e u r o p e a n j o u r n a l o f p h a r m a c e u t i c a l s c i e n c e s 2 7 ( 2 0 0 6 ) 9199

available at www.sciencedirect.com

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

A strategy for preclinical formulation development using

GastroPlusTM as pharmacokinetic simulation tool and a
statistical screening design applied to a dog study

Martin Kuentz , Sonja Nick, Neil Parrott, Dieter Rothlisberger

F. Hoffmann-La Roche Ltd., Pharmaceutical and Analytical R&D, Bldg./Lab. 072/338, Grenzacherstr., CH-4070 Basel, Switzerland

a r t i c l e

i n f o

a b s t r a c t

Article history:

The aim of this paper is to propose a pharmaceutical risk assessment strategy that goes

Received 21 February 2005

beyond the usual characterisation of a clinical candidate molecule according to the bio-

Received in revised form 11 August

pharmaceutical classication system (BCS). This strategy was evaluated for a new CNS drug

2005

with poor solubility and good permeability. In a rst step, GastroPlusTM was used to simulate

Accepted 20 August 2005

the absorption process based on preformulation data. These input data involved a physico-

Available online 10 October 2005

chemical drug characterisation including drug solubility measurements in simulated physiological media, as well as permeability determination. Further computer simulations were

Keywords:

conducted to determine the sensitivity to changes of selected input values. Thus, oral

Simulation

bioavailability prediction was studied as a function of the particle size and drug solubility.

Bioavailability

The second part of the presented strategy for preclinical formulation development was to

Dog studies

test specially designed formulations in a 23 screening factorial plan using the dog as the

Factorial design

animal model. The factors were the dosage form, food effect and dose strength. One of the

Clinical formulation

two experimental formulations was a capsule lled with the micronised drug, whereas the
other formulation was a surfactant solution of the drug. Accordingly, a worst case formulation was compared with a best case drug solution over the clinically relevant dose range
in fasted and fed dogs. The results of the computer simulation indicated that a fraction of
the dose is dissolved in the stomach and precipitates partially in the small intestine. The
simulation predicted almost full drug absorption during the GI transit time. Interestingly,
the simulation implies that stomach drug solubility had little impact on overall fraction
absorbed. The results also showed that changes of particle size and reference solubility
within two orders of magnitude hardly affected the oral bioavailability. This in silico deduction was subsequently compared with the results of the dog studies. Indeed a surfactant drug
solution showed no clear biopharmaceutical superiority over a solid capsule formulation on
the average of both dose strengths in fasted and fed dogs. Despite the substantial variability
of the in vivo data, the factorial screening design indicated marginal signicant interaction
between the dose level and feeding status. This can be viewed as a ag for the planning of
further studies, since a potential effect of one factor may depend on the level of the other.
In summary, the GastroPlusTM simulation together with the statistically designed dog study
provided a thorough biopharmaceutical assessment of the new CNS drug. Based on these
ndings, it was decided to develop a standard granulate in capsules for phase I studies.
More sophisticated formulation options were abandoned and so the clinical formulation
development was conducted in a cost-efcient way.
2005 Elsevier B.V. All rights reserved.

Corresponding author. Tel.: +41 61 688 38 70; fax: +41 61 688 86 89.
E-mail address: martin.kuentz@roche.com (M. Kuentz).
0928-0987/$ see front matter 2005 Elsevier B.V. All rights reserved.
doi:10.1016/j.ejps.2005.08.011

92

1.

e u r o p e a n j o u r n a l o f p h a r m a c e u t i c a l s c i e n c e s 2 7 ( 2 0 0 6 ) 9199

Introduction

A biopharmaceutical assessment of drug substances is crucial

for different phases of the development process. In an early
phase, pharmaceutical proling should help to rate candidate
molecules in terms of their druglike properties (Balbach and
Korn, 2004; Kerns and Di, 2003). The pivotal challenge is to nd
an appropriate molecule for preclinical and clinical development. Herein a drug categorisation according to the biopharmaceutical classication system (BCS) (Amidon et al., 1995) is
helpful. Once the candidate selection is made, the BCS concept is also valuable in view of project planning. The choice
of an adequate clinical formulation scenario is difcult, as it
has to meet the general project timelines and must take the
involved risks into consideration. Those risks can be considerable with biopharmaceutically challenging drugs like those in
BCS class II, III or IV. A straightforward technical development
could lead to a formulation that does not show adequate exposure in humans. On the other hand, de-risking, in the form
of parallel development of sophisticated formulations, is cost
intensive. This is especially the case for bioavailability testing in animals. An extensive formulation screening would not
only bind substantial resources, but is also uncertain in terms
of its relevance for humans.
This article presents an additional biopharmaceutical
assessment program in the preclinical development phase.

Fig. 1 shows when this assessment program should be conducted in the framework of other formulation activities. The
assessment directly affects the choice of the clinical development concepts. Depending on the development strategy one
can assign meaningful development resources about 1 year
before the rst human is dosed.
The biopharmaceutical assessment program consists of
two steps. The rst step is a computer simulation that should
help to elucidate the hurdles for oral bioavailability, as well as
identify critical parameters of a drug formulation. The simulation may also reveal some parameters that only have a limited
impact on the biopharmaceutical performance of a formulation, which is of equal importance. This rst step is only an
in silico model, but the physiology of the human situation is
simulated. The second part of the biopharmaceutical assessment is to test proof of concept formulations in statistically
designed pharmacokinetic experiments. This second step has
the limitations of the animal model, but this time in vivo data
are generated.
The eld of computer simulations with physiologically
based pharmacokinetic absorption models has been covered
in several articles (Yokoe et al., 2003; Agoram et al., 2001; Norris et al., 2000; Plusquellec et al., 1999). Recently, Dannenfelser
et al. (2004) reported the application of a computer simulation
for a clinical dosage form development. This prots from a better quality of input data in comparison to the discovery phase
since a rst pharmaceutical proling including BCS classica-

Fig. 1 Gantt chart of the relevant formulation development activities including the new additional biopharmaceutical
assessment program.

93

e u r o p e a n j o u r n a l o f p h a r m a c e u t i c a l s c i e n c e s 2 7 ( 2 0 0 6 ) 9199

tion is normally available. The underlying model in Dannenfelsers work as well as in the present paper goes back to the
compartmental absorption and transit model (CAT) (Yu et al.,
1996a,b) a more versatile model than the single-tank mixing
model (Sinko et al., 1991). The CAT model predicts absorption
through the intestinal tract and takes into account the ow
of the drug through the digestive tract, which is divided into
a set of compartments. The model was further developed to
the so-called advanced CAT (ACAT) model (Agoram et al., 2001)
and implemented in a commercially available software package called GastroPlusTM .
Based on GastroPlusTM computer simulations a biopharmaceutical evaluation can be made as to which factors could
have an impact on the oral bioavailability. The results should
lead to the design of experimental formulations. This second step of the proposed strategy should roughly assess the
difference between a good formulation and a rather poor
formulation in vivo using the dog as pharmacokinetic model.
A recent paper emphasizes the use of design of experiment
techniques (DoE) to gain maximal information from a minimal number of animals (Kuentz et al., 2003). It allows in the
present case to study the effect of the galenical formulation
on pharmacokinetic parameters depending on the dose and
feeding status of the animals. The results of using both, computer simulation and DoE in preclinical bioavailability testing
are discussed with special emphasis on further planning of
the clinical formulation activities.

2.

Materials and methods

2.1.

Characterization of the drug substance

R1315 (MW; 409.41 g/mol) is a hydrogen sulphate with a pKa

value of 5.9, and a c log P of 5.5 (log D7.4 of 4.9). The melting
point (DSC determination at 5 C/min) displayed an onset of
241.2 C and a peak at 245.4 C. The solubility of the drug was
in aqueous systems overall very low (<1 g/mL at pH values
higher than 5). The drug solubility was also tested in physiologically relevant media (Galia et al., 1998; Dressman et al.,
1998; Ingels and Augustijns, 2003). The solubility was highest with 0.2 mg/mL at room temperature in simulated gastric
uid (SGF at pH 1.2), whereas in fed simulated standard vehicle (FeSSIF) 0.1 mg/mL were dissolved at pH 5 and 0.05 mg/mL
at pH 6.5. In fasted simulated standard vehicle (FaSSIF at pH
6.5) only 0.002 mg/mL could go into solution. These solubility
values were determined by HPLC after equilibration for 3 h at
room temperature.
The drug exhibited supersaturated solutions in presence of mixed micelles. In FeSSIF at pH 5 the initial solubility was about 0.6 mg/mL, whereas in FeSSIF at pH 6.5
roughly 0.3 mg/mL were dissolved in the beginning. The
initial kinetic solubility is hence approximately six times
higher than the equilibrium value in FeSSIF. Both solutions showed after 30 min a bit less than half of the initial
solubility.
The drug compound was determined as highly permeable
in the Parallel Articial Permeability Assay (PAMPA) (Kansy
et al., 1998) and was rated as a BCS class II compound.

Table 1 Fractional 252 design for screening of

excipients in view of R1315 drug solubilisation
No.

Cremophor
RH40 (%)

Tween
20 (%)

Solutol
HS 15
(%)

PVP
K30
(%)

1
2
3
4
5
6
7
8
9
10
11

3.5
7
3.5
0
0
3.5
7
7
0
7
0

3.5
0
3.5
7
0
3.5
0
7
0
7
7

3.5
7
3.5
0
7
3.5
0
0
0
7
7

2.5
0
2.5
0
5
2.5
0
5
5
5
0

Propylene
glycol (%)
5
10
5
10
0
5
0
0
10
10
0

2.2.
Materials and manufacture of the experimental
formulations
The drug R1315 was synthesized by the chemical development department of F. Hoffmann-LaRoche Ltd. in Basel. The
excipients Cremophor RH40 (polyoxyl 40 hydrogenated castor), Solutol HS 15 (polyethylene-660-hydroxystearate) and
Kollidon 30 (polyvinylpyrrolidone, PVP K30) were obtained
from BASF (Germany), whereas the Tween 20 (polysorbate 20
or polyethylene-20-sorbitan-monolaurat) was purchased from
Fluka (Switzerland). The surfactants are all non-ionic and all
excipients were used as received.
The excipients were tested with the drug base of R1315
in a quarter fraction 252 statistical design to achieve maximal drug solubilisation. The different compositions can be
inferred from Table 1. The nally selected surfactant solution
contained 5% Cremophor RH40 with the two concentrations
of 1 and 2 mg/mL drug base of R1315.
For the solid dosage form the drug was manually lled in
hard gelatine capsules of size No. 0 (Capsugel, France). All capsules were prepared and weighed on the same day of animal
dosing.

2.3.
Solubility measurements and analytical HPLC
method
The formulation samples for solubility measurements were
equilibrated at 25 1 C during 24 h, ltrated (Millipore 0.22 m
pore size) and consecutively analyzed by HPLC (Waters
Alliance) with a reversed phase column (Symmetry C18 5 m
from Waters). The column temperature was 40 C. A gradient method was used with a mobile phase consisting of 50%
50 mM Tris in water at pH 7.0 (adjusted with HCl) and 50% acetonitrile. The second eluent consisted of 50% 50 mM Tris in
water at pH 7.0 (adjusted with HCl) and 70% acetonitrile. The
ow rate was 1 mL/min and UV detection was at 245 nm.

2.4.
Computer simulation of absorption using
GastroPlusTM
Computer simulations were performed with the software
GastroPlusTM V.4.0 (Simulations Plus Inc., Lancaster, USA),

94

e u r o p e a n j o u r n a l o f p h a r m a c e u t i c a l s c i e n c e s 2 7 ( 2 0 0 6 ) 9199

which uses a physiologically based absorption model with

nine compartments corresponding to different segments
of the digestive tract. GastroPlusTM uses a set of differential equations to model the amount of drug that is
released, dissolved and absorbed for the nine compartments.
This advanced compartmental absorption and transit model
(ACAT) is described in Agoram et al. (2001) and further details
of the computer program are given under the homepage
http://www.simulations-plus.com/.
The pH values in the segments can be inferred from the
literature (e.g. Dressman et al., 1998) and GastroPlusTM V.4 uses
the following default values for the human fasted condition:
stomach pH 1.7, duodenum pH 6.0, jejunum pH 6.26.4, Ileum
pH 6.67.5 and colon (caecum) pH 5.0.
Given a known solubility at any single pH, the solubility
at all other pH values was estimated taking into account the
pKa value. Fig. 3 shows the example of lower limit solubility
versus pH prole. The prole takes into account the measured
value in FaSSIF (pH 6.5) as a worst case solubility, because it
is an equilibrium value at room temperature (RT). It is a well
established approach to set an upper limit for the solubility
increase as a function of pH. The default value for bases in
GastroPlusTM is a factor of 50. We used instead of this arbitrary
factor a cut off solubility measured in SGF at pH 1.2. It is again
expected that the in vivo solubility at 37 C is higher. Accordingly, the parameter sensitivity analysis considered additional
solubility versus pH proles with solubility values in a range
of up to 100 times higher than displayed by Fig. 3 to cover a
broad range of possible in vivo solubilities.
GastroPlusTM also includes a mean precipitation time. This
parameter is understood as a mean time for particles to precipitate from solution when the local concentration exceeds
the drug solubility. Usually a default value of 900 s is assumed,
however, based on the solubility experiments in articial
intestinal uids, a rst estimate of 1800 s was set and consequently a range of 18018,000 s checked in the parameter
sensitivity analysis.
In GastroPlusTM , the dissolution rate constant is given by:
Kd = 3

C s Cl
rT

where
is the NoyesWhitney diffusion coefcient,  the density of the drug particle, r the radius, T the diffusion layer
thickness, Cs the solubility and Cl is the lumen concentration.
The diffusion layer thickness was taken as equal to the particle
radius.

The relevant input and GastroPlusTM default parameters

are presented in Table 2. Most of the compound parameters
for R1315 were directly obtained from the preformulation data.
The permeability result from the PAMPA assay was converted
to an effective human jejunal permeability, Peff based on a
correlation built with 20 reference drugs (Roche in-house reference). It was assumed for R1315 that no carrier-mediated
transport mechanism is involved.
Theoretically, the absorption rate coefcient, ka (i) for each
compartment is the product of the effective permeability
value, Peff (i) for the compartment and the absorption scale
factor, (i) for the compartment. This absorption scale factor is in theory simply the ratio of the surface area to volume, which reduces to 2/R, where R is the radius of the
compartment. Since the individual effective permeability values are unknown for the different compartments the usual
GastroPlusTM practice was followed to use only the one
estimate for the effective jejunum human permeability as
described above. Accordingly, the absorption scaling factors
were adjusted using software default values based on the
so-called log D model. This model considers the inuence
of the log D on the effective permeability. In other words,
as the ionised fraction of a compound increases the effective permeability decreases. GastroPlusTM includes further
default values for the transit times in the human GI tract,
which enables to predict the rate and extent of drug absorption.
For simulation of plasma concentrations some estimates
of human pharmacokinetic parameters are required. It was
known for this compound that for three preclinical species in
vitro data obtained with primary cultures of hepatocytes gave
reasonable estimates of the observed in vivo clearance. Therefore, human clearance was estimated from intrinsic clearance measured in human hepatocytes and scaled to in vivo
using physiologically based principles (Zuegge et al., 2001).
The estimated human hepatic clearance was 0.15 L/(h kg).
From this clearance value, assuming a liver blood ow of
1.2 L/(h kg) and a blood/plasma ratio of unity, 12.5% of the
drug was estimated to be extracted on rst pass through the
liver.
Simulations of absorption following dosing of 160 mg of an
immediate release form of R1315 were performed and also at
double this dose. Then a parameter sensitivity analysis was
performed where the particle size and drug solubility values
were varied and their impact on the oral bioavailability ascertained.

Table 2 Input variables and accepted default values for the GastroPlusTM simulation
General simulation and compound parameters

Physiological parameters

Pharmacokinetics

MW: 409.41 g/mol

c log P: 5.5
pKa : 5.9
Dosage form; immediate release capsule with 160 mg dose
Lower limit reference solubility (pH 6.5) 0.002 mg/mL
Mean precipitation time: 1800 s
Particle density: 1.2 g/mL
Effective permeability: 4.4 104 cm s1
Effective particle radius: 5 m

Human fasted conditions

Absorption model: log D model
Stomach transit time: 0.1 h
Dose volume: 250 mL
Small intestine transit time: 3.3 h
Small intestine radius: 1.2 cm
Small intestine length: 300 cm
Colon volume: 1200 mL

Body weight: 70 kg
First pass extraction (xed): 12.5%
Blood to plasma concentration ratio = 1
Clearance: 0.15 L/(h kg)
Vc: 1.9 L/kg

e u r o p e a n j o u r n a l o f p h a r m a c e u t i c a l s c i e n c e s 2 7 ( 2 0 0 6 ) 9199

95

Table 3 23 design for dog pharmacokinetic studies

No.

Formulation

Dose level
(mg/kg)

Feed status

1
2
3
4
5
6
7
8

Capsule
Capsule
Solution
Capsule
Solution
Solution
Capsule
Solution

2
2
2
4
4
2
4
4

Fed
Fasted
Fasted
Fed
Fasted
Fed
Fasted
Fed

2.5.

Fig. 2 Standardised effects of solubility enhancement for

R1315 as Pareto chart.

Pharmacokinetic experiments in beagle dogs

Male beagle dogs (weighing 1014 kg, obtained from the Roche
in-house colony) were used for the oral bioavailability studies. The dosing was performed according to a randomized 23
factorial design outlined in Table 3. Two doses 2 and 4 mg/kg
were administered orally as drug in a capsule as well as a drug
solution under both, fed and fasted conditions. In the fed condition, the animals received 200 g commercially available dog
food 30 min prior to the administration of the test article. In the
fasted condition, food was withdrawn overnight before dosing
and was not given during the experiment.
Blood samples (approximately 1 mL each) were drawn from
the cephalic vein at predose, and 0.5, 1, 1.5, 2, 3, 4, 6, 8 and
24 h postdose. Collection tubes contained EDTA as anticoagulant. Plasma was obtained by centrifugation and stored
deep-frozen at approximately 20 C until analysis. A selective
LCMS/MS method was used for the quantication of R1315.
The drug R1315 and an internal standard were isolated from
plasma samples after protein precipitation by on-line solid
phase extraction and separated from other constituents of the
sample by narrow-bore HPLC. Detection was accomplished
utilising ion spray MS/MS in positive ion selected reaction
monitoring mode. The limit of quantication was 1 ng/mL. The
assay performance was monitored using quality control samples spiked with known concentrations of R1315.
AUC0-inf (area under the plasma concentrationtime curve)
was estimated by non-compartmental analysis (applying the
linear trapezoidal rule), using the pharmacokinetic evaluation
program WinNonlin Pro . Extrapolation to innity was performed from the last measurable concentration using the rate
constant of the apparent terminal plasma level prole.

and so a degree of freedom of ve was obtained for the estimate of the total error.
A Pareto chart was selected to rank the estimated effects in
decreasing order of magnitude (Fig. 2). The length of each bar
is proportional to the standardised effect. This standardised
effect is the estimated effect divided by its standard error. The
vertical line of the plot marks those effects that are statistically signicant. Bars that extend beyond the line correspond
to effects that are statistically signicant at the 95% condence
level.
Screening designs were proposed before for preclinical
tests in animals (Kuentz et al., 2003). In the current study, a 23
full factorial design was conducted in a single block. The effect
of the formulation, the clinical dose, as well as the food effect
was evaluated in parallel. The full factorial design allowed further to estimate the interactions of the different factors (none
of the effects or rst order interactions are confounded corresponding to a resolution V). This detection of interactions is
an advantage of the full factorial design that has on the other
hand here only a rather low degree of freedom (d.f.) for estimation of the total error, so that the statistical power of this study
plan is rather low. The evaluation was therefore conducted on
a 90% signicance level to obtain an acceptable probability for
. The calculated power curve shows that a true effect would
have to be at least seven to eight times greater than its standard deviation to have adequate probability of effect detection.
Thus, only marked effects can show signicance in this design.
The statistical evaluations were calculated using the software
package Stagraphics Plus V. 5.0 (Manugistics Inc.).

3.

Results and discussion

2.6.
Statistical design of experiment techniques and
evaluation of the data

3.1.

Development of experimental formulations

There are introductions to the eld of planning factorial or

fractional factorial designs (Kleppmann, 2003; Lewis et al.,
1999). Senderak et al. (1993) was one of the pioneers to use
design of experiment techniques to optimize drug solubility
in an oral solution. In the present study, a quarter fraction 252
factorial design was used for screening purposes. The effect of
three surfactants, a polymer and a co-solvent in view of drug
solubilisation was evaluated. The compositions of the eleven
mixtures are displayed by Table 1. Since the design has only a
resolution of III, the main effects are confounded with interactions. These interactions between two factors were neglected

The development of a drug solution exhibiting maximal solubility was carried out as a part of the toxicological formulation
development. Initial solubilisation studies used the drug base.
Different surfactants were evaluated regarding their potential to get the best drug solubility. Fig. 2 shows the effects of
the different excipients on drug solubilisation. The presence
of Cremophor RH40 enabled highest drug concentrations in
solution, whereas Solutol HS15 and Tween 20 displayed
lower solubilisation capacities. The three non-ionic surfactants incorporate the drug mainly in micelles. The extent of
solubilisation is affected by the nature of the surfactant as well

96

e u r o p e a n j o u r n a l o f p h a r m a c e u t i c a l s c i e n c e s 2 7 ( 2 0 0 6 ) 9199

as of that of the drug. Within a homologues series of a given

surfactant one may state as a rule of thumb that the solubilisation capacity of a hydrophobic drug increases with the length
of the hydrocarbon chain and the number of ethylene oxide
units. The latter effect is mainly due to an increased number of
micelles per mole (Florence and Attwood, 1998). Certainly, the
comparison of structurally very different surfactants is more
difcult, which makes an experimental ranking indispensable.
The polymer PVP K30 showed a slight, but signicant effect
on drug solubilisation, while this was interestingly not the
case for the propylene glycol. The lacking co-solvent effect
must be understood within the concentration range examined. It should be further highlighted that interactions of
the excipients were not considered in the evaluation of the
252 screening design. The effect of the co-solvent was estimated from mixtures with surfactants that predominantly
solubilised the drug. The involved micelle formation could be
hindered by the presence of propylene glycol, which can be
supported by thermodynamic arguments. The driving force
of micelle formation lies in a positive entropy change arising from the release of structured water that occurs when
hydrophobic surfactant chains form the core of a micelle
(Martin, 1993). The positive entropy change should therefore
be highest in pure water systems, whereas aqueous solutions
of propylene glycol are likely to exhibit a less pronounced
entropy change. In essence, a co-solvent can have a negative interaction with surfactants that solubilise the drug by
micellation. This is reected by the results showing that a
combination of surfactant with co-solvent does not increase
solubilisation of R1315 and so Cremophor RH40 was used
alone with the drug in the nal experimental formulation.
Based on the obtained results different solutions with varying amounts of Cremophor RH40 were tested in view of solubilising at least 0.2% (w/w) R1315. Finally, an aqueous 5%
(w/w) Cremophor RH40 vehicle was used to prepare a 1 and
2 mg/mL solution of R1315. Both solutions contained 0.18%
methylparaben and 0.02% proylparaben as preservatives. An
HPLC analysis of the drug solution was performed directly
after the manufacture and following 10 days storage at both 5
and 25 C. The concentrated drug solution showed an increase
of the total degradation products of +0.36% stored at 5 C, and
+2.17% at 25 C, respectively. A daily ad hoc preparation was
decided for this type of experimental drug solution to avoid
any stability issue.
This rst experimental formulation is favourable from a
biopharmaceutical point of view over formulations, where the
drug is crystalline. The extreme case was represented by the
experimental formulation of pure drug substance in a capsule,
since there is no granulation step or additional excipients that
could improve the poor wetability of R1315. Stability tests were
conducted with the solid drug R1315 at different temperatures
and humidities. No degradation could be detected following
1 week storage at 40 C/75% R.H., which indicates sufcient
chemical stability at least for an ad hoc usage of the drug.

the complexity of the GastroPlusTM model and the numerous

input data some care is needed in interpretation of results.
Our usual strategy with GastroPlusTM is therefore to verify the
simulations against in vivo results in one or more preclinical species before making prospective simulations in human.
In this study, checks on the reasonableness of predicted oral
absorption in rat (in terms of consistent plasma proles; data
not shown) were made before proceeding with the human
simulations and sensitivity analysis. The timelines in Fig. 1
show that these modelling and simulation activities using animal in vivo data precede the biopharmaceutical assessment
program described in this article.
The results of the rst human simulation with a single
immediate release dose of 160 mg R1315 are shown in Fig. 4.
A fraction of the dose goes into solution in the acidic environment of the stomach and there is later a partial precipitation of
drug base in the intestine. However, the precipitated drug has
sufcient time to be re-dissolved and is well absorbed due to
the high permeability of the drug. Accordingly, the simulation
implies that stomach solubility of the drug has little impact
on the overall fraction absorbed. The drug permeation and redissolution are overlapping processes with the fast removal
of drug from the intestine creating a constant sink and promoting further base to go into solution. The fraction absorbed
is high, and the oral bioavailability is slightly reduced by liver
metabolism during the rst pass.
The simulation was repeated with a doubled dose of 320 mg
R1315. Fig. 5 shows that at this higher dose level there is a
lower fraction of the dose dissolved in the stomach. The full
dissolution step, involving also precipitated drug, takes more
time than observed with the lower dose, but the GI transit time
is sufcient and so the bioavailability values for both doses are
very similar in simulations over 24 h.
In a sensitivity analysis, the pharmaceutically relevant
parameters of mean particle size and drug solubility were
taken as variables over a 100-fold range (at a dose of 160 mg
R1315). The predicted dependence of oral bioavailability on
reference solubility at pH 6.5 (assuming shifted solubility versus pH proles similar to Fig. 3 is shown in Fig. 6. The oral

3.2.
Results from the physiologically based computer
simulation
The validity of any computer modelling is dependent on the
quality of both the model and of the input data. In view of

Fig. 3 Drug solubility vs. pH prole used as lower limit of

the solubility values of a set of computer simulations.

e u r o p e a n j o u r n a l o f p h a r m a c e u t i c a l s c i e n c e s 2 7 ( 2 0 0 6 ) 9199

Fig. 4 Simulated drug amounts (%) for the 160 mg dose of

R1315 in human.

bioavailability was hardly affected by reference solubility values (pH 6.5) in the range of 0.0020.2 mg/mL. Thus, higher
solubility values may increase the rate of absorption, but not
its extent. We further repeated single simulations at a 160
and 320 mg dose at 0.02 and 0.2 mg/mL reference solubility
to study the dissolution and absorption process analogues
to Figs. 4 and 5. The amount of dissolved drug was greatly
increased and complete absorption was reached at an earlier
time point so that almost no colonic absorption was predicted.
This is different from the results at the lower reference solubility 0.002 mg/mL (Figs. 4 and 5) where absorption in the colon
played a role.
The parameter sensitivity analysis was repeated with varying mean effective particle sizes ranging from 0.5 to 50.0 m.
A similar bioavailability prediction was obtained with the different particle sizes (Fig. 6). However, there is a tendency
towards slightly reduced bioavailability at the lower drug solubility of 0.002 mg/mL and the upper 50.0 m particle size, but
within the examined ranges the bioavailability was not greatly
affected. The parameter sensitivity analysis was repeated with
a doubled dose of 320 mg R1315. A similar picture was observed
as displayed for the lower dose by Fig. 6. We further varied the
precipitation time between 180 and 18 000 s at a dose of 320 mg
and found again that oral bioavailability was hardly affected.
This indicates that even at a relatively fast precipitation of
drug, the re-dissolution and permeation are fast enough to
make the given dose bioavailable.
The set of simulations was completed by varying the effective permeability from 0.44 to 44 104 cm s1 . A nearly constant bioavailability was predicted with a slight decrease at
the lower permeability limit, which was comparable to that
displayed by Fig. 6. Accordingly, permeability seems not to be
an issue for the given drug in the examined high permeability
range.
The observed insensitivity of input parameters on oral
bioavailability holds of course only within the inspected limits. However, the examined ranges of solubility, precipitation time and particle size are of technological relevance.
Also the doses of R1315 simply reect the clinically relevant
range, as it spans the anticipated doses in the planned human
studies.

97

Fig. 5 Simulated drug amounts (%) for the 320 mg dose of

R1315 in human.

Dannenfelser et al. (2004) reported a simulation of a poorly

soluble drug where solubility and drug particle size clearly
inuenced absorption. A surfactant solution and solid dispersion of the drug reached ten times higher bioavailabilities (at
the given dose) in dogs than a blend of the micronised drug
with microcrystalline cellulose. The present drug R1315, however, is different from that simulated by Dannenfelser et al.
(2004). As a result, the simulation results using R1315 indicate
that within the examined ranges, a particle size reduction or
solubility enhancement by technological means may not lead
to increased absorption and nally higher oral bioavailability
values.

3.3.

Evaluation of the pharmacokinetic study in dogs

Whereas the rst part of the preclinical formulation strategy

is an in silico assessment of the human situation, the second part uses experimental data from an in vivo model. The

Fig. 6 Parameter sensitivity analysis of the oral

bioavailability (%) as a function of reference solubility at pH
6.5 (mg/mL) [dark squares], as well as effective particle
radius (m) [light squares] at a dose of 160 mg R1315.

98

e u r o p e a n j o u r n a l o f p h a r m a c e u t i c a l s c i e n c e s 2 7 ( 2 0 0 6 ) 9199

Table 4 Calculated effects and interactions of the 23

design in terms of the AUC0-inf (ng h/mL)
Estimated effects and interactions
A: formulation
B: dose level
C: feeding status
AB
AC
BC

AUC0-inf ng h/mL
19.1
160.6
18.6
109.6
55.4
205.1

31.1 (0.649)
31.1 (0.122)
31.1 (0.657)
31.1 (0.176)
31.1 (0.326)
31.1 (0.096)

Standard errors are based on the total error (from ANOVA) with 1d.f.
The P-values from the ANOVA are displayed in brackets.

Fig. 7 Plasma levels of individual dogs that received a

solution. Diamonds hold for proles of 2 mg/kg dose,
whereas those of the 4 mg/kg dose are represented by
squares. The light symbols hold for the fasted condition
and the bold symbols for fed dogs.

beagle dog was chosen for the tests of experimental formulations. There are certainly differences in drug absorption and
especially metabolism between dog and man. Sutton (2004)
recently reviewed these physiological differences in view of a
dosage form performance. The importance of factors like GI
transit times or pH clearly depend on the drug candidate and
its dose. However, the scope of using the present animal model
was primarily to learn about relative effects in vivo.
Fig. 7 shows the experimental plasma levels as a function of
time from the dogs that received the drug solution, i.e. the dogs
no. 3, 5, 6 and 8 (Table 3). These plasma versus time proles
can be compared with those from dogs where a capsule was
administered (Fig. 8).
There is great variability in the data and a clear difference
between the two formulations cannot be directly noticed. The
statistical evaluation conrms that there is on the average
not a signicant difference between the drug solution and the
capsule formulation (Table 4). This absence of a marked formulation effect in the dog model can be compared with the
results of the computer simulation. The in silico experiments

Fig. 8 Plasma levels of individual dogs that received a

capsule. Diamonds hold for proles of 2 mg/kg dose,
whereas those of the 4 mg/kg dose are represented by
squares. The light symbols hold for the fasted condition
and the bold symbols for fed dogs.

showed that formulation differences are levelled in terms of

oral bioavailability.
The effect of the dose between 2 and 4 mg/kg was on the
average 160 31.1 ng h/mL (Table 4), which is again not a
marked effect and the P-value was 0.122. The corresponding
P-value for the potential food effect was even much higher
(Table 4) so that food consumption in the present case may
at least on the average be of a lesser biopharmaceutical relevance.
The values of the factor interactions provided a signicant
result on the 90% condence level with P = 0.096 in case of dose
level and feeding status. This would mean that an inuence
of the dose depends on the level of the food factor. Such an
interaction is not easy to rationalise on a mechanistic basis.
Certainly additional experiments are needed for mechanistic explanations, but this would go beyond the scope of the
present screening study.
Generally, it can be stated that the considerable variability
of the data limit the precision to estimate effects and interactions. However, the results indicated the absence of pronounced effects in terms of the AUC0-inf . This was particularly
the case for the two formulation principles that are very different from a technological viewpoint and should represent a
worst and a best case in view of oral bioavailability.

4.

Conclusions

The new CNS drug R1315 was initially proled to have poor solubility and high permeability. Such a rst biopharmaceutical
assessment has usually a direct impact on the selection of the
formulation strategy for phase I clinical trials. In the present
work, an alternative procedure was followed. Prior to initiating
a resource intensive formulation development, we opted for a
more detailed biopharmaceutical assessment. A GastroPlusTM
computer simulation was conducted in view of dosage form
performance. Drug solubility and particle size appeared to be
hardly relevant for the oral bioavailability within two orders of
magnitude. This conclusion emphasises the importance of a
more thorough biopharmaceutical view than just the assignment of the BCS class. Based on the poor solubility, the drug
R1315 would a priori be expected to show dependence of the
oral bioavailability on parameters like the particle size or solubility. It is, however, crucial to know about such dependence in
a range that can be inuenced by technological means. Therefore, two experimental formulations were manufactured that

european journal of pharmaceutical sciences

span the range from the particulate drug in a capsule to a

surfactant drug solution. A pharmacokinetic testing in beagle dogs was run as a factorial design to examine the effect
of the formulation in parallel with a potential food effect in a
clinically foreseen dose range. The results indicated that the
technologically very different formulations were not very different with respect to exposure on average.
Accordingly, it was decided for R1315 that no sophisticated
drug delivery system was to be developed in parallel to a
standard dosage form for phase I clinical trials. The development strategy was different from what was initially planned
for this BCS class II compound. It can therefore be concluded
that GastroPlusTM computer simulation in conjunction with a
factorial design of the pharmacokinetic animal trials led to a
deeper mechanistic understanding this again resulted in considerable resource savings.

Acknowledgements
The authors wish to thank Dr. R. Moog for providing the preformulation data of R1315, Dr. N. Burki for the analytical support,
Dr. B. Lausecker for the measurement of plasma levels, Mr.
P. Schrag for conducting the dog experiments, as well as J.M.
Clavey for the preparation of the drug formulations.

references

Agoram, B., Woltosz, W.S., Bolger, M.B., 2001. Predicting the

impact of physiological and biochemical processes on oral
drug bioavailability. Adv. Drug. Deliv. Rev. 50, S41S67.
H., Shah, V.P., Crison, J.R., 1995. A
Amidon, G.L., Lennernas,
theoretical basis for a biopharmaceutical drug classication:
the correlation of an in vitro drug product dissolution and
in vivo bioavailability. Pharm. Res. 12 (3), 413420.
Balbach, St., Korn, Ch., 2004. Pharmaceutical evaluation of early
candidates the 100 mg-approach. Int. J. Pharm. 275, 112.
Dannenfelser, R.M., He, H., Joshi, Y., Bateman, S., Serajuddin,
A.T.M., 2004. Development of clinical dosage forms for a
poorly water soluble drug: application of polyethylene
glycol-polysorbate 80 solid dispersion carrier system. J.
Pharm. Sci. 93 (5), 11651175.
Dressman, J.B., Amidon, G.L., Reppas, Ch., Shah, V.P., 1998.
Dissolution testing as a prognostic tool for oral drug
absorption: intermediate release dosage forms. Pharm. Res.
15 (1), 1122.
Florence, A.T., Attwood, D., 1998. Physicochemical Principles of
Pharmacy, third ed. Macmillan Press Ltd, pp. 245246.

27

( 2 0 0 6 ) 9199

99

Galia, E., Nicolaides, E., Horter, D., Lobenberg,

R., Reppas, C.,
Dressman, J., 1998. Evaluation of various dissolution media
for predicting in vitro performance of class I and II drugs.
Pharm. Res. 15, 698705.
Ingels, F.M., Augustijns, P.F., 2003. Biological, pharmaceutical,
and analytical considerations with respect to the transport
media used in absorption screening system, Caco-2. J.
Pharm. Sci. 92 (8), 15451558.
Kansy, M., Senner, F., Gubernator, K., 1998. Physicochemical
high throughput screening: parallel articial membrane
permeation assay in the description of passive absorption
processes. J. Med. Chem. 41, 10071010.
Kerns, E.H., Di, L., 2003. Pharmaceutical proling in drug
discovery. Drug Discov. Today 8 (7), 316323.
Kleppmann, W., 2003. Taschenbuch Versuchsplanung. Produkte
und Prozesse optimieren. Hanser, Munich, pp. 93154.

Kuentz, M., Rothlisberger,

D., Richter, W., 2003. Design of
experiment (DoE) methods maximize information from a
minimal number of animals in special cases of preclinical
bioavalability testing. Pharm. Dev. Technol. 8 (4), 453458.
Lewis, G.A., Mathieu, D., Phan-Tan-Luu, R., 1999.
Pharmaceutical Experimental Design. Marcel Dekker, New
York, pp. 23150.
Martin, A., 1993. Physical Pharmacy, fourth ed. Lea & Febiger,
Philadelphia, pp. 414418.
Norris, D.A., Leesman, G.D., Sinko, P.J., Grass, G.M., 2000.
Development of predictive pharmacokinetic simulation
models for drug discovery. J. Control. Release 65, 5562.
Plusquellec, Y., Efthymiopoulos, C., Duthil, P., Houin, G., 1999. A
pharmacokinetic model for multiple sites discontinuous
gastrointestinal absorption. Med. Eng. Phys. 21, 525532.
Senderak, E., Bonsignore, H., Mungan, D., 1993. Response
surface methodology as an approach to optimization of an
oral solution. Drug Dev. Ind. Pharm. 19 (4), 405424.
Sinko, P.J., Leesman, G.D., Amidon, G.L., 1991. Prediction
fraction dose absorbed in humans using a macroscopic
mass balance approach. Pharm. Res. 8, 979988.
Sutton, S.C., 2004. Companion animal physiology and dosage
form performance. Adv. Drug Deliv. Rev. 56, 13831389.
Yokoe, J., Iwasaki, N., Haruta, S., Kadano, K., Ogawara, K.,
Higaki, K., Kimura, T., 2003. Analysis and prediction of
absorption behavior of colon-targeted prodrug in rats by
GI-transit-absorption model. J. Control. Release 86, 305313.
Yu, L.X., Lipka, E., Crison, J.R., Amidon, G.L., 1996a. Transport
approaches to the biopharmaceutical design of oral drug
delivery systems: prediction of intestinal absorption. Adv.
Drug Deliv. Rev. 19, 359376.
Yu, L.X., Crison, J.R., Amidon, G.L., 1996b. Compartmental
transit and dispersion model analysis of small intestinal
transit ow I humans. Int. J. Pharm. 140, 111118.
Zuegge, J., Schneider, G., Coassolo, P., Lave, T., 2001. Prediction
of hepatic metabolic clearance in mancomparison and
assessment of prediction models. Clin. Pharmacokinet. 40,
553563.