European Journal of Pharmaceutical Sciences 115 (2018) 286–295

Contents lists available at ScienceDirect

European Journal of Pharmaceutical Sciences

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

In vitro characterization of ritonavir formulations and correlation to in vivo T

performance in dogs
⁎
Hao Xua, Silvia Krakowb, Yi Shia, Joerg Rosenbergb, Ping Gaoa,
a
NCE-Formulation Sciences, Drug Product Development, AbbVie Inc., North Chicago, IL 60064, USA
b
NCE-Formulation Sciences, AbbVie Deutschland GmbH Co. KG, Ludwigshafen, Germany

A R T I C L E I N F O A B S T R A C T

Keywords: Ritonavir (RTV) is a weakly basic drug with a pH-dependent solubility. In vitro characterization of dissolution
Ritonavir and supersaturation behaviors of three PEG-8000 based amorphous solid dispersions (ASD) and a physical blend
Absorption (PB) with crystalline drug were performed in the biomimetic media (e.g., FaSSGF, FaSSIF, FaSSIF-V2). A two-
Biomimetic stage dissolution test and a biphasic dissolution-partition test at the small scale (referred as to biphasic test) were
Dissolution
employed with intention to examine the in vitro and in vivo relationship (IVIVR) with retrospective PK data in
IVIVR
dog model.
Liquid-liquid phase separation
Biphasic The two-stage dissolution test revealed a high degree of supersaturation of RTV from these ASDs accompanied
BCS by the occurrence of liquid–liquid phase separation (LLPS) in the biomimetic media. A rapid decrease of ap-
Partition parent RTV concentrations of these ASDs was associated with significant precipitation upon the pH shift of the
Relative bioavailability dissolution medium, revealing the important role of “the gastric stage”. In comparison, the biphasic test revealed
a lower degree of supersaturation of RTV that is attributed to removal of RTV through partition into octanol,
acting as “the absorption compartment”. These two dissolution tests provide characterization of the super-
saturation state with a complex, dynamic interplay among dissolution, precipitation and partition processes.
Results of both in vitro dissolution tests are in good agreement with in vivo results in dogs. In addition, three
commercial generic RTV drug products were examined by the biphasic test. Agreement was also obtained be-
tween the RTV concentrations in octanol at 3 h from these generic drug products and their corresponding re-
lative bioavailability in dogs.

1. Introduction precipitation and partition occur simultaneously (Gao et al., 2016;

The design of supersaturation sustaining formations as an effective
strategy to enhance the in vivo absorption drugs with low aqueous
solubility has been extensively explored in the last decade (Gao and
Morozowich, 2005; Brouwers et al., 2009; Frank et al., 2014a). Char-
acterization of in vitro dissolution, and, in particular, characterization
of the supersaturated state associated with these enabling formulations
(e.g., amorphous solid dispersions), are highly desirable in order to
optimize their bioperformances (Baghel et al., 2016l; Newman et al.,
2012; Gao and Shi, 2012; Kostewicz et al., 2014; Nguyen et al., 2017;
Lu et al., 2017). Diverse two-compartment dissolution methods with pH
change have been utilized to simulate the dissolution and transit of
dosage forms from the stomach to the small intestine in vivo (Kostewicz
et al., 2014; Nguyen et al., 2017; Lu et al., 2017; Tsume et al., 2017). In
contrast, a two phase dissolution-partition test method with pH shift,
referred to as the biphasic test, has been developed in order to simulate
a dynamic environment in which three processes of drug dissolution,
Pestieau and Evrard, 2017). A practical biphasic system combining an
USP IV apparatus (with a flow cell) and an USP II dissolution apparatus
was first reported by Gao and his co-workers in 2009 (Vangani et al.,
2009). The biphasic test method permits the dissolution of the drug
product in various aqueous media under a non-sink condition and si-
multaneous partition of the dissolved drug into an organic phase that
acts as an “absorption compartment”. The partition of the drug is driven
by its free drug fraction in the aqueous phase, mimicking the absorption
in vivo. Therefore, the concentration-time profile of the drug in the
organic phase may serves as an output for establishing in vitro and in
vivo relationships (IVIVR). Applications of the biphasic method to BCS
II drugs including AMG517 (Vangani et al., 2009), celecoxib (Shi et al.,
2010), ABT-072 (Shi et al., 2016), fenofibrate (Xu et al., 2018), and
related formulations have been reported with IVIVR. In addition,
multiple investigations of biphasic tests for several poorly water soluble
drugs by other research groups demonstrated its broad utility for drug
product development (Pestieau and Evrard, 2017; Frank et al., 2014a,

⁎
Corresponding author.
E-mail address: ping.gao@abbvie.com (P. Gao).

https://doi.org/10.1016/j.ejps.2018.01.026
Received 19 September 2017; Received in revised form 4 December 2017; Accepted 11 January 2018
0928-0987/ © 2018 Elsevier B.V. All rights reserved.
H. Xu et al.
2014b; Locher et al., 2016; Locher et al., 2016).
Amidon et al. developed a mechanistic model to describe the drug
transport phenomenon associated with the two-phase dissolution
system (Mudie et al., 2012). This approach assumes a first-order ab-
sorption kinetics and a relatively high fraction absorbed in vivo (Fa).
The kinetics of drug partitioning from the aqueous phase to the organic
phase is based on simultaneous equilibria and mass transfer of dissolved
drug through the water-organic interface. The in vitro partitioning rate
coefficient, kp (equal to (AI/Va) ∗ PI; where AI is the surface area of the
interface, Va the volume of dissolution medium, and PI the partition
coefficient through the interface), reflects the rate of drug partitioning
into the organic phase. The physiological relevance of this test method
is that the in vitro kp approximates the in vivo absorption rate coeffi-
cient, ka as described below.
A A
kp = ⎛ I PI ⎞
⎜ ⎟ = k a = ⎛ Peff ⎞
⎝ Va ⎠in vitro ⎝ V ⎠in vivo
where A is the surface area for absorption in vivo, V the volume of the
GI fluid, and Peff the permeation coefficient through the membrane.
Based on a known (or estimated) ka and observed (or estimated) PI, one
can adjust AI/Va so that kp and ka become similar or equal. The biphasic
test employed in our laboratory possesses a flexibility of different
scaling factors (A/V) associated with the small and large scales in-
tended to mimic different physiologies of dogs and humans. As we have
reported the application of the biphasic test methods to characterize
varying formulation technologies of an weak acid drug ABT-072 of
extremely low solubility, good agreements were obtained between the
in vitro profiles at both small and large scales and in vivo PK results in
dogs and humans, respectively (Shi et al., 2016).
Oral drug delivery is a highly complex field with many unknowns.
Animal models have been widely used for preliminary evaluation of
new chemical entities before reaching the patient in a clinical set-
ting—namely, in terms of safety and minimizing toxicity, but also in
enabling optimization of formulation characteristics and specific drug
targeting, among others. Elucidating the mechanisms of drug absorp-
tion following oral administration has remained a long-standing goal
across a multitude of scientific disciplines, namely, for the purposes of
predicting human safety as well as pharmacokinetics and pharmaco-
dynamics for potential drug candidates. It has been well recognized that
animal models remain a poor simulacrum of human physiology, given
that many species are often fundamentally distinct in terms of their GI
anatomy and physiology as compared with humans. Nonetheless, they
have been frequently put to use in preclinical studies, justified by a
similarity to humans by key parameters, such as GI fluids with pH
variation, the transit times and/or the regional absorption.
Although significant progress has been made in recent years to
improve in vitro techniques for assessment of drug characteristics and
bioperformance, these approaches are not considered a reliable sub-
stitution of in vivo evaluation. Indeed, there remain significant gaps in
our knowledge of animal and human physiologies, yet the animal
models continue to be used in both academic and industrial environ-
ment in drug development and preclinical studies. Development of
biomimetic in vitro test methods has been partially supported by re-
liance on animal models as intermediary models for evaluation of
compounds through in vitro–in vivo correlations. Valuable information
of specific physiological and anatomical parameters influencing drug
absorption, especially the complex dynamics from the gut, is difficult to
simulate in vitro. Such understanding from animal models enables ex-
trapolation to humans, difficult to obtain otherwise. As the dog model
still plays an important role in drug product development, it is desirable
to develop appropriate in vitro tests and examine their relevance to in
vivo performance in dogs.
RTV (its molecular structure shown in Fig. 1A) is a HIV protease
inhibitor. It is a very weak base with pKa values of 1.8 and 2.6 (Law
et al., 2001; Law et al., 2004). The crystalline Form I and II possess a
287
European Journal of Pharmaceutical Sciences 115 (2018) 286–295
high aqueous solubility at pH < 1 and extremely low solubility of
2 μg/mL at pH 4–7 (Fig. 1B). RTV was revealed to have a low to mod-
erate intestinal permeability and is a Pgp substrate with moderate to
high efflux ratio indicated by Caco-2 results (Law et al., 2004; Aungst
et al., 2000; Alsenz et al., 1998). Clinical evaluation results of RTV oral
exposure with respect to the effect of Pgp and MRP1 expression in HIV
infected human subjects are consistent with this assessment (Meaden
et al., 2002). Therefore, RTV is classified as a BCS IV drug (Law et al.,
2004). It is of interest to note that RTV showed a high bioavailability of
~95 to 100% in dogs at the dose of 5 to 10 mg/kg (Kempf et al., 1995;
Hsu et al., 1998). Although the absolute bioavailability of RTV in
human subjects has not been determined, the fraction of dose absorbed
at the dose of 600 mg was estimated to be in the rage of 60 to 80%
(Kempf et al., 1995; Hsu et al., 1998). Such a high bioavailability of
RTV was hypothesized due to either saturation of the Pgp efflux pump
(Law et al., 2004) or other factors in vivo that minimized the active
transporter effect (Alsenz et al., 1998).
Prototype PEG-8000 based ASD formulations containing RTV of
10%, 20% and 30% (w/w), and one physical blend of crystalline drug
and PEG polymer (Table 2) were previously examined in dogs under the
fasting state (Law et al., 2004). That study applied a dissolution test
under a sink condition and did not investigate the supersaturation be-
haviors of RTV from such ASD formulations. Commercial RTV products
under the brand name of Norvir® (AbbVie) including the solution
(80 mg/mL), tablet (100 mg strength), and powder (100 mg/unit) were
examined by two biomimetic in vitro tests including the two-stage
dissolution test and the biphasic test at a large scale (Xu et al., 2017).
This was to characterize dissolution and precipitation behaviors of RTV
in the aqueous media and establish their relevance to pharmacokinetic
results in human subjects. These results revealed that the relationship
between supersaturation and solubilization of RTV from related PVP-
VA based ASDs is complex and highly media dependent. In addition to
achieve an agreement of the in vitro-in vivo correlation (IVIVC), we
gained mechanistic understanding of the food effect from the tablet and
powder formulations observed in vivo (Xu et al., 2017).
Relative bioavailability of RTV products manufactured by various
generic pharmaceutical companies was examined against the com-
mercial Norvir® capsule using an internally developed dog model at
Abbott/AbbVie (Garren et al., 2010). Although some commercial gen-
eric RTV drug products were selected for evaluation in dogs, they have
not been examined by the biomimetic in vitro tests at AbbVie.
The purpose of this work was to (1) evaluate RTV prototype PEG-
8000 ASDs by two biomimetic in vitro tests including the two-stage
dissolution test and the biphasic test and (2) evaluate three commercial
generic RTV drug products manufactured by Emcure, Hetero and Cipla.
This study was aimed to assess the utility of biomimetic in vitro test
methods and provide mechanistic understanding of the dynamic re-
lationship among the dissolution, precipitation and partition kinetics of
RTV in the aqueous media. These tests were evaluated to examine the
relevance of in vitro profiles to the in vivo exposure of all RTV for-
mulations tested in dogs.
2. Materials and Methods
2.1. Materials
Crystalline RTV was obtained in house. Poly(ethylene glycol) 8000
(PEG-8000, Carbowax) was obtained from Dow Chemical Company
(USA). Ethanol, octanol, acetonitrile, maleic acid, sodium hydroxide,
acetic acid, sodium acetate and sodium chloride were purchased from
Sigma (St. Louis, MO). The FaSSGF. FaSSIF and FaSSIF-V2 powders
were purchased from Biorelevant (UK).
2.1.1. ASD Formulations and Generic Drug Products
The preparation of the ASDs containing 10%, 20%, and 30% in PEG-
8000 matrix and 10% physical mixture (PB) containing crystalline RTV
H. Xu et al.
was reported in detail previously (Law et al., 2004). An alcohol solution
containing both RTV and PEG-8000 was dried with the use of a roto-
vapor at 75 °C, followed by subsequent cooling and drying overnight at
40 °C. The ASD materials were manually ground with a mortar and
pestle. After manually sifted to collect particle sizes between 149 and
420 μm, the ASD granules were examined by PXRD to verify the
amorphous state of RTV (data not reported). These granules were
manually filled into a hard gelatin capsule with an appropriate weight
for 50 mg RTV/unit.
Commercial generic tablet products containing 100 mg RTV
(Manufacturer, lot) were purchased including: Empetus (Emcure, lot
01A06001), Empetus (Hetero, lot EP80301), and Ritomune (Cipla, lot
K70556). These were the same products evaluated in the previous PK
study in dogs (Garren et al., 2010).
2.2. Methods
2.2.1. Two-Stage Dissolution Test
The two-stage dissolution test method is schematically shown in
Fig. 2A. The drug product was loaded to Erlenmeyer flask (250 mL)
containing 58 mL of the FaSSGF and agitated at 250 revolution-per-min
(rpm) in an orbit rotator for 30 min as ‘gastric-dissolution stage’. Then,
67 mL of the FaSSIF with a higher strength of 1.875× (pH 6.5) was
added into the FaSSGF in order to obtain the targeted composition of
FaSSIF-V2 after mixing with the FaSSGF. Samples were taken from the
dissolution medium at preselected time points. These samples were
centrifuged for 5 min at 21,000×g force. The supernatant (300 μL) was
pipetted out and mixed with an equal amount of methanol (300 μL) and
analyzed for RTV content by an Agilent 1290 HPLC system.
2.2.2. Biphasic Dissolution Test
This biphasic dissolution-partition method is shown schematically
in Fig. 2B. It involves an USP II dissolution apparatus in combination
with an USP IV apparatus (Sotax). A standard CE 7 flow cell (USP IV
apparatus, Sotax) with a diameter of 2.2 cm was used in this study. The
USP II apparatus has six vessels and each contained 58 mL of the aqu-
eous dissolution medium (a non-sink condition for the RTV dose) and
50 mL of octanol (a sink condition for the RTV dose) above the aqueous
medium. The aqueous and octanol phases were allowed to saturate with
each other by mixing for 30 min prior to use. The aqueous-octanol in-
terfacial surface area (A) of the vessel with a diameter of 4 cm is ap-
proximately 12.6 cm2, with an aqueous volume of 100 mL (sum of the
volume in flow cell, tubing and vessel), yielding a scaling factor A/V of
0.13.
288
European Journal of Pharmaceutical Sciences 115 (2018) 286–295
Fig. 1. (A) Molecular structure of RTV and (B) its
thermodynamic equilibrium solubility of crystal-
line Forms I and II versus the pH of the buffer.
Each RTV formulation (e.g., tablet, capsule) was loaded into the
flow cell (USP IV apparatus). The flow cells were initially filled with the
FaSSGF (a total volume of 42 mL including the cell and tubing) that was
circulated in a close loop for 30 min at a flow rate of 5 mL/min. This is
considered as “the gastric dissolution stage”. Then, the flow cell was
connected through a loop of Teflon tubing to the aqueous phase of the
USP II vessel containing 58 mL of the FaSSIF-V2 with a higher strength
of 1.71× (pH 6.5); this was to obtain the targeted composition of the
FaSSIF-V2 after mixing with the FaSSGF (Fig. 1B).
A dual paddle mounted on the regular USP II apparatus rotated at
50 rpm in order to provide sufficient mixing in both aqueous and or-
ganic phases. The test was run for 3 h with data collection. This is
considered as “the small intestine stage”. The RTV concentration profile
in octanol was monitored by UV probe (Rainbow®, Pion Inc.) at
240 nm. The aqueous samples were taken by cannulas and syringes and
centrifuged for 5 min at 21,000 ×g force. The supernatant was pipetted,
diluted with equal amount of ethanol, and analyzed for RTV content by
an Agilent 1100 HPLC system. The dissolution vessels (USP II) and flow
cells (USP IV) were housed in their corresponding water baths with the
temperature maintained at 37 ± 0.2 °C.
2.2.3. HPLC Assay
Determination of RTV concentration in the aqueous phase was
performed by HPLC analysis using a 4.6 × 50 mm 2.5 μm XBridgeTM
C18 column (Waters). Samples were eluted with a mobile phase of
mixed buffers (A: water with 0.1% formic acid; B: acetonitrile with
0.1% formic acid; 2%–40% B at 0–1 min; 40%–90% B at 1–7 min). The
flow rate of the mobile phase was 1 mL/min. RTV concentrations were
calculated based on the areas under the HPLC curves at UV wavelength
of 240 nm. Analytical determination of RTV content was based on the
average of three measurements.
3. Results and Discussion
This work applied two biomimetic in vitro tests including the two-
stage dissolution test and the biphasic test to RTV PEG-8000 based ASD
formulations and generic drug products. The goal was to characterize
dissolution and precipitation behaviors of RTV in the aqueous media
and examine the relevance of corresponding in vitro profiles of the RTV
ASD formulations to their pharmacokinetic data in the dog model. The
use of biomimetic media was a key experimental factor of the in vitro
tests. Simulated gastric and intestinal fluids intended for the dog model
were recommended based on the physiology considerations (Grignard
et al., 2017). Upon carefully reviewing, we considered that the
H. Xu et al. European Journal of Pharmaceutical Sciences 115 (2018) 286–295

Fig. 2. Schematic illustration of (A) the two-stage

dissolution test and (B) the biphasic test.

simulated human GI fluids (i.e., the FaSSGF and FaSSIF) were accep- Table 1
table and suitable due to insignificant difference between them and the Biopharmaceutical properties of RTV.
biomimetic media for the dog model.
Description Value

3.1. Part 1: Characterization of RTV PEG-8000 Based ASDs by Two Molecular weight 720.9 g/mol
Log D (pH 6.8) 4.3 (Ref (Law et al., 2004))
Biomimetic Dissolution Methods pKa (weak bases) 1.8, 2.6 (Ref (Meaden et al., 2002))
Solubility in FaSSGF (pH 1.6) ~500 μg/mL
The pH-solubility profile of RTV is shown in Fig. 1B. RTV crystalline Solubility in FaSSIF-V2 (pH 6.5) 7.4 ± 1.1 μg/mL
Form I and II possesses ~400 μg/mL at pH 1.0 and show a significant LLPS in FaSSIF 28 μg/mL
LLPS in FaSSIF-V2 ~40 μg/mL
reduction of solubility decreases to ~1 μg/mL at pH 6.8 (Law et al.,
Solubility in octanol 19 ± 2 mg/mL
2001; Law et al., 2004). Form II has an equilibrium solubility of ~2 μg/
mL at pH 6.8. The equilibrium solubility of crystalline RTV and LLPS
concentrations in biomimetic media (e.g., FaSSGF, FaSSIF, FaSSIF-V2) and a PB mixture of 10% (w/w) crystalline RTV were obtained in the
were determined and are reported in Table 1. FaSSGF and FaSSIF accordingly and are shown in Fig. 3. RTV con-
centration in the samples taken out from the FaSSGF or FaSSIF media
3.1.1. Assessment of Supersaturation, LLPS and Precipitation of the Two- was determined. Due to the coexistence of multiple species including
Stage Dissolution Test the molecularly dissolved drug, drug solubilized by bile salt micelles,
The two-stage dissolution test involved the use of 58 mL FaSSGF the LLPS and solid precipitates (in submicron/μm sizes), the cen-
(pH 1.6) medium initially for 30 min and then an addition of 67 mL trifugation process could not separate the drug present in micelles and
FaSSIF (pH 6.5) (the total volume of the dissolution medium was nano-sized particles from molecularly dissolved drug. Therefore, the
125 mL) (Fig. 2A). This was to mimic the in vivo gastrointestinal pro- RTV concentration determined in this study is referred to as the ap-
cess in the dog model with a pH shift. RTV concentration-time profiles parent concentration. Indeed. the complexity and challenge of analytical
of the PEG-8000 based ASDs containing 10%, 20% and 30% (w/w) RTV determination of the drug present in different formats were well

289
H. Xu et al.
500
450
400
10% ASD
350
20% ASD
300
Concentration (ug / ml)
30% ASD
250
10% PB
200
150
100
50
0 corporated into the RTV-rich colloidal phase (Alhalaweh et al., 2016).
-30 0 30 60 90 120 150 180
Time (min)
Fig. 3. Apparent RTV concentration-time profiles of the PB and ASD formulations ob-
tained in the FaSSGF (time between −30 to 0 min) and the FaSSIF (time between 0 and
180 min) by the two-stage dissolution test. Error bars represented one std. dev. of n = 3.
discussed in (Kanzer et al., 2010; Tho et al., 2010).
Few RTV apparent concentrations were obtained in the FaSSGF
(pH 1.6), or the “gastric stage”, as shown in the time axis of −30 to
0 min in Fig. 3. (The time zero here is referred as to the time when the
FaSSIF was added). Because RTV possesses a pH-dependent solubility,
apparent RTV concentrations of ~120 to ~350 μg/mL were observed
from these three PEG-8000 based ASDs. The ASD with 10% (w/w) RTV
showed a rapid dissolution, reaching an apparent concentration of
~350 μg/mL prior to the pH shift. The 20% and 30% (w/w) RTV ASDs
possessed a noticeably slower dissolution and reached the apparent
concentration of ~200 and ~125 μg/mL, respectively at the end of
gastric stage (or at t = 0 in Fig. 3). The significant difference of the
apparent peak concentration among these PEG-8000 ASDs is indicative
of the difference of their dissolution rate. It is worth noting that these
apparent RTV concentrations from ASDs in the FaSSGF (pH 1.6) ob-
served after 30 min are lower than the RTV solubility in the FaSSGF of
~500 μg/mL (Table 1). These results suggests that RTV in the FaSSGF
was present in the molecularly dissolved state and RTV release is dis-
solution rate limited and governed by the drug loading in the PEG-8000
ASD. The higher RTV load in the ASD resulted in a slower dissolution
rate. In the previous investigation (Law et al., 2004), dissolution pro-
files of these ASDs and the PB mixture were obtained in 900 mL 0.1 N
HCl that provided a sink condition for 50 mg RTV. The rank order of
these dissolution profiles also revealed the significant effect of the drug
loading upon the rate and extent of dissolution (Law et al., 2004).
All three ASDs exhibited a rapid reduction of RTV concentrations to
~50 μg/mL immediately upon addition of the FaSSIF. The 10% (w/w)
ASD maintained a concentration of ~50 μg/mL during the time of
15–60 min while 20% and 30% (w/w) ASDs exhibited lower apparent
concentrations of ~35 and ~25 μg/mL, respectively. Apparent RTV
concentrations of the three ASDs decreased to ~15 μg/mL after 60 min
and remained unchanged during the course of 3 h.
Note that the LLPS concentration of RTV in the FaSSIF was de-
termined to be 28 μg/mL (Table 1). Solubilization of RTV by the mixed
micelles with the bile salts in the FaSSIF contributes to this con-
centration. The presence of small particles (in the submicron and mi-
cron range) in the medium and the apparent RTV concentration of
~50 μg/mL during the time of 15–60 min are indicative of a super-
saturated state sustained by the presence of LLPS generated from the
ASDs. In principle, one would expect to observe RTV concentrations
European Journal of Pharmaceutical Sciences 115 (2018) 286–295
close to its LLPS that is the limit of the RTV concentration in this
medium. The higher apparent concentration of 50 μg/mL than the LLPS
concentration was most likely an overestimation of the RTV con-
centration due to difficulties to remove nanosize particles and micelles
in these samples by centrifugation. Apparent RTV concentrations de-
creased to ~15 μg/mL after 60 min, and this was significantly lower
than that of the LLPS. This is attributed to de-supersaturation by RTV
crystallization in the dissolution medium. Although RTV was reported
to be a slow crystallizer and has a long induction time for crystal nu-
cleation (> 2 h) (Ilevbare et al., 2013; Ilevbare and Taylor, 2013;
Alhalaweh et al., 2016), this observation is consistent with our recent
investigation of RTV in the same dissolution media from its commercial
products (ASD based formulation) (Xu et al., 2017).
Previous investigations revealed a lack of adherence of RTV to the
polymer upon dispersion in aqueous media (Tho et al., 2010) and RTV
LLPS concentrations are not affected by the presence of polymers in-
cluding PVP, PAA, HPMC, HPMCAS, and CAAdP (Alhalaweh et al.,
2016) These facts indicate that the polymer is not substantially in-
210 Presence of the PEG polymer appeared incapable of stabilizing the su-
persaturated state of RTV and maintaining its amorphous solubility as
one expected.
In contrast, the 10% (w/w) PB containing crystalline RTV showed a
low apparent RTV concentration of ~10 μg/mL in the FaSSGF that
decreased to few μg/mL in the FaSSIF at t = 60 min (Fig. 3). The
crystalline RTV substance in the PB exhibited such low apparent con-
centration in the FaSSGF, approximately 10–30 fold lower than those
from PEG-8000 ASDs. This is indicative of an extremely slow dissolu-
tion rate of crystalline RTV in the FaSSGF. The 10% (w/w) PB yielded a
low apparent RTV concentration during the whole course. It is expected
that the PEG polymer in the ASD does not affect the solubility and
dissolution rate of crystalline RTV.
3.1.2. Assessment of Supersaturation, LLPS and Precipitation of the
Biphasic Test
The scaling factor, AI/Va, of the biphasic test is dictated by the
geometry of the vessel (AI) and the volume of the dissolution medium
(Va). Because few sizes of commercial dissolution vessels are available,
a standard vessel of 100 mL was used at the small scale in this study.
The volumes of both aqueous and octanol in the vessel were 59 and
50 mL, respectively, due to practical constraints to maintain a sufficient
aqueous volume and achieve an appropriate height. Although the
partition coefficient, PI of the drug is mainly dictated by its molecular
properties and the composition of the aqueous medium, it may be
slightly affected by the stirring rate due to variation of the diffusion
layer thicknesses (Mudie et al., 2012). Given the fixed experimental
variables of the biphasic test, the PI of the drug is pre-defined rather
than an adjustable parameter.
The hydrodynamic conditions (e.g., the flow rate and dual paddle
speed) employed in the biphasic test play a significant role in the dis-
solution and partition kinetics affecting drug concentration-time pro-
files in both aqueous and octanol phases as we revealed in previous
reports (Xu et al., 2018; Mudie et al., 2012; Xu et al., 2017) In this
work, the hydrodynamic condition was set to employ a dual paddle
speed of 50 RPM and a flow rate of 5 mL/min based on our existing
knowledge. No exploration and/or optimization of the hydrodynamic
condition aimed to the IVIVR were conducted in this study. Multiple
studies of the biphasic test at the small scale (most are unpublished)
obtained in our laboratory concluded that the current experimental
parameter package is optimal due to good correlation with PK profiles
of several poorly soluble drugs observed in dogs (Shi et al., 2016).
The three RTV ASDs and the 10% PB were evaluated by the biphasic
test with sequential use of the FaSSGF and FaSSIF-V2 media. As shown
in Fig. 4A, apparent RTV concentrations of the three ASDs in the
FaSSIF-V2 were similar around 28 to 33 μg/mL at 15 min after mixing
of the FaSSGF (pH 1.6) with the FaSSIF-V2 (pH 6.5). Associated with
290
H. Xu et al. European Journal of Pharmaceutical Sciences 115 (2018) 286–295

A Aqeous phase accumulation of RTV in octanol. The change of the flux observed is
45 worth commenting. The high flux is presumably associated with a
highly supersaturated state of RTV upon immediate mixing of the
40
FaSSGF with the FaSSIF-V2 (pH shift) that dictates a rapid de-super-
35 saturation by partition accompanied by precipitation. Later, the supply
Concentration (ug / ml)

10% ASD of the RTV from the aqueous medium is dictated by the dissolution rate
30
20% ASD of undissolved ASDs and precipitates. Similar flux values of the ASDs
25 were observed among 10%, 20% and 30% drug loading after 90 min.
30% ASD
20 This implies that the similar dissolution rate was mainly associated with
10% PB RTV precipitates rather than the undissolved ASDs (if there was a small
15
amount of original ASDs remaining). The high flux of the 10% ASD
10 exhibited a superior performance, reaching a RTV concentration of
~270 μg/mL in octanol at 3 h, corresponding to 13.5 mg RTV. The 20%
5
and 30% ASDs yielded RTV concentrations of 160 and 70 μg/mL in
0 octanol at 3 h, respectively, corresponding to 8 and 3.5 mg RTV, re-
0 30 60 90 120 150 180 210 spectively. In contrast, the 10% PB exhibited a constantly low flux and
yielded a concentration of 25 μg/mL in octanol at 3 h. This corresponds
Time (min) to only 1.25 mg RTV in octanol.

3.1.3. Comparison of the Two Biomimetic Dissolution Tests

B. Octanol phase As described above, the two-stage dissolution test revealed that
350 apparent RTV concentrations of the three ASDs in the FaSSIF were
maintained at ~50 μg/mL between 10 and 60 min after precipitation
300 due to pH shift (Fig. 3). These values are significantly higher than the
Concentration (ug / ml)

LLPS of 28 μg/mL in the FaSSIF, suggesting an overestimation of the

250
200
150
100
50
0 The difference of the apparent RTV concentrations between the two
0 30 60 90 120 150 180 210
Time (min)
Fig. 4. (A) Apparent RTV concentration-time profiles in the FaSSIF-V2 and (B) cumula-
tive RTV concentration-time profiles in octanol of the PB and ASDs by the biphasic test.
Error bars represented one std. dev. of n = 3.
the increase of pH value of the medium, RTV concentrations gradually
decreased to ~10 μg/mL at 2 h and remained unchanged. During the
course of 3 h, apparent RTV concentrations were lower than the LLPS in
FaSSIF-V2 (~40 μg/mL, Table 1), suggesting a lower degree of super-
saturation. The gradual decrease of RTV concentration is presumably
attributed to partition into octanol. In contrast, the 10% PB showed a
consistently low apparent RTV concentration of ~5 μg/mL during the
whole course of the test. It is worth noting that this concentration is
slightly lower than its solubility of 7 μg/mL in the FaSSIF-V2 (Table 1).
The RTV concentration-time profiles of the three ASDs in octanol
are plotted in Fig. 4B. These profiles exhibited a lag time of ~20–30 min
before RTV concentrations increased substantially. This lag time is most
likely associated with RTV diffusion through the water-octanol inter-
face coupled with the hydrodynamic convection. The flux of RTV of
these formulations reflected by the corresponding slope of the con-
centration –time profile within the time of 30–90 min shows a trend
below:
10%ASD > 20%ASD > 30%ASD > > 10%PB
As the flux of RTV into octanol is solely driven by the free RTV
fraction in the FaSSIF-V2, this trend is indicative of the difference of
RTV dissolution rate among these formulations. The flux of all three
ASDs significantly decreased after ~90 min (Fig. 4B) with slow
RTV concentration in the medium due to difficulties to separate nano-
size particles including LLPS and micelles from the molecularly dis-
solved drug. In comparison, apparent RTV concentrations of these ASDs
in the FaSSIF-V2 by the biphasic test are noticeably low (~15 to 30 μg/
mL compared to the LLPS of 40 μg/mL in FaSSIF-V2) between 10 and
60 min accompanied by steady accumulation of RTV in octanol (Fig. 4).
The decrease of the RTV apparent concentration in the aqueous
medium is primarily driven by partition into octanol rather than the
precipitation.
biomimetic dissolution tests is of interest and deserves comment. The
two-stage dissolution test appears to overestimate the degree and
duration of supersaturation of RTV and this is attributed to the ex-
perimental methodology. That is, the maximum apparent RTV con-
centration is dictated by the amorphous solubility (or the LLPS con-
centration) and the portion of dissolved RTV exceeding the amorphous
solubility is removed through precipitation. Thus, the same apparent
RTV concentration was essentially observed for all three ASDs. Thus, it
is difficult to differentiate them by such test. In contrast, the biphasic
test employs the octanol phase acting as an “absorption compartment”.
A constant removal of dissolved RTV through partition regardless of the
supersaturation state occurs simultaneously as the dissolution and
precipitation take place. As a result, the complex dynamic equilibria
among dissolution, precipitation and partition are established in the
aqueous media during the entire course (Shi et al., 2016; Xu et al.,
2018; Frank et al., 2014b). A lower degree of apparent supersaturation
of RTV of these ASDs in the aqueous media evidenced by the low ap-
parent concentration is a realistic reflection of such dynamics. Further,
the performance of these ASDs can be readily differentiated by direct
measurement of the amount of RTV in octanol that is directly associated
with the free drug concentration in the aqueous medium.
Amidon et al. have repeatedly pointed out that the single aqueous
based dissolution tests with pH shift under the non-sink condition, in
general, may lead to an overestimation of the degree and duration of
supersaturation of poorly water soluble drugs (Tsume et al., 2014;
Takeuchi et al., 2014; Tsume et al., 2017). In particular, the BCS class II
drugs (with high permeability and low solubility) will be quickly ab-
sorbed upon dissolution and the drug concentration in the GI lumen
would be lower than one observed in the in vitro dissolution vessels
(Hens et al., 2015; Bevernagea et al., 2013; Psachoulias et al., 2012;
Psachoulias et al., 2011). Introduction of “the absorption compartment”

291
H. Xu et al. European Journal of Pharmaceutical Sciences 115 (2018) 286–295

16

14

12

AUC(µg.hr/mL) in dog model

10

6
y = 0.9016x + 1.0776
4 R² = 0.9428

0
0 5 10 15

AUC (mg.hr/mL) in two-stage

dissolution test

15

Fig. 5. Mean plasma concentration–time profiles of RTV after a single oral dose to beagle 12
AUC(µg.hr/mL) in dog model

dogs. RTV was administered in (a) 10% PB, or PEG-8000 based ASDs containing (b) 10%,
(c) 20%, and (d) 30% (w/w) RTV.
Permission from Ref. (Law et al., 2004).
9

such as the biphasic system would be a superior approach for char-

acterizing supersaturation as this method mimics all key dynamic
processes in vivo and the concentration profiles obtained in the organic
6
y = 0.0519x + 1.1262
phase are indirect measurement of the free drug available for partition. R² = 0.9124

3
3.1.4. Establishment of IVIVR for Both the Two-Stage Dissolution and
Biphasic
Mean plasma concentration-time profiles of RTV from the three
ASDs of 10%, 20%, 30% (w/w) RTV and the 10% (w/w) PB after a 0
single oral dose to beagle dogs at a dose of 50 mg were reported and are 0 100 200 300
shown here in Fig. 5 for readers' benefit and clarity (Law et al., 2004). A RTV conc.(µg/mL) in octanol at 180 min
rapid increase of the mean plasma concentration of RTV within the first
1 h and corresponding high Cmax values of RTV are characteristic of the Fig. 6. IVIVR plot of the in vivo AUC of RTV from the PB and ASDs in dogs against (A) the
ASDs as compared to the counterparts of 10% (w/w) PB with crystalline corresponding in vitro AUC by the two-stage dissolution test and (B) the RTV con-
centration in octanol at 3 h by the biphasic test.
RTV. The rapid absorption of RTV in dogs under the fasting state is
clearly driven by the supersaturation of RTV as revealed by the two-
stage dissolution test and indicates the critical role of supersaturation Table 2
on in vivo exposure. Apparent RTV concentration profiles from the two- PEG-8000 based ASD and physical blend (PB) formulations of RTV and their corre-
sponding PK data in dogs under the fasting state.
stage dissolution test represent the driving force for in vivo absorption.
(From ref. (Law et al., 2004)).
Therefore, the mean in vitro AUC from each formulation was integrated
between 0 and 180 min and used for IVIVR. Fig. 6A shows a plot of in Formulation Strength Description AUC (0-∞) Cmax (mg/mL)
vivo AUCs of RTV from these three ASD formulations and the PB for- (mg/unit) (μg h/mL)
mulation (Table 2) against their corresponding in vitro AUCs. A linear
10% ASD 50 mg PEG-8000 ASD 14.04 ± 1.49 5.48 ± 0.64
relationship is obtained with R2 of 0.94. with 10% RTV
Because the RTV concentration in octanol is directly related to the 20% ASD 50 mg PEG-8000 ASD 11.46 ± 1.75 4.42 ± 0.80
free drug available for partition, the concentration observed at 3 h re- with 20% RTV
presents the amount partitioning into “the absorptive compartment” 30% ASD 50 mg PEG-8000 ASD 7.00 ± 1.03 3.12 ± 0.30
with 30% RTV
from each formulation. Fig. 6B exhibits a linear relationship between
10% PB 50 mg PEG-8000 with 0.64 ± 0.34 0.40 ± 0.24
the in vivo AUCs in dogs and the RTV concentrations of these ASD 10% crystalline
formulations and the PB formulation in octanol at 3 h with R2 of 0.91. RTV

3.2. Part 2: Evaluation of RTV Generic Drug Products with the Biphasic Test
examined against the Norvir® capsule using an internally developed dog
model (Garren et al., 2010). The results for 12 of 16 experimental RTV
In the previous study, relative bioavailability of RTV products
ASD formulations in dogs were found to be predictive of relative
manufactured by various generic pharmaceutical companies was

292
H. Xu et al. European Journal of Pharmaceutical Sciences 115 (2018) 286–295

Fig. 7. Mean plasma concentrations of RTV fol-

lowing an oral dose of 100 mg in dogs under the
fasting state. A: Empetus (Emcure); B: Empetus
(Hetero); C: Ritomune (Cipla) vs. the common
reference product Norvir® capsule.
Permission from ref. (Garren et al., 2010).

exposures observed in healthy human subjects with respect to predic- A. Aqueous phase
tion of bioequivalence and non-bioequivalence. 25
Three commercial generic RTV tablet products manufactured by Emcure
Emcure, Hetero and Cipla were examined by the biphasic test at small 20
Hetero

Concentration (ug / ml)

scale under the same experimental conditions (the flow rate, paddle Cipla
speed, volume of the biomimetic media and octanol, etc.) described in
15
Part 1. However, we have no knowledge of the formulation composi-
tion, the physical state of RTV substance, and the manufacturing pro-
10
cess of these generic products. Although the Norvir® capsule was used
as the reference product in the previous study, it is unavailable and not
5
tested in this work. The pharmacokinetic profiles of these generic RTV
drug products reported are shown in Fig. 7 for readers' benefit and
convenience for discussion (Garren et al., 2010). 0
0 30 60 90 120 150 180 210
Apparent RTV concentration-time profiles of the three generic
products were obtained in the FaSSIF-V2 following exposure to the Time (min)
FaSSGF and are shown in Fig. 8A. The profiles of the Hetero and Cipla
tablets are similar with an initial RTV concentration of ~23 μg/mL and B. Octanol phase
then gradually decreasing to ~10 μg/mL at 3 h. In contrast, the Emcure 160
tablet exhibited a low apparent concentration of ~7 μg/mL initially,
and this may be related to a slow disintegration that was visually ob- 140
Concentration (ug / ml)

served during the experiment. It was surprising that the apparent RTV
120
concentrations of the Emcure tablet gradually increased from 7 to
~12 μg/mL at 3 h. It is worth noting that this profile has been verified 100
and is reproducible. This is opposite to our common observation that
RTV apparent concentrations in such aqueous media decrease due to 80
precipitation and partition (Xu et al., 2017; Tho et al., 2010). It is
60
difficult to offer any hypothesis to account for an increase of RTV
concentration without any knowledge of the Emcure tablet. 40
RTV concentration–time profiles of the three generic tablet products
in octanol are shown in Fig. 8B. These profiles all show a lag time of 20
~60 min, significantly longer than those from PEG-8000 based ASDs
0
(Fig. 4B). The profile of the Hetero tablet shows a significantly higher 0 30 60 90 120 150 180 210
flux and a higher concentration of 120 μg/mL at 3 h than those of the
Cipla tablet (85 μg/mL at 3 h), corresponding to 6 mg and 4.25 mg in
octanol at 3 h, respectively. The Emcure tablet exhibited a constant and Time (min)
low flux, reaching 25 μg/mL at 3 h, resulting in 1.25 mg in octanol. It is Fig. 8. (A) Apparent RTV concentration-time profiles in the FaSSIF-V2 and (B) cumula-
worth noting that the aqueous concentration-time profiles of three tive RTV concentration-time profiles in octanol of the three commercial generic RTV
generic RTV tablet products (100 mg dose) (Fig. 8A) show similarities products by the biphasic test. Error bars represented one std. dev. of n = 3.
to those of the PEG-8000 based ASDs (50 mg dose) (Fig. 4A). First,
generation of a low degree of supersaturation of RTV in the aqueous and AUC than those of the Cipla tablet. In contrast, the Empure tablet
medium from the Hetero and Cipla tablets is indicated by their con- showed an extremely low Cmax and AUC against the Norvir® capsule.
centrations below the LLPS concentration of 40 μg/mL. Secondly, these The RTV concentration in octanol at 3 h represents the amount in
profiles show a gradual decrease during the course of test (3 h). Al- “the absorption compartment” in vitro. The relative F% of the three
though the aqueous profiles of Cipla and Hetero tablets overlapped generic RTV products observed in dogs (reported in Table 3) is plotted
without differentiation (Fig. 8A), a significantly higher flux was ob- against their corresponding RTV concentration in octanol at 3 h and
served in octanol from the Hetero tablet than that of the Cipla tablet, shown in Fig. 9. A linear relationship with a correlation coefficient of
leading to a significantly higher RTV concentration at 3 h (Fig. 8B). The 0.97 was obtained, indicative of good IVIVR.
difference of the flux values among these RTV products is consistent As we briefly described in the Introduction section, the mechanistic
with their pharmacokinetic characteristics shown in Fig. 7. Note that model of the two-phase dissolution system indicates that the physio-
the Hetero tablet showed a faster absorption in vivo with a higher Cmax logical relevance of this method is based on that the in vitro

293
H. Xu et al. European Journal of Pharmaceutical Sciences 115 (2018) 286–295

Table 3 8000 based ASDs (with 50 mg dose) and up to ~82 mg from the Hetero
Commercial generic RTV products tested in this work and their relative bioavailability in tablet (with 100 mg dose and the relative F% in Table 3) within a time
dogs under the fasting state (Garren et al., 2010).
course of 3 h (comparable to the GI transition time in dogs). However,
Manufacturer Strength (mg/ Product description Relative F% only 13.5 mg RTV in octanol from 10% (w/w) PEG-8000 ASD and 6 mg
unit) (AUC) RTV in octanol from the Hetero tablet are obtained at 3 h; these
amounts are about 3.5 to 14 times less than the amounts desired based
Emcure 100 Tablet under the name of 1⁎
on the extent of absorption. These data indicate that the in vitro par-
Empetus
Hetero 100 Tablet under the name of 87⁎ tition rate, (AI/Va)PI, of RTV at the small scale is significantly lower
Empetus than the in vivo absorption rate. The in vitro PI at the small scale is
Cipla 100 Tablet under the name of 36⁎ estimated to be somewhat smaller than that at the large scale due to the
Ritomune hydrodynamic conditions employed in the biphasic test. While the in
⁎ vivo scaling factor, A/V, in humans is estimated to be ~2 (Mudie et al.,
The value was determined against the AUC of a reference product, Norvir® capsule.
2012), it is estimated to be ~7 in dogs (Anon, n.d.) that is ~50 folder
larger than the in vitro scaling factor, AI/Va, of 0.13 used in this study.
100
Assuming that the dog and human subjects have a similar in vivo Peff
value of RTV (Dahlgren et al., 2016), such analysis suggests that the
y = 0.8831x - 24.288 scaling factor at the small scale is most likely the primary factor re-
R² = 0.9675
80 sponsible for the low amount of RTV partitioning into octanol during
the course of 3 h. Without a comprehensive mathematical model to
account for the complex, dynamic processes encountered in the bi-
In vivo relative F%

phasic test, the preliminary results may suggest that the in vitro parti-
60 tion rate, kp, of RTV at the small scale is likely significantly smaller than
the in vivo absorption rate, ka in dogs.

40
20
0 RTV was observed from these ASD formulations accompanied by the
0 50 100
RTV conc. in octanol at 3 h
Fig. 9. The relative F% of RTV observed in dogs from three commercial generic products
against the RTV concentration in octanol at 3 h by the biphasic test at the small scale.
partitioning rate coefficient, kp (equal to (AI/Va) ∗ PI) approximates the
in vivo absorption rate coefficient, ka (equal to (A/V) ∗ Peff) (Mudie
et al., 2012). The biphasic test may involve a scaling factor, AI/Va, in
the range of 0.1 to 0.5 due to experimental constrains; that is ap-
proximately an order of magnitude smaller than the corresponding
physiological A/V value in vivo (Mudie et al., 2012). In contrast, the PI
of drug substances from water to octanol is reported to be in a wide
range of 2–30 × 10−4 cm/s (pending on the hydrodynamic condition)
with several BCS II drugs including piroxicam, nimesulide, valproic
acid, ibuprofen and felodipine (see Tables 3 and 7 in Mudie et al.
(2012)). These PI values are roughly an order of magnitude larger than
the typical permeability Peff in vivo reported (Mudie et al., 2012).
Therefore, the biphasic test may possess an in vitro kp of a given drug
that would be similar (or equal) to its in vivo ka. Indeed, our recent
investigation of three commercial RTV products by the biphasic test at
the large scale (intended to simulate humans) with the use of same
biomimetic media confirms (1) a good matching between the partition
profiles in vitro and the absorption profiles in vivo and, (2) ~50–70% of
RTV dose partitioning into octanol comparable to the absorbable
amount in human (Xu et al., 2017).
The current experimental set up of the biphasic test at the small scale
is designed to simulate the GI situation in dogs. The absolute bioa-
vailability of RTV in beagle dogs was reported to be approx. 100%
(Kempf et al., 1995; Hsu et al., 1998). Assuming that the in vitro bi-
phasic test simulates the dissolution-absorption in dogs, one would
expect up to ~47 mg RTV partitioning into octanol from the 10% PEG-
4. Conclusion
In this study, three PEG-8000 based ASD containing 10%, 20% and
30% RTV and 10% PB containing crystalline drug were evaluated using
two biomimetic in vitro dissolution tests. As RTV is a weak base, the pH
alteration of the aqueous media involved in both dissolution tests is of
critical importance. When RTV was dissolved in the FaSSGF (pH 1.6)
and transited to the FaSSIF or FaSSIF-V2 (pH 6.5), supersaturation of
150 LLPS and significant amount of precipitation. In vitro AUCs of these
apparent RTV concentration-time profiles of the PB and three ASD
formulations obtained by the two-stage dissolution test are integrated
and show good agreement with their corresponding AUCs in dogs.
The biphasic test at the small scale effectively differentiates the PB
and PEG-8000 based ASDs of RTV as well as three RTV commercial
generic tablet products. The degree of supersaturation of RTV from
these three PEG-8000 ASDs revealed by the biphasic test appears lower
than those from the two-stage dissolution test. RTV concentrations of
the PB and three ASDs as well as the three generic tablet products
observed in octanol at 3 h show agreement with their in vivo AUCs in
dogs.
Aqueous profiles of RTV obtained by these two biomimetic dis-
solution tests are complimentary in nature, providing insight regarding
to the dynamic interplay among dissolution, precipitation and parti-
tioning kinetics. Although these two dissolution tests are developed
based on different concepts and involve different experimental appa-
ratus and parameters, both methods demonstrate its capability for
achieving IVIVR. It is worth noting that the presence of “the absorption
compartment” with the biphasic test provides insight of the dynamic
relationship among dissolution, precipitation and partitioning kinetics
that is difficult to obtain otherwise.
The experimental set up and the hydrodynamic condition of the
biphasic test have a significant effect upon drug concentration profiles
in both aqueous and octanol phases. Preliminary results may suggest
that the in vitro partition rate, kp, of RTV in the biphasic test at the
small scale is significantly (maybe 3–15 times) smaller than the in vivo
absorption rate, ka in dogs. This warrants further investigation and
potential improvement of the biphasic test at the small scale.
Acknowledgements and Disclosure
We thank Drs. Eric Schmitt and Devalina Law of AbbVie for their

294
H. Xu et al. European Journal of Pharmaceutical Sciences 115 (2018) 286–295

enlightening discussions on this project. We also thank Socrates Vela formulations in aqueous medium assessed by asymmetrical flow field-flow fractio-
and Weili Wang of AbbVie for technical assistance. H.X. thanks AbbVie nation. J. Pharm. Biomed. Anal. 53, 359–365.
Kempf, D.J., et al., 1995. ABT-538 is a potent inhibitor of human immunodeficiency virus
for their financial support of the AbbVie Postdoctoral Research protease and has high oral bioavailability in humans. Proc. Natl. Acad. Sci. 92,
Program. 2484–2488.
This study was funded by AbbVie Inc. AbbVie participated in the Kostewicz, E.S., Abrahamsson, B., Brewster, M., Brouwers, J., Butler, J., Carlert, S.,
Dickinson, P.A., Dressman, J., Holm, R., Klein, S., Mann, J., McAllister, M., Minekus,
study design, research, data collection, analysis and interpretation of M., Muenster, U., Mullertz, A., Verwei, M., Vertzoni, M., Weitschies, W., Augustijns,
data, as well as writing, reviewing, and approving the publication. All P., 2014. In vitro models for the prediction of in vivo performance of oral dosage
authors are AbbVie employees and may own AbbVie stock. forms. Eur. J. Pharm. Sci. 57, 342–366.
Law, D., Krill, S.L., Schmitt, E.A., Fort, J.J., Qiu, Y., Wang, W., Porter, W.R., 2001.
Physicochemical considerations in the preparation of amorphous ritonavir-poly
References (ethylene glycol) 8000 solid dispersions. J. Pharm. Sci. 90 (8), 1015–1025.
Law, D., Schmitt, E.A., Marsh, K.C., Everitt, E.A., Wang, W., Fort, J.J., Krill, S.L., Qiu, Y.,
2004. Ritonavir–PEG 8000 amorphous solid dispersions: in vitro and in vivo eva-
Alhalaweh, A., Bergström, C.A.S., Taylor, L.S., 2016. Compromised in vitro dissolution
luations. J. Pharm. Sci. 93 (3), 563–570.
and membrane transport of multidrug amorphous formulations. J. Control. Release
Locher, K., Borghardt, J.M., Frank, K.J., Kloft, C., Wagner, K.G., 2016. Evolution of a
229, 172–182.
mini-scale biphasic dissolution model: impact of model parameters on partitioning of
Alsenz, J., Steffen, H., Alex, R., 1998. Active apical secretory efflux of the HIV protose
dissolved API and modelling of in vivo-relevant kinetics. Eur. J. Pharm. Biopharm.
inhibitos saquinavir and ritonavir in Caco-2 cell monolayers. Pharm. Res. 15 (3),
105, 166–175.
423–428.
Lu, E., Li, S., Wang, Z., 2017. Biorelevant test for supersaturable formulation. Asian J.
Anon n.d. The in vivo scaling factor, A/V, in dogs is estimated to be in the range of 4 to 26
Pharm. Sci. 12, 9–20.
cm-1 as reported in Deanna Mudie Ph.D. Thesis, University of Michigan, 2014. The
Meaden, E.R., Hoggard, P.G., Newton, P., Tjia, J.F., Aldam, D., Cornforth, D., Lloyd, J.,
average of A/V in dogs is estimated to be 6.8 cm-1 based on the degree of flatness of
Williams, I., Back, D.J., Khoo, S.H., 2002. P-glycoprotein and MRP1 expression and
1.7 (from K. Sugano). Theoretical investigation of passive intestinal membrane per-
reduced ritonavir and saquinavir accumulation in HIV-infected individuals. J.
meability using Monte Carlo method to generate drug-like molecule population. Int.
Antimicrob. Chemother. 50, 583–588.
J. Pharm. 373: 55–61 (2009). Estimation of a scaling factor, A/V, in dogs from private
Mudie, D.M., Shi, Y., Ping, H., Gao, P., Amidon, G.L., Amidon, G.E., 2012. Mechanistic
communication with Drs. Deanna Muddie and Greory Amidon.
analysis of solute transport in an in vitro physiological two-phase dissolution appa-
Aungst, B.J., Nguyen, N.H., Bulgarelli, J.P., Oates-Lenz, K., 2000. The influence of donor
ratus. Biopharm. Drug Dispos. 33 (7), 378–402.
and reservoir additives on Caco-2 permeability and secretory transport of HIV pro-
Newman, A., Gregory, K., Zografi, G., 2012. Assessing the performance of amorphous
tease inhibitors and other lipophilic compounds. Pharm. Res. 17 (10), 1175–1180.
solid dispersions. J. Pharm.Sci. 101 (4), 1355–1377.
Baghel, S., Cathcart, H., O'Reilly, N.J., 2016. Polymeric amorphous solid dispersions: a
Nguyen, M.A., Flanagan, T., Brewster, M., Kesisoglou, F., Beato, S., Biewenga, J., Crison,
review of amorphization, crystallization, stabilization, solid-state characterization,
J., Holmg, R., Li, R., Mannaert, E., McAllister, M., Mueller-Zsigmondy, M., Muenster,
and aqueous solubilization of biopharmaceutical classification system class II drugs.
U., Ojala, K., Page, S., Parr, A., Rossenu, S., Timmins, P., Van Peer, A., Vermeulen, A.,
J. Pharm. Sci. 105, 2527–2544.
Langgutha, P.A., 2017. Survey on IVIVC/IVIVR development in the pharmaceutical
Bevernagea, J., Brouwersa, J., Brewsterb, M.E., Augustijns, P., 2013. Evaluation of gas-
industry – past experience and current perspectives. Eur. J. Pharm. Sci. 102, 1–13.
trointestinal drug supersaturation and precipitation: strategies and issues. Int. J.
Pestieau, A., Evrard, B., 2017. In vitro biphasic dissolution tests and their suitability for
Pharm. 453, 25–35.
establishing in vitro-in vivo correlation: a historical review. Eur. J. Pharm. Sci. 102,
Brouwers, J., Brewster, M., Augustijns, P., 2009. Supersaturating drug delivery systems:
203–219.
the answer to solubility-limited oral bioavailability? J.Pharm. Sci. 98 (8), 2549–2572.
Psachoulias, D., Vertzoni, M., Goumas, K., Kalioras, V., Beato, S., Butler, J., Reppas, C.,
Dahlgren, D., Roos, C., et al., 2016. Regional intestinal permeability in dogs: bio-
2011. Precipitation in and supersaturation of contents of the upper small intestine
pharmaceutical aspects for development of oral modified-release dosage forms. Mol.
after administration of two weak bases to fasted adults. Pharm. Res. 28, 3145–3158.
Pharm. 13, 3022–3033.
Psachoulias, D., Vertzoni, M., Butler, J., Busby, D., Symillides, M., Dressman, J., Reppas,
Frank, K.J., et al., 2014a. What is the mechanism behind increased permeation rate of a
C., 2012. An in vitro methodology for forecasting luminal concentrations and pre-
poorly soluble drug from aqueous dispersions of an amorphous solid dispersion? J.
cipitation of highly permeable lipophilic weak bases in the fasted upper small in-
Pharm. Sci. 103 (6), 1779–1786.
testine. Pharm. Res. 29, 3486–3498.
Frank, K.J., Locher, K., Zecevic, D.E., Fleth, J., Wagner, K.G., 2014b. In vivo predictive
Shi, Y., Gao, P., Gong, Y., Ping, H., 2010. Application of a biphasic test for character-
mini-scale dissolution for weak bases: advantages of pH-shift in combination with an
ization of in vitro drug release of immediate release formulations of celecoxib and its
absorptive compartment. Eur. J. Pharm. Sci. 61, 32–39.
relevance to in vivo absorption. Mol. Pharm. 7 (5), 1458–1465.
Gao, P., Morozowich, W., 2005. Development of supersaturatable SEDDS (S-SEDDS)
Shi, Y., Erickson, B., Jayasankar, A., Lu, L., Marsh, K., Menon, R., Gao, P., 2016. Assessing
formulations for improving the oral absorption of poorly soluble drugs. Exp. Opin.
supersaturation and its impact on in vivo bioavailability of a low-solubility com-
Drug Del. 3, 97–110.
pound ABT-072 with a dual pH, two-phase dissolution method. J. Pharm. Sci. 105
Gao, P., Shi, Y., 2012. Characterization of supersaturatable formulations for improved
(9), 2886–2895.
absorption of poorly soluble drugs. AAPS J. 14 (4), 703–713.
Takeuchi, S., Tsume, Y., Amidon, G.E., Amidon, G.L., 2014. Evaluation of a three com-
Gao, P., Shi, Y., Miller, J.M., 2016. Development and application of in vitro two-phase
partment in vitro gastrointestinal simulator dissolution apparatus to predict in vivo
dissolution method for poorly water soluble drugs. In: Webster, G., Bell, R., Jackson,
dissolution. J. Pharm. Sci. 103 (11), 3416–3422.
D. (Eds.), A book Chapter in “Poorly Soluble Drugs: Dissolution and Drug Release”.
Tho, I., Liepold, B., Rosenberg, J., Maegerlein, M., Brandl, M., Fricker, G., 2010.
Pan Stanford Publishing.
Formation of nano/micro-dispersions with improved dissolution properties upon
Garren, K.W., Rahim, S., Marsh, K., Morris, J.B., 2010. Bioavailability of generic ritonavir
dispersion of ritonavir melt extrudate in aqueous media. Eur. J. Pharm. Sci. 40,
and lopinavir/ritonavir tablet products in a dog model. J. Pharm. Sci. 99 (2),
25–32.
626–631.
Tsume, Y., Mudie, D.M., Langguth, P., Amidon, G.E., Amidon, G.L., 2014. The bio-
Grignard, E., Taylor, R., McAllister, M., Box, K., Fotaki, N., 2017. Considerations for the
pharmaceutics classification system: subclasses for in vivo predictive dissolution
development of in vitro dissolution tests to reduce or replace preclinical oral ab-
(IPD) methodology and IVIVC. Eur. J. Pharm. Sci. 57 (152–16).
sorption studies. Eur. J. Pharm. Sci. 99, 193–201.
Tsume, Y., Matsui, K., Searls, A.L., Takeuchi, S., Amidon, G.E., Sun, D., Amidon, G.L.,
Hens, B., Brouwers, J., Corsetti, M., Augustijns, P., 2015. Supersaturation and pre-
2017. The impact of supersaturation level for oral absorption of BCS class IIb drugs,
cipitation of posaconazole upon entry in the upper small intestine in human. J.
dipyridamole and ketoconazole, using in vivo predictive dissolution system: gastro-
Pharm. Sci. 105 (9), 2677–2684.
intestinal simulator (GIS). Eur. J. Pharm. Sci. 102, 126–139.
Hsu, A., Granneman, G.R., Bertz, R.J., 1998. Ritonavir clinical pharmacokinetics and
Vangani, S., Zhou, P., Del-Barrio, M., Chiu, R., Cauchon, N., Gao, P., Medina, C., Jasti, B.,
interactions with other anti-HIV agents. Clin. Pharm. 35 (4), 275–291.
2009. Dissolution of poorly water-soluble drugs in biphasic media using USP 4 and
Ilevbare, G.A., Taylor, L.S., 2013. Liquid-liquid phase separation in highly supersaturated
fiber optic system. Clin. Res. Reg. Affairs. 26 (1–2), 8–19.
aqueous solutions of poorly water-soluble drugs: implications for solubility enhan-
Xu, H., Vela, S., Shi, Y., Morroum, P., Gao, P., 2017. In vitro characterization of ritonavir
cing formulations. Cryst. Growth Des. 13 (4), 1497–1509.
commercial drug products by biorelevant dissolution methods with relevance to their
Ilevbare, G.A., Liu, H., Pereira, J., Edgar, K.J., Taylor, L.S., 2013. Influence of additives on
in vivo performance. Mol. Pharm. 14, 3801–3814.
the properties of nanodroplets formed in highly supersaturated aqueous solutions of
Xu, H., Vela, S., Shi Y., Marroum, P., Gao, P., 2018. Developing quantitative in vitro-in
ritonavir. Mol. Pharm. 10 (9), 3392–3403.
vivo correlation (IVIVC) for fenofibrate immediate release formulations with the
Kanzer, J., Hupfeld, S., Vasskog, T., Tho, I., Hölig, P., Mägerlein, M., Fricker, G., Brandl,
biphasic dissolution-partition test method. J. Pharm. Sci. 107 (1), 476–487.
M., 2010. In situ formation of nanoparticles upon dispersion of melt extrudate

295