics 473 (2014) 356365

2.7. Data analysis

The dissolution curves obtained with analogous experimental
conditions were compared using the similarity factor, f2, dened by
Eq. (1), described by Moore and Flanner (1996).
8"
9
#0:5
n
<
=
1X
2
wt Rt  Tt
 100
(1)
f 2 50log 1
:
;
n t1
where Rt is the reference assay at time point t, Tt is the test assay at
time point t, n is the number of pull points, and wt is an optional
weight factor. The similarity factor directly compares the difference between the percent drug dissolved for a test and a reference
formulation. The percent drug dissolved was calculated relative to
the highest amount of drug dissolved from the tested formulations
within each experimental setup. No weight factor was used. When
two proles are identical, f2 = 100, when f2 = 50, there is an average
difference of 10%. FDA has set a public standard of 50  f2  100 to
indicate similarity between two dissolution proles (FDA, 2000).
AUC08 h for the observed and simulated plasma concentration
time proles was calculated using GraphPad Prism 6 (GraphPad
Prism Software, San Diego, CA, USA).
2.8. Experimental approach for in silico simulations
In silico simulations of small intestinal absorption were
performed using the GI-Sim absorption model, previously
described by Sjgren et al. (2013).
Physicochemical properties, e.g. intrinsic solubility and diffusivity in water, as well as intestinal permeability were gathered
from the literature (Sjgren et al., 2013). Fenobrate particle size
in the MD formulations was determined by laser diffraction and
kindly provided by the manufacturer (LifeCycle Pharma,
Hrsholm, Denmark). The particle size of TriCor1 was set to
200 nm based on a literature value (Buch et al., 2010). The
absorption of fenobrate was determined as fenobric acid
plasma concentrationtime proles, as fenobrate is completely
hydrolysed by tissue and plasma esterases (Balfour et al., 1990;
Zhu et al., 2010). Pharmacokinetic parameters of fenobric acid
were acquired in analogy with previous intestinal absorption
predictions of fenobrate (Sjgren et al., 2013). Apparent
solubility of fenobrate from MD formulations, as well as TriCor1,
was based on the experimental data acquired from the dissolution
studies from the plateau level of the dissolutiontime proles.
The micellar volume fraction fmic in CPH fasted medium was
calculated based on the phospholipids and bile salts molecular
volumes and concentrations in the dissolution medium assuming
a molecular ratio of 2:1 in the micelles, as previously reported
(Mazer et al., 1980). Deviation in fenobrate solubility measured
after dissolution of the MD formulations compared to that
measured in pure CPH fasted medium was assumed to be a

359

consequence of altered fmic in the dissolution media, due to the

introduction of surfactants from the MD formulations. Specic fmic
values for each dissolution test were calculated using two
different approaches, based on single experiment observations
or based on multiple experiments observations. With the single
experiment approach, specic fmic values for each dissolution test
were calculated based on the solubility differences of fenobrate
towards buffer, and CPH fasted medium under the assumption
that the buffer/micelle partitioning ratio of fenobrate was
unchanged by the introduction of surfactants. A specic fmic
value was calculated for each formulation for each of the volumes
of dissolutions setup investigated, i.e. 75, 100 and 1000 mL. The
calculated fmic values were then adopted in GI-Sim compensating
for the volume difference between respective dissolution test and
intestinal compartments, assuming a maximum possible increase
in micelles, i.e. all surfactants present in respective compartment.
Using the second approach to calculate the specic fmic values
(based on multiple experiments), the concentrations of each MD
surfactant, i.e. poloxamer 188, poloxamer 407, cremophor ELP and
SLS, in each GI-Sim compartment were calculated assuming that
all added surfactants were present in each compartment. The
apparent solubility of fenobrate determined in the dissolution
studies was related to the surfactant concentration, producing
linear relationships. Using linear regression, the apparent
solubility of fenobrate in each compartment was estimated
based on the calculated surfactant concentrations. Based on these
values, a new set of micellar volume fractions for the different
compartments was calculated.
To evaluate the results in silico, reference simulations of a
solution and a suspension, with the same particle size as the MD
formulations, were performed with the GI-Sim default value of
fmic = 2.0  104. Also, to investigate the performance of a direct
utilization of in vitro solubility measurements without accounting
for the potential increase in fmic, simulations were performed for
the hypothetical suspension when adopting the highest measured
fenobrate solubility in the dissolution tests. Finally, all simulations were performed adopting either a total small intestinal
volume of 533 mL or 105 mL as previously suggested by Heikkinen
et al. and Schiller et al., respectively, in order to examine the
implication of free luminal water volumes for the in silico
predictions (Heikkinen et al., 2012; Schiller et al., 2005).
3. Results
3.1. Dissolution studies
Formulation A and C, comprising the highest and lowest Cmax in
the clinical study, was tested in USP test media 13 (Table 2), all
providing sink conditions. The dissolutiontime proles are
depicted in Fig. 2. In all three sink condition setups, the dissolution
process was fast with >80% of the dose dissolved within 10 min.

Fig. 2. Mean dissolution proles  SD (n = 3) of different formulations containing 145 mg fenobrate per tablet (formulation A (black circles), formulation C (light grey
triangles)), using the USP paddle apparatus (USP test conditions 13, see Table 3).

360

R. Berthelsen et al. / International Journal of Pharmaceutics 473 (2014) 356365

Fig. 3. Mean (SD) dissolution proles (n = 3) of different formulations; formulation A (black circles), formulation B (dark grey squares), formulation C (light grey triangles)
and TriCor1 (black triangles) containing 145 mg fenobrate per tablet using (A) the USP paddle apparatus (1000 mL CPH fasted state media, 50 rpm, 37
C) or (B) the mini
paddle apparatus (100 mL CPH fasted state media, 75 rpm, 37
C). The amount of fenobrate dissolvable in the pure dissolution medium, determined in a separate solubility
study, is indicated by the dotted line.

None of the methods was able to discriminate the two

formulations (f2  64).
Dissolution proles obtained in 1000 mL and 100 mL CPH
fasted medium, using the standard USP 2 apparatus and the
mini paddle apparatus, respectively, are presented in Fig. 3 with
fenobrate solubility in the media marked by a dotted line. In
both studies, the three formulations appear to have similar
initial dissolution rates, but as no measurements where done
during the rst 2 min, it was not possible to determine it
directly. In the large volume dissolution experiment (Fig. 3A),
the three MD formulations displayed very similar dissolution
proles (f2  66) with fenobrate concentrations at the plateau
level of the dissolution curves (11.2  0.03 mg/mL, 11.5  0.2 mg/
mL and 10.8  0.2 mg/mL from formulation AC, respectively),
very similar to the solubility of fenobrate in the media
(10.9  0.08 mg/mL). In the small volume dissolution study
(Fig. 3B), formulation A releases the highest amount of its dose,
1.5  0.03%, followed by formulation B and C releasing
1.3  0.01% and 0.9  0.01% of their dose, respectively. Comparing
the amount of dose released, the small volume dissolution setup
ranked the formulations in the same order as the Cmax and
AUC01 found in the clinical study, namely A > B > C, whereas
the high volume dissolution study was unable the discriminate
the three formulations at any parameter. With the small volume
dissolution setup, the dissolution curves are all considered
different from each other as f2  46.
Dissolution proles were also determined in 75 mL CPH fasted
medium, using the mini paddle apparatus. Results are not shown
(available in the Supplementary data), as they are very similar to
those obtained in 100 mL, so decreasing the volume further did not
chance the picture.
Supplementry material related to this article found, in the
online version, at http://dx.doi.org/10.1016/j.ijpharm.2014.06.060.
After obtaining the right rank order for the MD-formulations
using the small volume dissolution setup, the dissolution prole
for TriCor1 was determined as a control, using the same method.
The dissolution prole of TriCor1 is also shown in Fig. 3B,
displaying very similar dissolution behaviour to formulation B
(f2 = 64) correlating well with the Cmax values obtained in the
clinical study.
Comparing the apparent solubility of fenobrate, measured as
the fenobrate concentration in the dissolution vessels after
60 min dissolution, it is evident that the surfactants in the
formulations greatly increase the solubility. For example, one
tablet of formulation A almost doubles the solubility of fenobrate
in 100 mL compared to that obtained in 1000 mL CPH fasted media,
18.7 mg/mL and 11.2 mg/mL, respectively.

3.2. In silico simulations

3.2.1. Reference predictions
Reference predictions of a hypothetical solution and suspension
(without fmic burst) using the equilibrium solubility in CPH fasted
medium are shown in Fig. 4. The simulated concentrationtime
prole for a solution showed greater resemblance to those
observed for the MD formulations, with high Cmax and a short
time to reach Cmax (Tmax), compared to the simulations for a
conventional suspension. The simulated exposure of a conventional suspension was signicantly lower than the observed
exposure from all MD formulations. Also, according to the
reference simulations, a general increase in Cmax can be expected
when adopting a smaller volume to the intestinal compartments in
silico, as this was observed for both the solution and the
suspension. Direct implementation of the maximum apparent
solubility (18.7 mg/mL) obtained from the dissolution study
performed in 100 mL (formulation A), not accounting for the
potential increase in fmic, increased the predicted, fabs, with 17%,
compared to the reference predictions of a conventional suspension (data not shown). However, as the exposure was still underpredicted compared to observations for the MD formulations, this
modelling approach was neglected, and further work was put into
estimation of the micellar volume fraction, as this factor seemed to

Fig. 4. Simulated plasma concentrationtime proles of fenobrate (145 mg)

administered as a hypothetical solution adopting either the volumes suggested by
Heikkinen et al. (2012) (533 mL, black dashed line) or Schiller et al. (2005)
(105 mL, grey dashed lines), or as a hypothetical suspension adopting either the
volumes suggested by Heikkinen et al. (2012) (533 mL, black dotted line) or Schiller
et al. (2005) (105 mL, grey dotted lines) to the intestinal compartments 27 in GISim. Solid lines show average observed fenobric acid plasma concentrations in
humans (n = 20) after oral administration of different fenobrate immediate release
formulations; formulation A (black), formulation B (dark grey) and formulation C
(light grey).

R. Berthelsen et al. / International Journal of Pharmaceutics 473 (2014) 356365

361

Table 4
Fenobrate solubility in CPH fasted medium with and without formulations (MD formulations AC and TriCor1). Calculated micellar fractions (fmic) in respective dissolution
tests implemented to the intestinal compartments in the prediction tool GI-Sim.
CPH fasted Formulation A

Formulation B

Formulation C

TriCor1

Dissolution test volume [mL]

S [mg/mL]
fmic (x104)

10.9
6.2

1000
11.2
6.5

100
18.7
10.7

1000
11.5
6.4

100
16.2
9.3

1000
10.9a
6.2

100
11.4
6.5

100
16.6
9.5

Intestinal compartment volumes based on Heikkinen et el.

(2012)
(2) Duodenum: 41.6 mL
(3) Jejunum 1: 154 mL
(4) Jejunum 2: 122 mL
(5) Ileum 1: 94.3 mL
(6) Ileum 2: 70.5 mL
(7) Ileum 3: 48.9 mL
Intestinal compartment volumes based on Schiller et al. (2005)

fmic
(x104)
2.0
2.0
2.0
2.0
2.0
2.0
fmic
(x104)
2.0
2.0
2.0
2.0
2.0
2.0

fmic
(x104)
10.4
4.3
4.9
5.7
7.0
9.0
fmic
(x104)
44.6
13.5
16.5
20.8
27.1
37.6

fmic
(x104)
12.9
4.9
5.7
6.8
8.4
11.1
fmic
(x104)
57.2
16.9
20.7
26.3
34.5
48.0

fmic
(x104)
6.0
3.1
3.4
3.8
4.4
5.3
fmic
(x104)
22.2
7.5
8.9
10.9
13.9
18.9

fmic
(x104)
9.4
4.0
4.5
5.3
6.4
8.2
fmic
(x104)
39.6
12.1
14.8
18.6
24.2
33.4

fmic
(x104)
2.0
2.0
2.0
2.0
2.0
2.0
fmic
(x104)
2.0
2.0
2.0
2.0
2.0
2.0

fmic
(x104)
2.7
2.2
2.2
2.3
2.4
2.6
fmic
(x104)
5.5
2.9
3.2
3.5
4.0
4.9

fmic
(x104)
10
4.1
4.7
5.5
6.7
8.7
fmic
(x104)
42.4
12.9
15.7
19.8
25.8
35.7

(2)
(3)
(4)
(5)
(6)
(7)
a

Duodenum: 8.19 mL
Jejunum 1: 30.4 mL
Jejunum 2: 24.1 mL
Ileum 1: 18.6 mL
Ileum 2: 13.9 mL
Ileum 3: 9.8 mL

Set to solubility in dissolution medium, for technical reasons (actual value: 10.8  0.3 mg/mL).

have the greatest impact on the simulated plasma concentration

time proles.
3.2.2. Single experiment correlations
A series of simulations was performed based on single
experiments, i.e. with data obtained from a single dissolution
study (n = 3). Calculated fmic values for the CPH fasted medium and
respective formulation and dissolution test are displayed in Table 4
along with the calculated fmic values implemented to the intestinal
compartments in the prediction tool GI-Sim.

Simulated curves employing the experimental results from

1000 mL and 100 mL dissolution test and compensating for an
increase in fmic are displayed in Fig. 5. Calculated AUC08 h for the
clinical dataset and simulated curves in Fig. 5 are displayed in
Table 5.
Best overall results were obtained when employing the smaller
volume for total intestinal volume in the in silico model. With the
larger intestinal volume (533 mL) and fmic-values (Fig. 5A and C),
the MD formulations overall behaved like conventional suspensions with at plasma concentration curves and low exposure

Fig. 5. Observed (solid lines) and simulated (dashed lines) plasma concentrationtime proles of fenobrate (145 mg) administered as MD formulations A (black), B (dark
grey) and C (light grey) adopting either the volumes suggested by Heikkinen et al. (2012) (533 mL, A and C) or Schiller et al. (2005) (105 mL, B and D) to the intestinal
compartments in GI-Sim. Solubility input, as well as micellar volume fraction, was based on dissolution tests of the MD formulation performed in either 1000 mL (A and B) or
100 mL (C and D) dissolution media (Table 4). In (D), the simulated plasma concentrationtime prole of TriCor1 is depicted by a dotted black line.

362

R. Berthelsen et al. / International Journal of Pharmaceutics 473 (2014) 356365

Table 5
Calculated AUC08 h for the clinical dataset and simulated curves in Fig.
Formulation

In vivo data

5.

Simulation data
Dissolution study volume
1000 mL

100 mL

Adopted intestinal volume (GI-Sim)

A
B
C
TriCor1

AUC08 h  std.
[ng  hr/mL]
64,614  12,702
55,355  13,471
51,104  13,852
54,554  11,969

533 mL

105 mL

533 mL

105 mL

AUC08 h
[ng  hr/mL]
18,247
22,328
14,497

AUC08 h
[ng  hr/mL]
42,605
50,156
26,288

AUC08 h
[ng  hr/mL]
24,290
21,502
15,179
31,388

AUC08 h
[ng  hr/mL]
52,992
48,825
30,976
59,831

despite the increased fmic adopted. Employment of the smaller

volume (105 mL) and subsequently increased fmic-values (Fig. 5B
and D) caused a large shift in the simulated plasma concentration
curves leading to a better correlation for the MD formulations. The
simulations based on data from the small dissolution study
(100 mL, Fig. 5C and D) resulted in correct Cmax rank order of the
three MD formulations, and predicted the AUC08 h of formulation
A and B to be within the standard deviation (SD) of the observed
data. In general, the simulated proles under-predicted the
exposure; the AUC08 h of the simulated curves was 82, 88 and
59% of the AUC08 h for the observed plasma concentrationtime
proles belonging to formulation AC, respectively.
The simulated plasma concentrationtime prole of TriCor1
based on input data from the mini paddle dissolution experiment
is in the same range as the MD formulations, correlating with the
clinical data. The predicted AUC08 h is well within the std. of the
observed data; however, comparing the Cmax rank-order, TriCor1
displays the highest Cmax leading to a wrong rank order of the
simulated results. This contrasts the in vitro dissolution data;
however, as TriCor1 is based on a different formulation principle, i.
e. particle size reduction, it might be difcult to compare the
different proles.

of each experiment. Fig. 6 shows this relation with the data tted
by linear regression producing a ne correlation for MD formulation A and B. For formulation C, the apparent solubility basically
remained unchanged throughout the dissolution experiments.
Fig. 7 shows the simulated plasma proles when employing the
smaller volume for total intestinal volume in the in silico model,
combined with apparent solubility input calculated for each GISim compartment based on linear regression ts from Fig. 6 (input
values are available in the Supplementary material, Table 1). The
simulations based on multiple experiments produced the overall
best correlations, with right Cmax rank order and AUC08 h within
the std. for the AUC08 h of the observed data, for all three MD
formulations. Compared to the observed data, the AUC08 h of the
simulated curves was 85, 95 and 80% of the AUC08 h for the
observed plasma concentrationtime proles belonging to formulation AC, respectively.
Supplementry material related to this article found, in the
online version, at http://dx.doi.org/10.1016/j.ijpharm.2014.06.060.

3.2.3. Multiple experiments correlations

As an alternative approach to the single experiment correlations described above, a set of simulations was performed based on
input data from multiple dissolution studies. The apparent
solubility of fenobrate after 60 min of dissolution, performed in
75, 100 and 1000 mL of CPH fasted medium, was correlated to the
calculated total surfactant concentration in the dissolution vessel

According to the United States Pharmacopeia (USP), dissolution testing of solid oral dosage forms should be done with a
quantity of the medium not less than three times that required to
form a saturated solution of the drug substance. This is done to
achieve sink conditions and retain high discriminatory power.
The usual volume of dissolution medium is 5001000 mL, with
900 mL as the most commonly used volume, but the use of higher
volumes for compounds with limited solubility is also allowed
(The United States Pharmacopeia and National Formulary, 2012).

Fig. 6. Apparent solubility (mean  SD) of fenobrate as a function of surfactant

concentration for the three MD formulations (A: poloxamer 407, B: poloxamer
188 + cremophor ELP and C: poloxamer 188 + SLS) in CPH fasted medium. The given
percentages refer to the total surfactant concentration from one tablet dissolved in
different dissolution volumes. The lines are made by linear regression with R2 = 0.99
(A; black circles), 0.98 (B; dark grey squares) and 0.82 (C; light grey triangles).

4. Discussion
4.1. USP dissolution methods and sink conditions

Fig. 7. Observed (solid lines) and simulated (dashed lines) plasma concentration
time proles of fenobrate (145 mg) administered as MD formulations A (black), B
(dark grey) and C (light grey) adopting the volume suggested by Schiller et al.
(2005) (105 mL) to the intestinal compartments in GI-Sim. Solubility input, as well
as micellar volume fraction, is available in the supporting information.

R. Berthelsen et al. / International Journal of Pharmaceutics 473 (2014) 356365

The dissolution tests described in USP monographs are designed

to test the quality of the drug product. For poorly soluble drugs,
the drug dissolution may be low using classical buffer systems
and standard apparatus, why solubility modiers, such as
surfactants, inorganic salts and organic co-solvents, are often
added to the dissolution medium to ensure sink conditions, e.g.
for fenobrate capsules the USP performance dissolution test is
prescribed to be done in, e.g. 0.05 M sodium lauryl sulphate in
1000 mL water (The United States Pharmacopeia and National
Formulary, 2012). In general, when dissolution tests are used in
quality control, sink conditions are a requirement to ensure 80%
drug release within a short time frame.
The dissolution studies performed under sink conditions
(Fig. 2) following USP methods did not discriminate between
the tested MD formulations. The reason for this is probably that the
entire dose easily can be dissolved due to the large volume, and the
solubility inuence from the different surfactants in the formulations is overruled in contrast to the in vivo situation where the
volume is much smaller. The sink conditions allowed more than
80% of the dose to be released within a short time frame but did not
in it-self ensure a good IVIVC. In general, when using the
dissolution model as a prognostic tool in the formulation
development phase, the emphasis should be on obtaining a good
IVIVC, utilizing biorelevant medium and often disregarding the
desire for sink conditions.
4.2. Modelling physiological volume of human intestinal water
content
As mentioned in the introduction, a common setup for
biorelevant dissolution is 5001000 mL of biorelevant medium
using the USP 2 apparatus (Dressman et al., 1998). In spite of this,
studies have shown that the volume in the fasted small intestine is
approximately 100 mL, depending on the amount of liquid emptying
from the stomach, absorption of uid through the gut wall, and
intestinal transit time (Mudie et al., 2010; Schiller et al., 2005).
Several studies, in which 500 mL has been used as the dissolution
volume, have resulted in ne IVIVC, e.g. Wagner et al. forecasted the
food effect of compound A utilizing 500 mL of media simulating the
fasted and fed state small intestine (Wagner et al., 2012). However,
based on the measurements of the human intestinal water content
made by Schiller et al., it was interesting to examine the effect of
different in vitro dissolution volumes (Schiller et al., 2005).
In the present study, both the in vitro and the in silico models
illustrated that the volume chosen to simulate in vivo intestinal
water content is of great importance. This was not surprising as
fenobrate has such a high dose number (D0 = (dose/dissolution
volume)/Csolubility; Amidon et al., 1995) making the amount of dose
dissolved highly volume dependent. When the solubility is the ratelimiting step for oral absorption, as for BCS class II compounds with
dose numbers signicantly larger than one, including fenobrate,
any impact on the solubility (e.g. presence of surfactant or cosolvent, precipitation and water volume) in the GI tract will directly
affect the fraction absorbed (Sutton, 2009). In vivo and in vitro,
surfactants form mixed micelles with bile salts and phospholipids
present in the bulk medium, which will affect the in situ solubility of
fenobrate, as well as it is partitioning into these micelles. Both the
micellar volume fraction and the in situ solubility see Table 3, will in
turn be affected by the volume in which the dose is dissolved and
thereby affect the fraction absorbed. Therefore, to get the best
prediction of in vivo behaviour, it is important to choose the volume
resembling in vivo conditions as closely as possible. This is also
supported by another in silico study performed by Sutton (2009),
using GastroPlusTM, where it was shown that best ts for solubility
limited compounds was associated with an estimated small
intestinal water volume of about 130 mL (Sutton, 2009).

363

The amount of dose dissolved was very low ranging from 0.9%
in the mini paddle setup to 8.2% in the standard USP 2 apparatus
caused by the inherent low aqueous solubility of fenobrate. To
get a higher amount of the dose to dissolved, the dose could be
lowered or the volume increased to resemble sink conditions.
However, in order to be able to evaluate the effect of the added
surfactants in the correct surfactant to drug ratio in an in vivo
relevant volume, the dissolution studies were done with high
doses (145 mg fenobrate) using entire tablets. The low amounts
of dose dissolved might make the correlation to in vivo drug
release difcult. However, since the model produced a good IVIVC,
the experiments suggest that the initial release prole obtained in
a small in vivo relevant dissolution volume with the entire clinical
dose is relevant for the prediction of in vivo release, at least in the
case of immediate release tablets containing fenobrate. Furthermore, the correlation to in vivo release is made more reliable by
combining in vitro data with in silico simulations integrating a
multitude of factors related to the formulation performance like
solubility, colloidal transport effect on apparent permeability,
inuence of free vs colloidal bound drug on absorption and
interplay between dissolution and permeability.
4.3. Mini paddle dissolution apparatus
The mini paddle setup was developed as a scaled down version
of the standard USP 2 paddle apparatus to reduce the sample size
and the volumes of test media needed, but retain the reliability and
predictability of the standard test apparatus (Klein, 2006). The
model has previously been used for stomach dissolution (250 mL)
(Wagner et al., 2012) and small scale dissolution studies (Klein
et al., 2012; Klein and Shah, 2008). To the best of our knowledge,
this is the rst time the apparatus has been used for dissolution
studies of whole tablets with clinical doses in small volumes, for
which purpose the apparatus showed great promise. The described
mini paddle dissolution model produced data leading to a much
better IVIVC compared to the standard USP paddle dissolution
model using 1000 mL dissolution medium. However, to test the
models utility for dissolution testing of tablets containing clinical
doses in small volumes more studies should be done using
different compounds and formulations.
4.4. Combining in vitro and in silico models
The combination of in vitro and in silico models can lead to a
better prediction of the oral absorption, as in silico models are able
to consider several aspects besides disintegration and dissolution
typically captured by dissolution tests; e.g. effects on absorption by
partitioning into micelles and particle and micelle drifting in the
aqueous boundary layer, when based on in vitro test input data for
the concrete compound/formulation (Sjgren et al., 2013; Sutton,
2009). The combination of in vitro data with in silico simulations
also makes for more relevant comparisons with clinical data, as it
enables comparison of predicted and observed plasma concentrationtime proles instead of, e.g. a dissolutiontime prole with a
plasma concentrationtime prole. The inclusion of a physiologically based in silico absorption model may also potentially be used
to improve the understanding of the factors limiting the intestinal
absorption by theoretical variation of critical factors (Sjgren et al.,
2013). In the present study, the in silico simulations emphasized
the impact of water volume and distribution into micelles for the
prediction of oral absorption of the poorly water soluble
compound fenobrate. To predict the plasma-proles for the
three MD formulations, fenobrates apparent solubility, measured
after 60 min of dissolution in the various dissolution experiments,
was used as input data instead of standard solubility measured in
water or a biorelevant medium. This was done to include the

364

R. Berthelsen et al. / International Journal of Pharmaceutics 473 (2014) 356365

potential effect on the release prole of fenobrate from the

surfactants embedded in the different formulations. As previously
discussed, the surfactants will form mixed micelles with the bile
salts and phospholipids in the biorelevant medium, increasing the
apparent solubility of fenobrate within these micelles. The
micellar volume fraction is also affected by the formation of more
mixed micelles, as well as fenobrates partitioning in and out of
these micelles. In silico simulations only accounting for the effect
on the apparent solubility, under-predicted the exposure for all
three MD formulations, emphasising the importance of simulating
the dynamic aspects related to the formation of micelles, including
effects on apparent solubility and micellar volume fraction.
However, as the best simulations were not in perfect correspondence with the clinical data, the model can still be improved. For
example, a better understanding of how mixed micelles of bile
salts, phospholipids and potential additional surfactants are
formed and how drugs partition into these micelles and
subsequently are released before absorption could potentially
improve the in silico model further.
The best correlations obtained here were based on input data
from multiple dissolution studies. This made the model more timeconsuming and more expensive but also led to the best
correlations. Physiologically-based pharmacokinetic (PBPK) models often comprise a trade-off; the more time and cost that are put
in to producing the input data, the more specic and reliable the
model may become. It is, therefore, important to clarify the
purpose and ambition of the model as well as its use, e.g. rank
ordering of different formulations or quantitative predictions of
plasma concentrationtime proles, and based on this specify the
needed level of input data vs simplicity.
Today PBPK modelling is often used as support for decision
making at different phases of the drug development, as it
comprises a technique to predict plasma concentrationtime
proles from preclinical in vitro and in vivo data. In a recent review
by Kostewicz et al., it was pointed out that PBPK models
successfully have been used to ascertain key issues in the
development of new drugs in vivo. However, there is still many
aspects that need to be addressed in order to fully exploit the utility
of the PBPK models to predict drug absorption. Presently, middle
out in silico approaches are often being used, in which in vivo data
are used to optimize or rene an existing PBPK model in a predict,
learn and conrm paradigm (Kostewicz et al., 2013). In the present
study, this middle out approach was used to examine the
implication of free luminal water volumes for the in silico
predictions, showing that smaller volumes than originally
estimated resulted in a better correlation to the clinical data.
5. Conclusions
Combining in silico and in vitro modelling proved to be a very
useful tool in the process of understanding the absorption
mechanism of the poorly water soluble compound fenobrate
from immediate release formulations. For formulations containing
poorly water soluble drugs and solubilizing surfactants, the use of
small biorelevant volumes in vitro and in silico was demonstrated to
be critical for in vitroin silico prediction of drug absorption. The
current work supports recent imaging data that the intestinal water
volume in the fasted state is about 100 mL. The in vitroin silico
modelling approach used showed potential as a simple and fast
prediction model, able to rank order different surfactant based
formulations of a low soluble model compound.
Funding
This work has been supported by the Predicting Drug
Absorption Innovation Consortium (PDA).

Acknowledgements
Special thanks to Karen Kleberg Hansen and Freja Jacobsen for
performing the solubility study.

References
Amidon, G.L., Lennernas, H., Shah, V.P., Crison, J.R., 1995. A theoretical basis for a
biopharmaceutic drug classication: the correlation of in vitro drug product
dissolution and in vivo bioavailability. Pharm. Res. 12, 413420.
Balfour, J.A., McTavish, D., Heel, R.C., 1990. Fenobrate. A review of its
pharmacodynamic and pharmacokinetic properties and therapeutic use in
dyslipidaemia. Drugs 40, 260290.
Bergstrom, C.A., Holm, R., Jorgensen, S.A., Andersson, S.B., Artursson, P., Beato, S.,
Borde, A., Box, K., Brewster, M., Dressman, J., Feng, K.I., Halbert, G., Kostewicz, E.,
McAllister, M., Muenster, U., Thinnes, J., Taylor, R., Mullertz, A., 2013. Early
pharmaceutical proling to predict oral drug absorption: current status and
unmet needs. Eur. J. Pharm. Sci. 57, 173199.
Buch, P., Holm, P., Thomassen, J.Q., Scherer, D., Branscheid, R., Kolb, U., Langguth, P.,
2010. IVIVC for fenobrate immediate release tablets using solubility and
permeability as in vitro predictors for pharmacokinetics. J. Pharm. Sci. 99, 4427
4436.
Do, T.T., Van, S.M., Mols, R., Annaert, P., Martens, J., Van, H.J., Vermant, J., Augustijns,
P., Van den Mooter, G., 2011. The conict between in vitro release studies in
human biorelevant media and the in vivo exposure in rats of the lipophilic
compound fenobrate. Int. J. Pharm. 414, 118124.
Dressman, J.B., Amidon, G.L., Reppas, C., Shah, V.P., 1998. Dissolution testing as a
prognostic tool for oral drug absorption: immediate release dosage forms.
Pharm. Res. 15, 1122.
Ehrhardt, M., Lindenmaier, H., Burhenne, J., Haefeli, W.E., Weiss, J., 2004. Inuence of
lipid lowering brates on P-glycoprotein activity in vitro. Biochem. Pharmacol.
67, 285292.
European Federation of Pharmaceutical Industries and Associations, 2003. Animal
welfare: 3R's Replace, Rene, Reduce. Online Source.
FDA., 2000. FDA Guidance for Industry: Waiver of In Vivo Bioavialability and
Bioequivalence Studies for Immediate-Release Solid Oral Dosage Forms Based
on a Biopharmaceutics Classication System. US Department of Health and
Human Services, Food and Drug Administration, Center for Drug Evaluation and
Research (CDER), August 2000. Online Source.
Galia, E., Nicolaides, E., Horter, D., Lobenberg, R., Reppas, C., Dressman, J.B., 1998.
Evaluation of various dissolution media for predicting in vivo performance of
class I and II drugs. Pharm. Res. 15, 698705.
Granero, G.E., Ramachandran, C., Amidon, G.L., 2005. Dissolution and solubility
behavior of fenobrate in sodium lauryl sulfate solutions. Drug Dev. Ind. Pharm.
31, 917922.
Heikkinen, AT, Baneyx, G, Caruso, A, Parrott, N, 2012. Application of PBPK modeling
to predict human intestinal metabolism of CYP3A substrates an evaluation
and case study using GastroPlus. Eur. J. Pharm. Sci. 47, 375386.
Holm, P., Norling, T., 2006. A tablet comprising a brate. [WO 2006084475 A2].
Patent.
Kataoka, M., Masaoka, Y., Yamazaki, Y., Sakane, T., Sezaki, H., Yamashita, S., 2003. In
vitro system to evaluate oral absorption of poorly water-soluble drugs:
simultaneous analysis on dissolution and permeation of drugs. Pharm. Res. 20,
16741680.
Kleberg, K., Jacobsen, J., Mullertz, A., 2010. Characterising the behavior of poorly
water soluble drugs in the intestine: application of biorelevant media for
solubility, dissolution and transport studies. J. Pharm. Pharmacol. 62, 113.
Klein, S., 2006. The mini paddle apparatus a useful tool in the early developmental
stage? Experiences with immediate-release dosage forms Dissolution Technol.
13, 611 Generic.
Klein, S., Buchanan, N.L., Buchanan, C.M., 2012. Miniaturized transfer models to
predict the precipitation of poorly soluble weak bases upon entry into the small
intestine. AAPS PharmSciTech 13, 12301235.
Klein, S., Shah, V.P., 2008. A standardized mini paddle apparatus as an alternative to
the standard paddle. AAPS PharmSciTech 9, 11791184.
Kostewicz, E.S., Aarons, L., Bergstrand, M., Bolger, M.B., Galetin, A., Hatley, O., Jamei,
M., Lloyd, R., Pepin, X., Rostami-Hodjegan, A., Sjogren, E., Tannergren, C., Turner,
D.B., Wagner, C., Weitschies, W., Dressman, J., 2013. PBPK models for the
prediction of in vivo performance of oral dosage forms. Eur. J. Pharm. Sci.
Lipinski, C.A., Lombardo, F., Dominy, B.W., Feeney, P.J., 1997. Experimental and
computational approaches to estimate solubility and permeability in drug
discovery and development settings. Adv. Drug Delivery Rev. 23, 325.
Mazer, N.A., Benedek, G.B., Carey, M.C., 1980. Quasielastic light-scattering studies of
aqueous biliary lipid systems. Mixed micelle formation in bile salt-lecithin
solutions. Biochemistry (Moscow) 19, 601615.
Moore, J.W., Flanner, H.H., 1996. Mathematical comparison of dissolution proles.
Pharm. Technol. 20, 6474.
Mudie, D.M., Amidon, G.L., Amidon, G.E., 2010. Physiological parameters for oral
delivery and in vitro testing. Mol. Pharmaceutics 7, 13881405.
Reppas, C., Vertzoni, M., 2012. Biorelevant in-vitro performance testing of orally
administered dosage forms. J. Pharm. Pharmacol. 64, 919930.
Schiller, C., Frohlich, C.P., Giessmann, T., Siegmund, W., Monnikes, H., Hosten, N.,
Weitschies, W., 2005. Intestinal uid volumes and transit of dosage forms as

R. Berthelsen et al. / International Journal of Pharmaceutics 473 (2014) 356365

assessed by magnetic resonance imaging. Alimentary Pharmacol. Ther. 22, 971
979.
Sjgren, E., Westergren, J., Grant, I., Hanisch, G., Lindfors, L., Lennerns, H.,
Abrahamsson, B., Tannergren, C., 2013. In silico predictions of gastrointestinal drug absorption in pharmaceutical product development: application
of the mechanistic absorption model GI-Sim. Eur. J. Pharm. Sci. 49, 679
698.
Sunesen, V.H., Pedersen, B.L., Kristensen, H.G., Mullertz, A., 2005. In vivo in vitro
correlations for a poorly soluble drug, danazol, using the ow-through
dissolution method with biorelevant dissolution media. Eur. J. Pharm. Sci.
24, 305313.
Sutton, S.C., 2009. Role of physiological intestinal water in oral absorption. AAPS J.
11, 277285.
Takano, R., Kataoka, M., Yamashita, S., 2012. Integrating drug permeability with
dissolution prole to develop IVIVC. Biopharm. Drug Dispos. 33, 354
365.

365

The United States Pharmacopeia and National Formulary, 2012. USP 35 - NF 30. USP
35 - NF 30. Rockville MD 20852, The United States Pharmacopeial Covention. US
Pharmacopeia. 6-9-0012. Serial (Book, Monograph).
van, H.P., Liu, X., Fahr, A., 2011. Drug delivery strategies for poorly water-soluble
drugs: the industrial perspective. Expert Opin. Drug Delivery 8, 14811500.
Wagner, C., Jantratid, E., Kesisoglou, F., Vertzoni, M., Reppas, C., Dressman, B., 2012.
Predicting the oral absorption of a poorly soluble, poorly permeable weak base
using biorelevant dissolution and transfer model tests coupled with a
physiologically based pharmacokinetic model. Eur. J. Pharm. Biopharm. 82,
127138.
Williams, H.D., Trevaskis, N.L., Charman, S.A., Shanker, R.M., Charman, W.N., Pouton,
C.W., Porter, C.J.H., 2013. Strategies to address low drug solubility in discovery
and development. Pharmacol. Rev. 65, 315499.
Zhu, T., Ansquer, J.C., Kelly, M.T., Sleep, D.J., Pradhan, R.S., 2010. Comparison of the
gastrointestinal absorption and bioavailability of fenobrate and fenobric acid
in humans. J. Clin. Pharmacol. 50, 914921.