European Journal of Pharmaceutical Sciences 24 (2005) 305–313

In vivo in vitro correlations for a poorly soluble drug,

danazol, using the flow-through dissolution
method with biorelevant dissolution media
Vibeke Hougaard Sunesen, Betty Lomstein Pedersen,
Henning Gjelstrup Kristensen, Anette Müllertz∗
Department of Pharmaceutics, The Danish University of Pharmaceutical Sciences, Universitetsparken 2, DK-2100 Copenhagen Ø, Denmark

Received 25 July 2003; received in revised form 28 September 2004; accepted 17 November 2004
Available online 6 January 2005

Abstract

The purpose of the study was to design dissolution tests that were able to distinguish between the behaviour of danazol under fasted and
fed conditions, by using biorelevant media. In vitro dissolution of 100 mg danazol capsules was performed using the flow-through dissolution
method. Flow rates were 8, 16 or 32 ml/min, corresponding to total volumes dissolution medium of 960, 1920 and 3840 ml, respectively.
The media used contained bile salt and phospholipid levels relevant for either fasted or fed conditions in vivo. Crude and inexpensive bile
components, Porcine Bile Extract and soybean phospholipids, were used as the bile source. The effect of adding different concentrations and
molar ratios of monoglycerides and fatty acids to the fed state media was investigated. In vivo release profiles under fasted and fed conditions
were obtained from a previous study by deconvolution [Sunesen, V.H., Vedelsdal, R., Kristensen, H.G., Christrup, L., Müllertz, A. 2005.
Effect of liquid volume and food intake on the absolute bioavailability of danazol, a poorly soluble drug, Eur. J. Pharm. Sci. 24, 297–303].
In the fasted state, the physiologically most relevant correlation with in vivo results was achieved with a medium containing 6.3 mM bile
salts and 1.25 mM phospholipids (8 ml/min). A medium containing 18.8 mM bile salts, 3.75 mM phospholipids, 4.0 mM monoglycerides and
30 mM fatty acids (8 ml/min) gave the closest correlation with fed state in vivo results. By using the flow-through dissolution method it was
possible to obtain correlations with in vivo release of danazol under fasted and fed conditions. Both hydrodynamics and medium composition
were important for the dissolution of danazol. In the fed state an IVIVC could only be obtained by including monoglycerides and fatty acids
in the medium.
© 2004 Published by Elsevier B.V.

Keywords: In vivo in vitro correlations; Flow-through cell; Dissolution; Biorelevant media; Danazol

1. Introduction In this study, we obtained in vivo in vitro correlations

If proven to be predictive of in vivo behaviour, the disso-
lution test might reduce the need for expensive human bioe-
quivalence studies (Shah and Lesko, 1995). By understand-
ing the relationship between physico-chemical properties of
the drug substance and physiological parameters relevant for
dissolution and absorption it might be possible to define the
most suitable dissolution test conditions.
∗ Corresponding author. Tel.: +45 35 30 64 40; fax: +45 35 30 60 30.
E-mail address: amu@dfuni.dk (A. Müllertz).
(IVIVCs) for a model compound, danazol, under fasted and
fed conditions. Establishment of IVIVCs for danazol was
performed on the basis of in vivo data from a previous study,
under fasted and fed state conditions (Sunesen et al., 2005). A
significant increase in the absolute bioavailability of danazol
when ingested together with a meal was found in the clinical
study, mainly explained by increased solubilization in the
presence of bile and lipolysis products in the small intestine,
and a delayed and prolonged gastric emptying period. This
finding is supported by Charman et al. (1993). The purpose
of the present study was to define appropriate combinations

0928-0987/$ – see front matter © 2004 Published by Elsevier B.V.

doi:10.1016/j.ejps.2004.11.007
306 V.H. Sunesen et al. / European Journal of Pharmaceutical Sciences 24 (2005) 305–313
of dissolution medium composition and hydrodynamics to
simulate the in vivo release of danazol.
Compendial dissolution media often fail to yield IVIVCs
for class II drugs because relevant physiological parameters
are not taken into account. When optimising medium compo-
sition the physiological relevance should always be of prime
consideration. In order to better predict in vivo behaviour,
Galia et al. (1998), introduced two biorelevant media, FaS-
SIF and FeSSIF, representing fasted and fed conditions of the
upper jejunum, respectively. Both media contained physio-
logically relevant concentrations of pure bile salt and phos-
phatidylcholine. However, the fed state medium does not
account for the presence of digested lipids in the intestine
and may lead to underestimates of dissolution for highly
lipophilic drugs (Nicolaides et al., 1999). In an in vitro study,
it was previously shown that the dissolution of danazol from
a capsule formulation was highly dependent on the choice of
medium (Galia et al., 1998). In the present study, the flow-
through dissolution equipment was chosen because it offers
the possibility of maintaining sink conditions throughout an
experiment, and is claimed to simulate in vivo hydrodynam-
ics more closely than the paddle or basket apparatus (Perng
et al., 2003; Qureshi et al., 1994). Since danazol is a neu-
tral compound with a poor aqueous solubility, dissolution is
expected to be much more important in the surfactant-rich
milieu in the small intestine than in the stomach. Therefore,
the work was limited to only using simulated intestinal me-
dia. The gastric residence time observed under in vivo con-
ditions (Sunesen et al., 2005) was taken care of mathemati-
cally.
2. Materials and methods
2.1. Chemicals
Danazol (USP) was obtained from Unikem A/S (Copen-
hagen, Denmark). The particle diameter denoted as the D
(10%), D (50%) and D (90%) was 1.58, 4.46 and 10.2 ␮m,
respectively (Malvern Mastersizer S, Malvern Instruments
Ltd., UK). Lactose monohydrate (Ph. Eur.) was purchased
from Unikem A/S (Copenhagen, Denmark). Hundred mil-
ligrams danazol and 100 mg lactose monohydrate was mixed
and filled in to hard gelatine capsules. The same batch of
Table 1
Composition and solubility of biorelevant dissolution media. Fa and Fe denotes fasted and fed state media, respectively
Water Fa(low) Fa(high) Fe
pH 6.8 6.8 5.6 ± 0.1
KH2 PO4 (mM) 29 29
TRIZMA® maleate (mM) 12
Bile salts (mM) 2.5 6.3 18.8
Phospholipids (mM) 0.5 1.25 3.75
Fatty acids (mM)
Monoglycerides (mM)
Solubility (␮g/ml) 0.42 ± 0.10 7.27 ± 0.89 18.6 ± 1.7 52.6 ± 0.8
danazol capsules was used in the in vivo and in vitro stud-
ies.
Porcine Bile Extract (containing 54–56% bile salt, aver-
age MW 493 g/mol, 0.9% phospholipids), oleic acid, tech.
(90% pure), KH2 PO4 , p.a. (≥99.0% pure), and TRIZMA®
maleate, reagent grade, were supplied by Sigma–Aldrich (St.
Louis, Missouri, USA). Two batches of Porcine Bile Extract
were used in the study. The bile salt content and composition
of each batch were determined by HPLC at Research Center
Foulum, Denmark. Porcine Bile Extract contained primar-
ily glycochenodeoxycholate (43%), glycocholate (25%) and
glycohyocholate (15%). Lecigran M (deoiled soybean phos-
phatidylcholine granules, containing 61% phospholipids, of
this 43% phosphatidylcholine, average MW 782 g/mol) was
a gift from Riceland Foods Inc. (Stuttgart, Arkansas, USA).
Dimodan MO 90/D distilled monoglycerides (1-monoolein,
90% pure) was kindly donated by Danisco A/S (Grindsted,
Denmark).
2.2. Composition of biorelevant media
The composition of the media used for dissolution and
solubility studies is shown in Table 1. Two classes of natu-
rally occurring surfactants were chosen to represent the major
components of human bile: bile salts (BS) and phospholipids
(PL).
Further two intestinal surfactants representing important
lipolytic products were used in the fed state media: fatty acids
(FA) and monoglycerides (MG). To avoid dissolution prob-
lems, BS, PL, FA and MG were dissolved in a small amount
of water by gentle stirring while heating to approximately
40 ◦ C for up to 5 h. When dissolved, the solution was di-
luted to the desired concentration with the appropriate buffer
solution. pH of the fasted state media were adjusted to 6.8.
pH values of the fed state media were approximately 5.5.
For stability reasons, all media were used within 24 h after
preparation.
Adjustment of the osmolality to in vivo values by adding
KCl to the fed state media was not possible due to precip-
itation. Comparative dissolution studies of danazol showed
no difference between fasted state media with and without
KCl addition, and osmolality of the media is presumably not
of great importance for the dissolution of danazol (data not
shown).
Fe(3.75:1) Fe(7.5:2) Fe(15:2) Fe(30:4)
5.6 ± 0.1 5.4 5.6 ± 0.1 5.5 ± 0.4
12 12 12 12
18.8 18.8 18.8 18.8
3.75 3.75 3.75 3.75
3.75 7.5 15 30
1 1 2 4
59.8 ± 1.0 77.3 ± 8.2 76.1 ± 1.9 139 ± 6.3
 V.H. Sunesen et al. / European Journal of Pharmaceutical Sciences 24 (2005) 305–313
2.3. Flow-through dissolution test
The dissolution tests were performed by using the flow-
through dissolution method (USP Apparatus 4). The auto-
mated system (Sotax Dissotest CE70, Sotax, Basel, Switzer-
land) contained a fraction collector and a medium splitter.
A computer controlled the collection of sample fractions at
predetermined intervals (5, 10, 15, 20, 25, 30, 40, 50, 60, 70,
80, 90, 100, 110, 120 min). The equipment was fitted with
22.6 mm cells (internal diameter) at 37 ± 0.1 ◦ C. A ruby ball
(Ø 5 mm) and 1 g of glass beads (Ø 1 mm) were placed in the
bottom of the cone to ensure laminar flow of the jet of fluid en-
tering the cell. The 100 mg danazol capsule was placed on the
top of the glass beads. A tablet holder prevented the capsule
from floating. Undissolved danazol particles were hold back
by a glass fibre filter GD-120 (Advantec, Tokyo Roshi Raisha
Ltd., Japan) in the top of the cone. The flow rate was 8, 16 or
32 ml/min. All experiments were performed in at least four
replicates. Media solutions were degassed with helium dur-
ing dissolution tests. From dissolution tests using fasted state
media 300 ␮l samples were diluted with 300 ␮l methanol.
From fed state dissolution tests 100 ␮l samples were diluted
with 400 ␮l methanol.
2.4. Solubility
Screw cap vials containing 20 ml medium were pre-
heated to 37 ◦ C before adding 10 mg (excess) danazol powder
(n = 3). The samples were gently rotated in a heating chamber
at 37 ◦ C. A 2 ml × 1 ml aliquots were removed after 1, 4, 8 and
24 h and centrifuged at 4500 rpm for 15 min in a preheated
centrifuge (37 ◦ C). The supernatant was filtered through a
0.45 ␮m Minisart RC4 filter (Sartorius AG, Göttingen, Ger-
many). The first seven drops of the filtrate were discarded
to circumvent the initial adsorption of danazol to the filters.
Two hundred microiliters of the filtered solution was diluted
with 300 ␮l methanol. Equilibrium solubility was achieved
when two consecutive determinations of concentration were
within 5% of each other.
2.5. HPLC analysis
Samples from solubility and dissolution experiments were
analysed by a reverse-phase HPLC assay (Naylor et al.,
1995). Detection of danazol was accomplished by ultravi-
olet absorption (285 nm). The reverse-phase column was a
Phenomenex Luna C18-RP, 4.6 mm × 250 mm column with
a Phenomenex Universal C18-RP guard column. The mobile
phase consisted of water:methanol:acetonitrile (30/30/40,
v/v). The flow rate was 2 ml/min and the injection volume
was 10 ␮l. The retention time of danazol was 7 min. Quan-
tification was done by use of external standards. The calibra-
tion curve was linear in the range 0.1–250 ␮g/ml. Inter-day
precision of the HPLC analysis was estimated by calculating
the standard deviation of measurements of the external stan-
dards on separate days. The coefficient of variation was then
307
calculated by dividing the standard deviation of the measure-
ments with the mean value. Calculations were made on the
highest and lowest concentration of external standards used
in the HPLC analysis. The inter-day precision was 8.8 and
3.5% of the lowest and highest standard, respectively. The
accuracy, determined as the average deviation from the the-
oretical concentration, was 6.3 and 2.8% of the highest and
lowest standards, respectively.
2.6. Surface tension
Surface tension was measured in the biorelevant me-
dia at 37 ◦ C by the Wilhelmy plate method (Krüss digital-
tensiometer K10T) (n ≥ 12).
2.7. Dynamic light scattering
The size distribution of micelles and vesicles in the me-
dia was determined by dynamic light scattering (DLS) mea-
surements on media without and saturated with danazol
(DynaPro-801TM , ProteinSolutions, USA) (n ≥ 6). The so-
lutions were filtered through 100 nm filters prior to analy-
sis. Autocorrelation of the collected photons from the mea-
surements gained information about the rate of diffusion of
the particles in the solutions. By using the Stokes–Einstein
equation, the software calculated the corresponding hydro-
dynamic sizes. A regularization algorithm was used to find
the best fitting size distribution of the particles present in the
media.
2.8. In vitro in vivo correlations
Predictive mathematical models describing the relation-
ship between the plasma concentration of danazol and in
vitro dissolution under fasted and fed conditions were devel-
oped using the software PDx-IVIVCTM (GloboMax® LLC,
Slough, Berkshire, UK). Level A IVIVCs, correlating the
entire in vitro and in vivo profiles were obtained by a two-
stage approach. Correlations were performed by using mean
data. First step included numerical deconvolution of the oral
plasma concentration curves by using the intravenous treat-
ment (Sunesen et al., 2005) as a reference, thereby obtaining
an in vivo release profile. Percentage released in vivo was
related to percentage dissolved in vitro by fitting a nonlinear
regression model to the data through an iterative process. The
best fit was determined based upon the mean square error of
the difference between the observed (Sunesen et al., 2005)
and the modeled release.
The model investigated was of the following form (Eq.
(1)) (Shepard and Farrell, 2002):
xvivo (t) = a1 + a2 xvitro (−b1 + b2 t) (1)
where xvivo (t) is the % released in vivo at time (t), a1 al-
lows for a difference between the initial in vitro and in vivo
drug release, a2 allows for a difference between in vitro and
308 V.H. Sunesen et al. / European Journal of Pharmaceutical Sciences 24 (2005) 305–313

in vivo drug release, b1 allows for a time shift between in

vitro and in vivo release, b2 allows for time scaling between
in vitro and in vivo release, and xvitro is the % release in
vitro.
The obtained model was then used to convert the in vitro
dissolution profile to the corresponding predicted bioavail-
ability profile. By using the intravenous treatment as a ref-
erence it was then possible to calculate the predicted in vivo
plasma profile. Finally, Cmax and AUC were calculated for
the predicted in vivo plasma profile. The predictability of
the developed model was evaluated by comparing predicted
and observed values of Cmax and AUC. For a reasonable
IVIVC, mean absolute prediction errors (PE) for Cmax and
Fig. 1. Mean deconvolution profiles of danazol after administration of
AUC should not exceed 10%. 100 mg capsules to healthy male volunteers under fasted and fed state con-
ditions (n = 8) (Sunesen et al., 2005).

3. Results relations will be relevant although the correlation is based on

3.1. Deconvolution profiles
The overall purpose of the present study was to develop
dissolution methods that can adequately predict the in vivo
behaviour of danazol under fasted and fed conditions, based
on results from a clinical study in healthy male subjects
(Sunesen et al., 2005). In the fasted state, danazol was ad-
ministered with 200 ml water. Under fed state conditions,
danazol was administered together with a lipid-rich meal. In
order to obtain in vivo release profiles, numerical deconvolu-
tion was performed (PDx-IVIVCTM ). Since the results from
Sunesen et al. (2005), indicate that first-pass metabolism of
danazol in the liver may be considerable, the calculated de-
convolution profiles show the bioavailability of danazol and
will not reflect the exact dissolution of the compound. How-
ever, if permeability is not limiting, and first-pass metabolism
is assumed to be linear, establishment of in vivo in vitro cor-
bioavailability rather than dissolution.
Deconvolution was done on mean data, using the intra-
venous treatment as the impulse response function. Fig. 1
shows that the bioavailability reached 13% of the dose when
danazol was administered in the fasted state. When adminis-
tered in the fed state, the bioavailability of danazol reached
46% of the dose.
3.2. In vitro dissolution
Two media were compared in order to cover the inter-
val of BS concentrations observed under in vivo fasted state
conditions. Resultant in vitro dissolution curves are shown
in Fig. 2. By using water as the dissolution medium at a
flow rate of 8 ml/min, less than 1% of the danazol dose was
dissolved after 2 h. A three-fold increase in total amount dis-
solved after 2 h was achieved by adding low levels of BS
and PL (Fa(low), 8 ml/min). Dissolution reached 15 ± 0.5%

Fig. 2. In vitro dissolution of 100 mg danazol capsules in media containing bile salts and phospholipids (mean ± S.D., n ≥ 4).
 V.H. Sunesen et al. / European Journal of Pharmaceutical Sciences 24 (2005) 305–313 309

Fig. 3. In vitro dissolution of 100 mg danazol capsules in media containing bile salts, phospholipids, fatty acids and monoglycerides (mean ± S.D., n ≥ 5).

after 2 h when a high fasted level of BS and PL was used
(Fa(high), 8 ml/min). Similar dissolution profiles were ob-
tained for Fa(high) at a flow rate of 8 ml/min, and for Fa(low)
at a flow rate of 32 ml/min.
By increasing the concentration of BS and PL to fed
state levels (Fig. 2), dissolution reached 28 ± 4.3% (Fe,
8 ml/min) after 2 h. Addition of FA and MG increased the
total amount dissolved after 2 h to 32 ± 6.6% (Fe(3.75:1),
8 ml/min), 45 ± 3.9% (Fe(15:2), 8 ml/min) and 49 ± 3.7%
(Fe(30:4), 8 ml/min) (shown in Fig. 3).
Concerning the curvature of the plots, rather smooth dis-
solution curves were obtained in media containing only BS
and PL, displaying a continuous decrease in dissolution rate
(Fig. 2). Addition of FA and MG to the media introduced
a slight bend in the dissolution curve after approximately
30 min, and dissolution levelled off more rapidly than in me-
dia without FA and MG. The change in dissolution rate was
particularly distinct for media containing high concentrations
of FA and MG.
3.3. Solubility
centration of surfactants in the media (BS + PL + FA + MG)
(r2 = 0.98). The equilibrium solubility of danazol was
achieved after 1 h in all media. Under all conditions tested,
the concentration of danazol in the samples reached its sat-
uration level after 10–15 min. In the remaining part of the
dissolution profiles, samples were unsaturated with danazol.
However, sink conditions (<20% of the saturation concen-
tration) were never achieved in media containing exclusively
BS and PL, and only prevailed in the four-component media
(BS, PL, FA, MG) in the last part of the dissolution curves.
3.4. Surface tension
Surface tension values of the media are shown in Table 2.
The critical micellar concentration (CMC) was estimated
by surface tension measurements on fasted media contain-
ing decreasing concentrations of BS and PL (molar ratio
5:1) in 29 mM KH2 PO2 (pH 6.8). CMC was below 0.5 mM
BS, which means that all media contained amphiphile levels
above their CMC.

The solubility of danazol was increased more than 300 3.5. Dynamic light scattering
times in media containing the highest concentration of am-
phiphiles, Fe(30:4), as compared to water (Table 1). Danazol The particle size measurements are presented in Table 2.
solubility was found to increase linearly with the total con- Fa(low) and Fe(3.75:1) were characterized by one single par-
Table 2
Characterization of biorelevant dissolution media with regard to surface tension and particle size (mean ± S.D.)
Media Surface tension (mN/m) (n ≥ 12) Particle sizea Rh (nm) (n ≥ 6)

Without danazol Saturated with danazol

Fa(low) 36.1 ± 0.5 95 ± 14 106 ± 5.4
Fa(high) 34.8 ± 1.1 65 ± 19 212 ± 61 65 ± 25 204 ± 54
Fe 30.1 ± 0.4 12 ± 2.6 117 ± 9.5 18 ± 2.0 195 ± 26
Fe(3.75:1) 29.2 ± 0.5 19 ± 3.2 22 ± 6.1
Fe(15:2) 28.5 ± 0.3 33 ± 17 131 ± 23 17 ± 3.8 108 ± 9.4
Fe(30:4) 25.0 ± 2.2 45 ± 18 146 ± 37 46 ± 24 152 ± 49
a In all cases, the smallest particle size was the dominating, and constituted on average between 74 and 99% of the mass. Only particles constituting >1% of

the mass were considered significant in the particle size distribution.

310 V.H. Sunesen et al. / European Journal of Pharmaceutical Sciences 24 (2005) 305–313

Table 3
Internal predictability of the IVIVC models, calculated by use of the software, PDx-IVIVCTM , using the best fit to Eq. (1)
In vitro dissolution In vivo treatment Cmax (ng/ml) AUC (h ng/ml)
Obs. Pred. PE (%) Obs. Pred. PE (%)
Fa(low), 8 ml/min Fasted 26.1 29.2 12 142 95.9 33
Fa(low), 16 ml/min Fasted 26.1 27.2 4.2 142 126 12
Fa(low), 32 ml/min Fasted 26.1 24.3 6.6 142 129 9.0
Fa(high), 8 ml/min Fasted 26.1 25.8 0.8 142 131 7.9
Fe, 8 ml/min Fasted 26.1 31.0 19 142 154 8.6
Fe, 8 ml/min Fed 68.2 54.6 20 558 364 35
Fe, 16 ml/min Fed 68.2 50.9 25 558 429 23
Fe(3.75:1), 8 ml/min Fed 68.2 57.1 16 558 386 31
Fe(3.75:1), 16 ml/min Fed 68.2 60.2 12 558 423 24
Fe(3.75:1), 32 ml/min Fed 68.2 66.8 2.1 558 453 19
Fe(15:2), 8 ml/min Fed 68.2 55.9 18 558 459 18
Fe(30:4), 8 ml/min Fed 68.2 63.2 7.3 558 497 11
Observed (Obs.) and predicted (Pred.) values of Cmax and AUC are shown. PE (%) denotes the percentage predction error.

ticle size, whereas the remaining media contained two par- Table 4
ticle sizes. Only particles making up >1% of the mass were Parameter estimates obtained in the nonlinear regression model for the dis-
solution media yielding good in vivo in vitro correlations
considered significant. The media contained particles with
hydrodynamic radii (Rh ) in the range 12–212 nm. The par- Dissolution test Simulation a1 a2 b1 b2
conditions
ticle size decreased when increasing the surfactant concen-
tration in media containing BS and PL alone. By addition Fa(low), 32 ml/min Fasted −0.607 1 0.194 0.343
Fa(high), 8 ml/min Fasted −0.507 1 0.251 0.321
of FA and MG to the fed state media a progressive increase Fe(30:4), 8 ml/min Fed −4.463 1 −0.173 0.139
in particle size was observed. No significant change in par-
ticle size was observed upon saturation of the media with
Since the endpoint of the in vitro dissolution profiles approx-
danazol, except in the Fe medium where the particle sizes
imates the in vivo bioavailability, the relative bioavailability
increased.
of the oral treatment is masked by the incomplete dissolution.
Thus, it is reasonable to use the default value of a2 = 1 in the
3.6. In vitro in vivo correlations model. The quality of the resultant nonlinear IVIVC model
was determined by comparing observed and predicted values
PDx-IVIVCTM , using the best fit to Eq. (1), performed of Cmax and AUC. A satisfactory internal predictability had
mathematical modelling to obtain a correlation between in prediction errors (PE) lower than 10% (Table 3). Satisfactory
vivo bioavailability and in vitro dissolution. Due to the dif- IVIVCs under fasted state conditions were obtained by using
ferent time scaling in vivo and in vitro, b2 was important Fa(high) at a flow rate of 8 ml/min, or by using Fa(low) at a
to achieve the best fit. Further, a1 and b1 were estimated. flow rate of 32 ml/min. All the tested in vitro dissolution con-
ditions under fed state conditions exceeded the 10% limits of
PE. However, the Fe(30:4) medium, containing 18.8 mM BS,
3.75 mM PL, 30 mM FA and 4.0 mM MG, was considered to
yield a satisfactory correlation with in vivo results. It yielded
7.3% prediction error of Cmax , and 11% prediction error of
AUC. Observed and predicted plasma concentration curves
are shown in Fig. 4. Parameter estimates obtained in the linear
regression model are shown in Table 4.

4. Discussion

Based on the results from the clinical study (Sunesen

Fig. 4. Observed and predicted plasma concentration profiles of danazol in
the fasted and fed states.
et al., 2005), and from current knowledge of the dissolu-
tion behaviour of poorly soluble lipophilic drugs in biorele-
vant dissolution media (Bakatselou et al., 1991; Naylor et al.,
1993, 1995; Shah et al., 1989, 1995; Dressman et al., 1998),
it was the intention to define appropriate combinations of
medium composition and hydrodynamics to obtain IVIVCs
by using the flow-through apparatus.
 V.H. Sunesen et al. / European Journal of Pharmaceutical Sciences 24 (2005) 305–313
4.1. Medium composition
Bile salts and phosphatidylcholine were chosen to repre-
sent the major components of human bile. Due to the high
cost of pure bile salts and phoshpatidylcholine, and the huge
amounts needed to perform flow-through dissolution exper-
iments, Porcine Bile Extract and a mixture of soybean phos-
pholipids were chosen as the bile source. Porcine Bile Ex-
tract is less expensive than pure bile salts but needs to be
assayed for the exact amount of bile salts in each batch be-
fore use, which may be a rather time consuming procedure.
Porcine bile was considered a reasonable choice, since the
composition was similar to human bile with regard to the
level of the two major bile salts, glycochenodeoxycholate
and glycocholate. Based on data from Schersten (1973), a
molar ratio of BS:PL (5:1) was chosen in both fasted and
fed state media. A low and a high level of bile salts were
used to cover the range of fasted state bile salt concentra-
tions (Lindahl et al., 1997). In all fed state media 18.8 mM
BS levels were used, based on results from Hernell et al.
(1990), Rautureau et al. (1981). Further, two intestinal sur-
factants were used in the fed state media to represent im-
portant lipolytic products originating from the digestion of
dietary lipids: fatty acids (FA) and monoglycerides (MG).
Long-chain lipids, oleic acid and 1-monoolein were chosen
as the FA and MG sources, respectively. Levels were selected
on the basis of measurements of human postprandial duode-
nal content (Hernell et al., 1990) and from in vitro lipolysis
results obtained by Zangenberg et al. (2001). It was the in-
tention to investigate the effect of adding different amounts
and molar ratios of FA and MG in a physiologically relevant
magnitude.
4.2. Hydrodynamics
Influence of hydrodynamics on the dissolution of dana-
zol was investigated by measuring dissolution at different
flow rates. The pump used in the flow-through apparatus had
a delivery range between 8 and 50 ml/min. Flow rates used
in the present study were 8, 16 and 32 ml/min, giving ax-
ial velocities in the flow-through cells (22.6 mm diameter)
of 2, 4 or 8 cm/min, respectively. From the in vivo part of
the study the small intestinal transit times were found to be
241 ± 58 min in the fasted state and 268 ± 106 min in the fed
state (Sunesen et al., 2005). Assuming that the total length
of the small intestine is 350 cm (Ho et al., 1983), the esti-
mated average axial velocity was 1.5 and 1.3 cm/min under
fasted and fed state conditions, respectively. Since the flow-
through dissolution apparatus is only able to simulate axial
drug transport it is reasonable to compensate for the lack of
radial drug transport by using higher flow rates than observed
in vivo.
The use of higher flow rates in vitro than observed in vivo
was expected to speed up the in vitro dissolution process as
compared to in vivo release. Thus, in vitro dissolution was
expected to run ahead of in vivo release which is confirmed by
311
the values of the obtained time-scaling parameters (b2 ). The
values of b2 (Table 4) were below 1 for all the test conditions
giving reasonable IVIVCs.
In order to limit the total volume of dissolution medium
to a magnitude relevant for physiological conditions, the in
vitro dissolution tests were only running for 120 min. Flow
rates of 8, 16 and 32 ml/min corresponded to total volumes of
960, 1920 and 3840 ml, respectively, during a single dissolu-
tion test. Apart from the water administered together with the
dosage form, the volume of liquid in the fasted small intes-
tine is between 120 and 350 ml (Dillard et al., 1965). Under
fed state conditions the volume in the upper small intestine is
highly variable comprising volumes up to 1600 ml (Fordtran
and Lochlear, 1966). Thus, a flow rate of 8 ml/min may be
considered the most appropriate choice of dissolution condi-
tions to simulate in vivo bioavailability of danazol.
4.3. In vivo in vitro correlations
Under fasted state conditions, two combinations of
medium composition and flow rate yielded almost identical
dissolution profiles and IVIVCs complying with the predic-
tion limits, Fa(low), 32 ml/min and Fa(high), 8 ml/min. Under
the conditions used, a four-fold increase in flow rate had the
same effect on the dissolution profile as a 2.5 times increase
in the BS and PL level. Apparently, it is possible to manip-
ulate the dissolution results, and other combinations of flow
rate and concentration of BS and PL in the molar ratio (5:1)
might as well yield IVIVCs. The Fa(high) medium had a sur-
face tension (34.8 ± 1.1 mN/m) very close to the mean value
determined by Pedersen et al. (2000), in fasted human intesti-
nal fluids (33.7 ± 2.8 mN/m). Further, the lower flow rate was
more physiologically relevant, so Fa(high) run at a flow rate
of 8 ml/min may be considered the most appropriate choice
of dissolution conditions to simulate in vivo bioavailability
of danazol in the fasted state.
The best predictability under fed state conditions was
achieved by using a medium containing all four compo-
nents, Fe(30:4). This combination of surfactants yielded a
dissolution curve with the highest initial dissolution rate
of the media investigated, displaying a curvature compara-
ble to the bioavailability profile. The surface tension was
25.0 ± 2.2 mN/m. This is in close agreement with the results
obtained by Luner and Vander Kamp (2001). They estimated
a surface tension value of approximately 25 mN/m of the
fed state upper gastro-intestinal tract, based on compositional
analysis. Different amphiphiles contribute differently to the
overall surface tension. The combination of amphiphiles in
the above-mentioned fasted and fed media seem to be phys-
iologically relevant with regard to surface tension. In con-
trast, the surface tensions of FaSSIF and FeSSIF media were
rather high, 50 and 37 mN/m, respectively (Luner and Vander
Kamp, 2001). Since the inclusion of lipid digestion products
in the fed state dissolution medium was a prerequisite for
obtaining an IVIVC, the use of FeSSIF may lead to underes-
timates of fed state dissolution behaviour, at least for danazol.
312 V.H. Sunesen et al. / European Journal of Pharmaceutical Sciences 24 (2005) 305–313
4.4. Solubility
The solubility of danazol in the media was found to be
the most important determinant of the initial dissolution rate.
Furthermore, the solubility enhancement of danazol was ap-
parently independent of the type of amphiphile added to the
media, which is in agreement with Zangenberg et al. (2001).
On a molar basis, an increase in amphiphile concentration
provided an equivalent increase in danazol solubility irre-
spective of the type of amphiphile used (e.g. BS, PL, TG or
FA). This implies that equivalent solubilities of danazol can
be obtained in media containing exclusively BS and PL, and
media containing all four components. However, the type of
amphiphile included in the media had a marked influence on
the curvature of the dissolution plots. Thus, different disso-
lution media providing the same solubility of danazol, but
containing different types of amphiphiles, are expected to
yield different dissolution profiles. To date, the possibility of
developing an IVIVC has been based on solubility estima-
tions only (e.g. the Biopharmaceutics Classification System,
Amidon et al., 1995).
4.5. Colloidal species
The significant solubility enhancement of danazol in the
presence of surfactants indicates that danazol was extensively
solubilized by the colloidal species present in the media,
which is in agreement with observations by Bakatselou et
al. (1991). During lipid digestion, lipolytic products includ-
ing FA and MG interact with BS and PL to form mixed mi-
celles (Rh 3–20 nm) and vesicles (Rh 40–60 nm). Depending
on the nature of the lipids being digested, additional col-
loidal species such as microemulsions (Rh 10–200 nm) might
as well be produced (Carey and Small, 1970; Cistola et al.,
1988). For BS/PL systems containing long-chain lipid diges-
tion products (lipid:BS > 1), vesicles and mixed micelles are
expected to coexist (Staggers et al., 1990). Only in the case of
Fe(15:2) and Fe(30:4), was the ratio of lipid:BS > 1. However,
these media contained aggregates in the vesicular and mi-
croemulsion size ranges, with no presence of mixed micelles.
Kossena et al. (2003), investigated media similar to Fe(30:4)
with regard to composition and danazol solubility by using
pure bile components and lipolysis products, and found Rh
of 4.1 and 70 nm. Aggregates in Fe(30:4) had Rh of 45 ± 18
and 146 ± 37 nm. A medium similar to Fa(high), but with
a slightly lower concentration of BS, contained exclusively
mixed micelles with hydrodynamic radius, 3.5 nm (Kossena
et al., 2003). In the present study, the smallest particle size
in Fa(high) was 65 ± 25 nm. The lack of small particles in
general in the media investigated may be due to limitations
in the regularization algorithm to adequately distinguish be-
tween particle sizes and/or the complexity of the media used.
Soy phospholipids were used as a source of phospholipids.
The FA composition of soy phospholipids differs from that
of bile phosphatidylcholine by a higher content of unsatu-
rated FA (oleic and linoleic acid). This may affect the phase
behaviour of the formed phases. Furthermore, the fact that
the media were filtered through 100 nm filters before mea-
surement of the size of the colloidal species may affect the
size distribution to some extent. The pattern of particle size
changes, at increasing concentrations of bile salt and phos-
pholipids, with or without lipid digestion products, has been
observed previously (unpublished data). The solubility and
thereby the dissolution rate was apparently not influenced by
the large particle sizes, suggesting that the payload of dana-
zol per diffusing particle was sufficient to offset the reduced
diffusivity.
In conclusion, it was possible to obtain IVIVCs for danazol
in the fasted and fed state by using the flow-through dissolu-
tion apparatus. Inexpensive crude bile components, of these
Porcine Bile Extract and soybean phospholipids, proved to
be valuable in obtaining IVIVCs, and questions the need for
pure bile salts and lecithin for drug development issues. Ad-
dition of lipid digestion products, fatty acids and monoglyc-
erides, was necessary for achievement of IVIVCs under fed
state conditions. It was possible to manipulate the dissolution
results by appropriate combinations of test conditions. The
initial dissolution rate of danazol was highly dependent on
the flow rate and the solubility of danazol in the dissolution
media. Further studies are needed to determine the discrim-
inatory power of the dissolution tests with regard to other
batches, generic products, manufacturing methods, etc.
Acknowledgements
Vibeke Hougaard Sunesen was the recipient of a bur-
sary from the Danish Medical Research Council, Centre for
Drug Design and Transport. The authors want to thank Irene
Klausen for technical assistance.
References
Amidon, G.L., Lennernäs, H., Shah, V.P., Crison, J.R., 1995. A theoretical
basis for a biopharmaceutical drug classification: the correlation of in
vitro drug product dissolution and in vivo bioavailability. Pharm. Res.
12, 413–420.
Bakatselou, V., Oppenheim, R.C., Dressman, J.B., 1991. Solubilization
and wetting effects of bile salts on the dissolution of steroids. Pharm.
Res. 8, 1461–1469.
Carey, M.C., Small, D.M., 1970. The characteristics of mixed micellar
solutions with particular reference to bile. Am. J. Med. 49, 590–608.
Charman, W.N., Rogge, M.C., Boddy, A.W., Berger, B.M., 1993. Ef-
fect of food and a monoglyceride emulsion formulation on danazol
bioavailability. J. Clin. Pharmacol. 33, 381–386.
Cistola, D.P., Hamilton, J.A., Jackson, D., Small, D.M., 1988. Ionization
and phase behavior of fatty acids in water: application of the Gibbs
phase rule. Biochemistry 27, 1881–1888.
Dillard, R.L., Eastman, H., Fordtran, J.S., 1965. Volume-flow relation-
ship during the transport of fluid through the human small intestine.
Gastroenterology 49, 58–66.
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, 11–22.
 V.H. Sunesen et al. / European Journal of Pharmaceutical Sciences 24 (2005) 305–313
Fordtran, J.S., Lochlear, T.W., 1966. Ionic constituents and osmolality of
gastric and small intestinal fluids after eating. Am. J. Dig. Dis. 11,
503–521.
Galia, E., Nicolaides, E., Hörter, D., Löbenberg, R., Reppas, C., Dress-
man, J.B., 1998. Evaluation of various dissolution media for predict-
ing in vivo performance of Classes I and II drugs. Pharm. Res. 15,
698–705.
Hernell, O., Staggers, J.E., Carey, M.C., 1990. Physical-chemical behavior
of dietary and biliary lipids during intestinal digestion and absorption.
2. Phase analysis and aggregation states of luminal lipids during duo-
denal fat digestion in healthy adult human beings. Biochemistry 29,
2041–2056.
Ho, N.F.H., Park, J.Y., Ni, P.F., Higuchi, W.I., 1983. Advanced quanti-
tative and mechanistic approaches in interfacing gastrointestinal drug
absorption studies in animals and humans. In: Crouthanel, W., Sarapu,
A.C. (Eds.), Animal Models for Oral Drug Delivery in Man: In
Situ and In Vivo Approaches. American Pharmaceutical Association,
Washington, DC, pp. 27–106.
Kossena, G.A., Boyd, B.J., Porter, C.J.H., Charman, W.N., 2003. Separa-
tion and characterization of the colloidal phases produced on digestion
of common formulation lipids and assessment of their impact on the
apparent solubility of selected poorly water-soluble drugs. J. Pharm.
Sci. 92, 634–648.
Lindahl, A., Ungell, A.-L., Knutson, L., Lennernäs, H., 1997. Character-
ization of fluids from the stomach and proximal jejunum in men and
women. Pharm. Res. 14, 497–502.
Luner, P.E., Vander Kamp, D., 2001. Wetting behavior of bile salt-lipid
dispersions and dissolution media patterned after intestinal fluids. J.
Pharm. Sci. 90, 348–359.
Naylor, L.J., Bakatselou, V., Dressman, J.B., 1993. Comparison of the
mechanism of dissolution of hydrocortisone in simple and mixed mi-
celle systems. Pharm. Res. 10, 865–870.
Naylor, L.J., Bakatselou, V., Rodriguez-Hornedo, N., Weiner, N.D., Dress-
man, J.B., 1995. Dissolution of steroids in bile salt solutions is modi-
fied by the presence of lecithin. Eur. J. Pharm. Biopharm. 41, 346–353.
Nicolaides, E., Galia, E., Efthymiopoulos, C., Dressman, J.B., Reppas,
C., 1999. Forecasting the in vivo performance of four low solubility
drugs from their in vitro dissolution data. Pharm. Res. 16, 1876–1882.
Pedersen, B.L., Müllertz, A., Brøndsted, H., Kristensen, H.G., 2000. A
comparison of the solubility of danazol in human and simulated gas-
trointestinal fluids. Pharm. Res. 17, 891–894.
313
Perng, C.-Y., Kearney, A.S., Palepu, N.R., Smith, B.R., Azzarano, L.M.,
2003. Assessment of oral bioavailability enhancing approaches for
SB-247083 using flow-through cell dissolution testing as one of the
screens. Int. J. Pharm. 250, 147–156.
Qureshi, S.A., Caillé, G., Brien, R., Piccirilli, G., Yu, V., McGilveray,
I.J., 1994. Application of flow-through dissolution method for the
evaluation of oral formulations of nifedipine. Drug Dev. Ind. Pharm.
20, 1869–1882.
Rautureau, M., Bisalli, A., Rambaud, J.C., 1981. Etude de la phase aque-
use intra-jejunale des sels biliaires et des lipides au cours de la di-
gestion d’un repas standard chez le sujet normal. Gastroenterol. Clin.
Biol. 5, 417–425.
Schersten, T., 1973. Formation of lithogenic bile in man. Digestion 9,
540–553.
Shah, V.P., Konecny, J.J., Everett, R.L., McCullough, B., Noorizadeh,
A.C., Skelly, J.P., 1989. In vitro dissolution profile of water-insoluble
drug dosage forms in the presence of surfactants. Pharm. Res. 6,
612–618.
Shah, V.P., Lesko, L.J., 1995. Current challenges and future regula-
tory directions in in vitro dissolution. Drug Inform. J. 29, 885–
891.
Shepard, T., Farrell, C., 2002. A theoretical evaluation of the PDx-IVIVC
equation: Applicability and limitations. In: Poster presented by Globo-
max at the 2002 AAPS Annual Meeting and Exposition, Toronto,
Canada.
Staggers, J.E., Hernell, O., Stafford, R.J., Carey, M.C., 1990. Physical-
chemical behavior of dietary and biliary lipids during intesti-
nal digestion and absorption. 1. Phase behavior and aggregation
states of model lipid systems patterned after aqueous duodenal
contents of healthy adult human beings. Biochemistry 29, 228–
240.
Sunesen, V.H., Vedelsdal, R., Kristensen, H.G., Christrup, L., Müllertz,
A. 2005. Effect of liquid volume and food intake on the absolute
bioavailability of danazol, a poorly soluble drug. Eur. J. Pharm. Sci.
24, 297–303.
USP, 2004. The United States Pharmacopeia/The National Formu-
lary. United States Pharmacopeial Convention Inc., Rockville,
USA.
Zangenberg, N.H., Müllertz, A., Kristensen, H.G., Hovgaard, L., 2001. A
dynamic in vitro lipolysis model. II. Evaluation of the model. Eur. J.
Pharm. Sci. 14, 237–244.