Advanced Drug Delivery Reviews 101 (2016) 122–142

Contents lists available at ScienceDirect

Advanced Drug Delivery Reviews

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

Physical chemistry of supersaturated solutions and implications for

oral absorption☆
Lynne S. Taylor a,⁎, Geoff G.Z. Zhang b
a
Department of Industrial and Physical Pharmacy, Purdue University, West Lafayette, IN, 47907, USA
b
Drug Product Development, Research and Development, AbbVie Inc., North Chicago, IL, 60064, USA

a r t i c l e i n f o a b s t r a c t

Article history: Amorphous solid dispersion (ASD) formulations are widely used for delivery of poorly soluble drugs for dissolu-
Received 29 January 2016 tion enhancement and bioavailability improvement. When administered, ASDs often exhibit fast dissolution to
Received in revised form 10 March 2016 yield supersaturated solutions. The physical chemistry of these supersaturated solutions is not well understood.
Accepted 11 March 2016
This review will discuss the concepts of solubility, supersaturation, and the connection to membrane transport
Available online 22 March 2016
rate. Liquid–liquid phase separation (LLPS), which occurs when the amorphous solubility is exceeded, leading
Keywords:
to solutions with interesting properties is extensively discussed as a phenomenon that is relevant to all enabling
Solubility formulations. The multiple physical processes occurring during dissolution of the ASD and during oral absorption
Amorphous are analyzed. The beneficial reservoir effect of a system that has undergone LLPS is demonstrated, both
Liquid–liquid phase separation experimentally and conceptually. It is believed that formulations that rapidly supersaturate and subsequently un-
Membrane transport rate dergo LLPS, with maintenance of the supersaturation at this maximum value throughout the absorption process,
Solubility enhancement i.e. those that exhibit “spring and plateau” behavior, will give superior performance in terms of absorption.
Absorption © 2016 Elsevier B.V. All rights reserved.

Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
2. Theoretical considerations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
2.1. Crystalline solubility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
2.2. Supersaturation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
2.3. Amorphous solubility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
2.4. Thermodynamics of mixing and liquid–liquid phase separation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
2.5. Liquid–liquid phase separation versus crystallization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
3. Experimental observations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
3.1. Experimental determination of amorphous solubility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
3.2. Methods to detect and characterize LLPS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
3.2.1. Light scattering techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
3.2.2. Fluorescence spectroscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
3.2.3. Other characterization approaches . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
3.3. The amorphous solubility advantage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
3.4. Additional factors impacting the experimentally observed amorphous solubility . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
3.4.1. Ionic strength . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
3.4.2. pH . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
3.4.3. Additional solutes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
3.5. Membrane transport and supersaturated solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
3.6. The amorphous solubility as the upper limit of supersaturation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
3.7. Solubilization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
3.8. Effect of solubilizing additives on amorphous solubility, supersaturation and flux . . . . . . . . . . . . . . . . . . . . . . . . . . . 132

☆ This review is part of the Advanced Drug Delivery Reviews theme issue on “Understanding the challenges of beyond-rule-of-5 compounds.”
⁎ Corresponding author at: Department of Industrial and Physical Pharmacy, College of Pharmacy, Purdue University, 575 Stadium Mall Drive, West Lafayette, IN, 47907, USA. Tel.: +1
765 496 6614; fax: +1 765 494 6545.
E-mail address: lstaylor@purdue.edu (L.S. Taylor).

http://dx.doi.org/10.1016/j.addr.2016.03.006
0169-409X/© 2016 Elsevier B.V. All rights reserved.
 L.S. Taylor, G.G.Z. Zhang / Advanced Drug Delivery Reviews 101 (2016) 122–142 123

4. Biological absorption of insoluble drugs from the ASD: A mass transport process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
4.1. Physical processes and considerations that impact dissolution of the ASD and absorption . . . . . . . . . . . . . . . . . . . . . . . . 133
4.1.1. Disintegration and erosion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
4.1.2. Driving force for dissolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
4.1.3. Kinetics of dissolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
4.1.4. Phase behavior of the ASD solids during dissolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
4.1.5. Phase behavior in aqueous phase . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
4.1.6. Physical stability of the drug-rich phase that evolves following LLPS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
4.1.7. Flux through GI membrane . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
4.2. The “reservoir” effect of LLPS: An ideal case for absorption . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
4.3. Interplay among the different physical processes during dissolution and absorption . . . . . . . . . . . . . . . . . . . . . . . . . . 136
4.3.1. Case 1: Instantaneous achievement of LLPS, whereby the nanodroplets are physically stable throughout the absorption process . . . 136
4.3.2. Case 2: Instantaneous achievement of LLPS, but system is not physically stable due to crystallization . . . . . . . . . . . . . . 137
4.3.3. Case 3: ASD is dissolving at a slow rate and LLPS is not achieved . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
4.3.4. Case 4: Crystalline suspensions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
4.4. Other considerations related to absorption . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
5. Concluding remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
Disclosure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140

1. Introduction
It has been almost 20 years since Lipinski first proposed the “Rule of
Five,” outlining desirable physicochemical properties, i.e. drug-like
properties, for efficient oral delivery of active pharmaceutical agents
[1]. However, the percentage of insoluble compounds in discovery and
development pipelines across the pharmaceutical industry continues
to increase. While 40% of approved drugs are considered insoluble,
nearly 90% of developmental compounds are poorly soluble [2]. More-
over, the extent of insolubility is also becoming worse. In the 1990's,
an equilibrium solubility of 100 μg/mL might have been considered
highly insoluble. Nowadays, pharmaceutical scientists feel fortunate to
have drug candidates with an aqueous solubility of 10 μg/mL. Regardless
of the reasons underlying the trends in suboptimal aqueous solubility of
new chemical entitities (NCE), formulation scientists are tasked with
designing formulations that deliver these intractable compounds to
patients in an efficient and consistent manner.
Over the years, great advancements have been made in this area.
Many formulation approaches have been explored and utilized to over-
come the challenge of poor solubility. Since the pre-requisite of absorp-
tion is the presence of drug molecules in solution, many formulation
approaches pre-dissolve the drug in the formulation using cosolvents
or surfactant-lipid blends [3]. For most solubility enhancing approaches
based on solid dosage forms, the theoretical basis is the Whitney–Noyes
equation, coupled with the Nernst–Brunner diffusion layer model,
which describes the dissolution of a solid:
dM DA DAC s
¼ ðC s −C Þ≈
dt h h
nanometer range has seen application in the delivery of insoluble com-
pounds [4,5]. Many approaches, such as co-formulating with solubiliz-
ing agents, using salts, co-crystals, and amorphous solids as the API,
and amorphous solid dispersions (ASD), have been employed to
enhance solubility.
In the past decade, there has been increasing realization that simply
increasing the dissolution rate or solubilizing the API is often insufficient
to achieve the desired bioavailability. Consequently, interest in delivery
systems that lead to supersaturation has burgeoned. In this context,
ASDs, where the API is typically molecularly mixed with or, less
commonly, physically suspended in a hydrophilic polymer matrix, are
commonly employed. When ASD formulations are administered,
dissolution in the gastrointestinal fluids often leads to supersaturation,
which drives rapid and sustained absorption. The application of
ASDs has gained much attention in the recent years as a result of funda-
mental work on amorphous pharmaceutical systems [6–15] and experi-
ence gained through development and launch of commercial products
[16–18]. Other formulation approaches or inherent molecular proper-
ties may also lead to supersaturation in vivo. Solubilizing components,
such as surfactants, that are added to the formulation, or present
in vivo, change the thermodynamic properties and phase behavior of
supersaturating systems, resulting in complex systems that are not
well understood. The physical chemistry of these supersaturated
solutions and the implications for oral absorption are the subject of
this review.
2. Theoretical considerations
ð1Þ
2.1. Crystalline solubility

where M is the mass of solute dissolved, dM/dt is the dissolution rate, The solubility of the crystalline solid is a solid–liquid equilibrium
D is the diffusion coefficient of solute, A is the surface area of the ex- (SLE) and is the concentration in solution following attainment of equi-
posed solid, h is the thickness of the diffusion layer, Cs is the solubility librium between the crystalline solid and the solution phase whereby
of solute in dissolution medium, and C is the concentration of solute in the solvent is not present in the crystalline phase and is given by [19,
the dissolution medium. Under sink conditions where C is approximate- 20]:
ly zero, the right hand side of the equation can be simplified.
Two parameters in Eq. 1, D and h, are dictated by the molecule and   μ ðcÞ  μ 0ðaÞ
W ð cÞ W ð cÞ
human gastrointestinal tract (GIT) conditions and are consequently ln xD γD ¼ D D
ð2Þ
RT
difficult to influence by formulation. However, both surface area and
solubility can be readily manipulated by formulation. Particle size where xW(c)
D is the mole fraction solubility of the crystalline drug (D) in
reduction of the active pharmaceutical ingredient (API) using various water, W, γW(c)
D is the liquid phase activity coefficient of the drug at sat-
milling techniques, direct generation of small API particles by rapid uration, R is the gas constant, T is the experimental temperature, and
(c)
precipitation or supercritical fluid technology, has been widely used to μD - μ0(a)
D is the chemical potential difference between the crystal
(c)
facilitate dissolution. More recently, size reduction of the API to the (μD ) and the pure supercooled liquid (μ0(a) D ) at the temperature of
124 L.S. Taylor, G.G.Z. Zhang / Advanced Drug Delivery Reviews 101 (2016) 122–142

interest. The supercooled liquid is considered as the standard state in the amorphous form lacks the long range interactions present in the
this derivation. crystal structure and:

W ðaÞ W ðaÞ
2.2. Supersaturation ln xD ¼  lnγ D ð5Þ

A solution is defined as being supersaturated when the chemical Eqs. 2 and 5 can be combined to express the solubility of the amor-
potential of the solute in the solution phase is higher than that of a sol- phous drug in terms of the crystal solubility and the chemical potential
ute at equilibrium with the most thermodynamically stable crystalline difference:
form. The fundamental supersaturation (S) can be expressed by the
following equation [21]: ðcÞ 0ðaÞ
W ðaÞ W ð cÞ μD  μD
ln xD ¼ ln xD  ð6Þ
RT
W ðSÞ ð cÞ W ðSÞ W ðSÞ
μD  μD γ x
lnS ¼ ¼ ln DW ðcÞ DW ðcÞ ð3Þ
RT γ D xD To arrive at Eq. 6, it is assumed that the liquid phase activity
coefficient of the compound dissolved in water is constant over the
Here μW(S)
(c)
- μD is the chemical potential difference between the range of concentrations encompassing the crystalline solubility and
D
drug in the supersaturated aqueous solution and in a saturated solution amorphous solubility, i.e. γW(c)D ≅ γW(a)
D . This is a reasonable assump-
(note, the chemical potential of the drug in the saturated solution is tion for poorly water soluble compounds where concentrations are
equivalent to the chemical potential of the crystalline drug, μ D ),
(c) very low.
W(S)
xD is the mole fraction concentration in the supersaturated solu- In order to use Eq. 6 to estimate the solubility of the amorphous
tion, and γW(S) is the liquid phase activity coefficient of the drug in drug, it is necessary to not only experimentally determine the crystal-
D
the supersaturated solution. For practical purposes, supersaturation line solubility, but also to estimate the chemical potential (free energy)
is generally expressed in terms of the ratio of the solution molar con- difference between the crystal and amorphous drug at the temperature
centrations: of interest. Various approaches have been used to estimate this param-
eter, all of which require experimental measurement of the enthalpy
C and temperature of fusion (ΔHf and Tm respectively), whereby more rig-
SC ¼ ð4Þ
C orous methods require determination of the heat capacity of the crystal
and supercooled liquid. Assuming that the heat capacity difference
(c)
where C is the molar concentration in the supersaturated solution between the solid and the liquid (ΔCp) is constant, then μD - μ0(a) D can
and C ⁎ is the molar concentration of the saturated solution. The be expressed by [20]:
concentration-based supersaturation (S C ) can be linked to that      
shown in Eq. 3 by the ratio γW(S) /γ W(c) . For solutions where the ðcÞ 0ðaÞ ΔH f Tm  T ΔC p T m  T ΔC p Tm
D D
μD  μD ¼ þ  ln ð7Þ
ratio of the activity coefficients is close to 1, the concentration- R T mT R T R T
based supersaturation is a good approximation of the thermodynam-
ically defined supersaturation. This approximation is likely to be The value obtained from Eq. 7, or variants thereof, can then be used
valid at very low supersaturations and/or for very dilute solutions together with the experimental crystal solubility and Eq. 6 to estimate
in simple media. the amorphous solubility. However, when using Eq. 7 to evaluate the
chemical potential difference, it is assumed that both the crystalline
2.3. Amorphous solubility and amorphous forms of the drug are pure. While this is a reasonable as-
sumption for non-solvated crystals, it is well known that amorphous
The solubility of an amorphous material can be described as the substances absorb water because of their liquid-like structure [30].
concentration achieved in solution following equilibrium between the Thus a small amount of water will mix with the amorphous drug during
solution phase and the amorphous material. Strictly speaking, such an the equilibration process with water, and alter the chemical potential of
equilibrium is impossible since the crystalline material is the thermody- the compound. The chemical potential of the drug in the amorphous
namically stable form. However, if crystallization is slow, then concen- material containing a given amount of water (μDL(a)) is given by:
trations in excess of the crystalline solubility (i.e. a supersaturated
LðaÞ 0ðaÞ LðaÞ
solution) can be achieved by dissolving amorphous material, and μD ¼ μD þ RTlnaD ð8Þ
under some conditions, a metastable equilibrium can exist between
the amorphous material and the aqueous solution. From a thermody- where μ 0(a)
D is the chemical potential of the pure amorphous drug at the
namic perspective, the amorphous solubility can be treated as a temperature of interest, T, and aL(a)
D is the activity of the drug in the
liquid–liquid equilibrium (LLE) [22], which enables the amorphous sol- water saturated amorphous phase (L). The activity of the drug depends
ubility to be predicted, as long as the crystalline solubility is known [23]. on its concentration in that phase (lnxDL(a)) as shown by Eq. 9:
Numerous approaches to estimate the solubility of an amorphous mate-
LðaÞ LðaÞ LðaÞ
rial in water have been presented in the literature [12,23–27]. The sim- aD ¼ xD γ D ð9Þ
plest approach is to assume that the amorphous material is a
supercooled liquid at the temperature of interest. This assumption where γDL(a) is the activity coefficient of the drug in the water saturated
holds if the glass transition temperature (Tg) of the drug of interest is amorphous phase (L).
lower than the temperature of interest. For compounds that are glasses When an amorphous drug is added to aqueous solution, assuming
at the temperature of interest, studies have suggested that extrapolating crystallization does not occur, the drug will dissolve into the bulk liquid
the properties of the supercooled liquid to that temperature enables a phase until it reaches its (amorphous) solubility limit, and water will
reasonable estimation of the amorphous solubility to be made [28,29]. penetrate into the amorphous drug until it is saturated with water, at
Herein, we will assume that the amorphous material in an aqueous en- which point a metastable equilibrium is established between these
vironment can be treated as a supercooled liquid for the purposes of un- two phases. The equilibrium is metastable because crystallization can
derstanding the solution phase behavior. occur from either the water-saturated amorphous drug or the bulk
For an amorphous drug, the mole fraction solubility (xW(a) D ) depends aqueous phase since the system is supersaturated. Importantly, the
on the liquid phase activity coefficient (γW(a)
D ), i.e. it is mainly governed presence of water in the amorphous drug will reduce the chemical po-
by the intermolecular forces between the solute and the solvent since tential. Eq. 6 can be modified using the relationships shown in Eqs. 8
 L.S. Taylor, G.G.Z. Zhang / Advanced Drug Delivery Reviews 101 (2016) 122–142
and 9, to account for the impact of the water on the chemical potential
of the amorphous drug:
ð cÞ
μD  μD
0ðaÞ  
W ðaÞ W ð cÞ LðaÞ LðaÞ
lnxD ¼ ln xD  þ ln xD γ D
RT
where xW(a)
D is the mole fraction of the compound in the bulk aqueous
phase (W), i.e. the solubility, xL(a)
D is the mole fraction of the drug in
the water-saturated amorphous phase (L) and γDL(a) is the correspond-
ing activity coefficient of the compound in this phase. From Eq. 10, it
is apparent that the amorphous solubility depends on the crystalline
solubility, the free energy difference between the crystal and the pure
amorphous material, and the amount of water present in the amor-
phous material following equilibration with water.
Murdande et al. [12] have provided a comprehensive thermody-
namic analysis of the amorphous solubility yielding Eq. 11 which is
essentially equivalent to Eq. 10:
ΔGc→a
C amorphous ¼ C eq  eIða2 Þ  e RT
where (Camorphous) is the amorphous solubility, Ceq is the experimental-
ly determined solubility of the crystal form, ΔGc →a is the free energy
difference between the crystal and amorphous form and -I(a2) is the
activity of the amorphous drug saturated with water, and is equivalent
to the term xL(a)
D γD
L(a)
in Eq. 10. This term can be considered as a correc-
tion factor that accounts for the change in the chemical potential of the
compound in the amorphous phase due to absorption of water. It can be
estimated from experimental data [12,26] by determining the number
of moles of water absorbed per mole of amorphous compound as a func-
tion of relative humidity and estimating the water content at a relative
humidity of 100 (water activity of 1). The activity of the compound in
the presence of the absorbed water is then estimated by applying the
Gibbs–Duhem equation to water sorption isotherm data obtained for
the amorphous compound. The detailed thermodynamic analysis is
presented in Ref. [12].
2.4. Thermodynamics of mixing and liquid–liquid phase separation
In this section, a link will be established between the amorphous
solubility and liquid–liquid phase separation, as inferred in the work
of Brick et al. [25] and described in detail by Paus et al. [22]. Hence a
brief review of the thermodynamics and phase behavior of two partially
miscible liquids is warranted. As is well known, not all liquids mix
completely, with a common pharmaceutical example being the n-
octanol–water system. A liquid mixture of octanol and water will split
into two liquid phases when combined together at most compositions.
In other words, water has a finite solubility in octanol, and if this solubil-
ity is exceeded, a water-rich phase will evolve. Likewise, octanol has a
finite solubility in water, and exceeding this concentration will lead to
the formation of an octanol-rich phase. Similarly, the liquid form of a
pharmaceutically relevant compound can have a finite solubility in
water, leading to a miscibility gap [31–33].
A free energy of mixing versus composition diagram for two partially
miscible liquids is shown in Fig. 1. Starting from the right hand side of
the diagram, with pure liquid water (W), it can be seen that the free en-
ergy of mixing becomes more negative as liquid or amorphous drug (D)
is added indicating that mixing of the two liquids is thermodynamically
favorable, until the composition xWD is reached. Likewise, commencing
with pure liquid or amorphous drug, the free energy of mixing de-
creases as water is added until a composition of xDW is reached. Howev-
er, at compositions between xDW and xWD, even though the free energy
of mixing is still negative, the system can achieve a lower overall free
energy by splitting into two separate liquid phases, one with a compo-
sition of xDW and the other with a composition of xWD, and is said to ex-
hibit a miscibility gap. In other words, a two phase system is expected to
evolve if the composition generated falls between xDW and xWD. The
125
ð10Þ
ð11Þ
Fig. 1. Free energy versus composition diagram for two liquids with a miscibility gap. D is
drug and W is water. Treating the amorphous drug as a liquid, the amorphous solubility is
given by the composition xWD. Figure adapted from Ref. [34].
compositions xDW and xWD are termed the binodal points or coexistence
concentrations, and represent the compositions of each phase that exist
in equilibrium following phase separation. At the binodal compositions,
the chemical potential of each species is the same in both phases, as
shown by the common tangent in Fig. 1. xDW is the drug-rich phase
and contains a small amount of water mixed with the liquid or amor-
phous drug, while xWD is the water-rich phase and contains a small
amount of drug dissolved in water. For compositions between the
binodal and spinodal points (compositions x DW and x WD), the homoge-
neous solution is metastable and formation of a new liquid phase occurs
through nucleation and growth of the new phase. The spinodal points,
x DW and x WD, are the inflections in the free energy versus composition
curve where ∂2G/∂ x2 = 0. For compositions between the spinodal
points, the homogeneous solution is unstable because small fluctuations
in composition that produce drug-rich and water-rich regions, will re-
sult in an overall decrease in free energy. Hence the formation of the
two liquid phases is spontaneous, occurring via spinodal decomposition
which is a diffusion controlled process. The process by which a homoge-
neous liquid splits into two liquid phases is termed liquid–liquid phase
separation (LLPS). If one of the new phases is a glass (i.e. the tempera-
ture is below the glass transition temperature of the water-saturated
drug-rich phase), the process has been termed glass-liquid phase
separation (GLPS) [29].
The concept of liquid–liquid phase separation can be extended be-
yond simple liquids, and has been used to explain the formation of
amorphous nanoparticles using a solvent shifting method, whereby a
poorly water soluble compound dissolved at a high concentration in a
water-miscible organic solvent was added to water [25]. The amor-
phous material was treated thermodynamically as a supercooled liquid
which exhibits a miscibility gap with water. The authors proposed that
the amorphous nanoparticles were formed by spinodal decomposition
due to a rapid change of added droplet composition into an unstable
region due to the counterdiffusion of the organic solvent (out of the
droplet) and water (into the droplet). Other examples of liquid–liquid
phase separation at temperatures below the melting point of the crys-
talline solid, where the crystalline form is the thermodynamically fa-
vored form, have been reported [31,32,35]. The thermodynamics
underlying liquid–liquid equilibria have also been used as an approach
to describe the amorphous solubility [22].
The link between LLPS and amorphous solubility is readily demon-
strated by considering the well-established concept that, at equilibrium,
the activity of a particular species, ai, in a multi-component system is
identical in all phases. Therefore, for two liquid phases in equilibrium,
126 L.S. Taylor, G.G.Z. Zhang / Advanced Drug Delivery Reviews 101 (2016) 122–142

whereby one phase is water-rich (W) while the second phase is drug-
rich (L) the thermodynamic activity of the drug (aD) is the same in
each phase:

W ðaÞ LðaÞ
aD ¼ aD ð12Þ

where aW(a)
D is the thermodynamic activity of the compound in the
drug-saturated bulk aqueous solution (W) and aL(a) D is the thermody-
namic activity of the compound in the water-saturated drug-rich liquid
(L). Herein, the term drug-rich liquid is used to encompass thermody-
namically stable liquids (i.e. compounds which are above their melting
points), as well as thermodynamically metastable supercooled liquids
or glasses (relevant for compounds which are most stable as the crystal-
line form under the experimental conditions). It should be noted that
for supercooled liquids and glasses, the LL equilibrium is metastable as
the crystalline form is the most stable form. Eq. 12 can be written in
the following form: Fig. 2. Schematic temperature-composition phase diagram for a drug (D) in water (W)
illustrating the crystalline and amorphous solubility curves. The numbers in parentheses
W ðaÞ W ðaÞ LðaÞ LðaÞ
lnxD γD ¼ ln xD γD ð13Þ indicate different regions of the phase diagram that are discussed further in the text. A
similar representation is presented in Ref. [22].

The amorphous solubility is equivalent to xW(a)

D , which is the mole
fraction of the solute in the water-rich phase, which in turn is deter- exceeded, the system can achieve a lower free energy by splitting into
mined by the activity of the solute in drug-rich phase and the activity two liquid phases, which can persist if the crystallization rate is slow.
coefficient of the drug in the aqueous phase. This is analogous to the The amorphous phase generated can be considered a precursor to crys-
crystalline solubility (Eq. 2) which depends on the activity of the tallization since the system is still supersaturated after LLPS has oc-
crystalline solid and the solution activity coefficient. curred and is expected to subsequently crystallize. Fig. 3 illustrates the
By referring to Fig. 1, the amorphous solubility corresponds to the initial liquid–liquid demixing followed by crystallization for a “slow
composition xWD, and it is apparent that, in the absence of crystalliza- crystallizer.” Such amorphous precursor phases have been widely stud-
tion, this composition can be approached from two directions. First, by ied for inorganic systems dissolved in water since they are thought to be
equilibrating an amorphous material with the aqueous phase. In this in- important in the area of biomineralization [37], but analogous studies
stance, we commence with pure amorphous drug which simultaneous- for organic systems are quite rare. Several research groups involved in
ly dissolves and absorbs water until the system is in metastable crystallization process development have observed LLPS of solutes in
equilibrium whereby no more material will dissolve. Second, we can organic/aqueous solvent mixtures (often termed “oiling out” in this
start with pure water, and add drug. In the absence of crystallization, context) followed by subsequent crystallization which occurs minutes
the solution will be homogeneous until the composition xWD (the amor- [36] to hours [34] later. LLPS is most commonly observed to occur in
phous solubility) is reached. For solutions containing higher amounts of systems where there is very high supersaturation, rapid generation
drug, liquid–liquid phase separation is thermodynamically favored, of supersaturation, when crystallization inhibitors or high levels of im-
yielding a drug-rich amorphous phase either through nucleation and purities are present and where there is inadequate mixing causing
growth or spinodal decomposition. As will be discussed subsequently, high local supersaturations [38]. LLPS also has been observed in aqueous
the second route, where liquid–liquid phase separation occurs by ex- solutions during the evaluation of solubility enhancing formulations
ceeding the amorphous solubility, is of relevance for drug delivery. Re- [39–42]. The drug-rich species formed following LLPS typically have a
gardless of the route, the final system is equivalent, and consists of size of 100–500 nm when characterized immediately after formation
two phases, a water-saturated drug-rich amorphous phase of composi- using techniques such as dynamic light scattering [29,39,40,43]. Growth
tion xDW, and a drug-saturated water-rich solution of composition xWD. of the drug-rich droplets/particles can subsequently occur, or can be
inhibited if additives such as polymers are present [40,44]. Thus some
2.5. Liquid–liquid phase separation versus crystallization compounds can clearly persist as a two phase solution which remains
supersaturated at a constant level (at the coexistence concentration)
A phase diagram for the SLE and LLE for a drug in water is shown
schematically in Fig. 2 [22,36].
The crystalline (xW(c)
D ) and amorphous (xW(a)D ) aqueous solubilities,
and the composition of the water-saturated drug-rich phase (xL(a) D ) at
a fixed temperature are noted on the x-axis. Region 1 shows where
the solution is undersaturated with respect to the crystalline solubility
and the solution is stable. The line dividing regions 1 and 2 is the solu-
bility line for the crystalline drug. In region 2, the solution is supersatu-
rated with respect to the crystalline form and therefore metastable.
Supersaturation provides the driving force for phase separation; in re-
gion 2, crystallization can occur if the barrier for crystal nucleation can
be overcome. Region 3 shows the concentration and temperature region
where liquid–liquid phase separation can occur; in region 3 either LLPS
or crystallization can take place. Because the formation of crystalline
material by nucleation and growth involves a variety of kinetic barriers,
LLPS may occur more rapidly than crystallization, in particular for com-
plex organic molecules for which organization into a crystal lattice is en-
tropically unfavorable. Thus for “slow crystallizers,” when high Fig. 3. Schematic diagram illustrating the phase behavior of slow and fast crystallizing
supersaturation is rapidly generated and the amorphous solubility is compounds in the context of the crystalline and amorphous solubility phase boundaries.
 L.S. Taylor, G.G.Z. Zhang / Advanced Drug Delivery Reviews 101 (2016) 122–142 127

until crystallization occurs. The relevance of this LLPS and the implica-
tions of an upper limit of supersaturation, as indicated by Fig. 3, to solu-
bility enhancing systems will be discussed subsequently.
In contrast, some compounds will crystallize rapidly at supersatura-
tions lower than the amorphous solubility. These compounds can be de-
scribed as “fast crystallizers,” and their behavior is illustrated in Fig. 3. In
a study of approximately 50 pharmaceutically relevant compounds,
around a third were found to be fast crystallizers, whereby they were
observed to crystallize within 2.5 min of the generation of a highly su-
persaturated solutions in aqueous media using the solvent shift tech-
nique [45]. In contrast, a third of the compounds evaluated failed to
crystallize after 1 h, even though a visible precipitate formed indicating
that liquid–liquid demixing occurred, while the remaining third of the
compounds crystallized at intermediate time points. The crystallization
kinetics of a compound can be altered through the addition of excipients
such as polymers or other substances that act as inhibitors [46–48]. This
in turn can facilitate LLPS by inhibiting crystallization, enabling the
generation and maintenance of higher supersaturations where LLPS
can be observed [49].

3. Experimental observations Fig. 4. Predicted versus measured amorphous solubility for a set of 23 compounds. A
line with a slope of 1 was added for reference. Data were compiled from Refs.
3.1. Experimental determination of amorphous solubility [28,29,35,39,43,53].

It is widely recognized that experimental determination of the scattering of light which can be readily detected. Indeed, solutions un-
amorphous solubility is challenging [27,50]. Common issues include dergoing LLPS often appear milky and/or with a bluish tinge when
not being able to prepare a pure amorphous reference, and perhaps viewed with the naked eye. In their classic study, Thomson et al. studied
most frequently, crystallization prior to attainment of the liquid–liquid LLPS in protein-water solutions using light scattering to determine the
equilibrium. A pure amorphous reference is prepared most convenient- binodal and spinodal concentrations as a function of temperature, en-
ly by melt quenching, spray drying or spin coating. If the preparation abling construction of a phase diagram [54]. Their approach was subse-
method yields an amorphous material, this solid is then added to an quently adopted to evaluate LLPS of a pharmaceutical compound where
aqueous medium and the concentration evolution with time is moni- liquid–liquid demixing was observed during evaluation of the crystalli-
tored until a plateau is achieved, using a similar approach to that used zation process [55], and to characterize the phase behavior of ritonavir
to determine the solubility of the crystalline form. For some compounds, in supersaturated aqueous solutions [35]. The light scattering that arises
good agreement has been observed between the experimentally ob- from the presence of a second phase of colloidal dimensions can be
served amorphous solubility and the predicted value [43]. However, readily detected using a ultraviolet (UV)/visible [56] or fluorescence
many systems crystallize prior to attainment of the amorphous solubil- spectrometer [57]. An example of data for nisoldipine obtained using a
ity, leading to large discrepancies between predicted and experimental- UV spectrometer is shown in Fig. 5. For this system, the concentration
ly observed values [26,27,50,51]. This has led to the development of of the drug in the aqueous solution was increased constantly by addition
alternative approaches. Lindfors and colleagues created rapidly- of a concentrated solution of the compound dissolved in an organic sol-
dissolving amorphous nanoparticles of the poorly water-soluble drug, vent using a syringe pump. The UV absorption spectrum was
felodipine, and elucidated the amorphous solubility by determining
the concentration at which the added nanoparticles stopped dissolving,
as inferred from light scattering data[52]. Another approach that has
been used recently is to generate the amorphous phase in situ by creat-
ing a solution concentration above the amorphous solubility [28,35]. In
this instance, it is possible to either measure the concentration where
the new phase is first detected using a method such as light scattering,
or to separate the two phases by centrifugation followed by analytical
determination of the drug concentration in the water-rich phase.
While this approach is also not infallible, good agreement between the
predicted and experimental amorphous solubility has been observed
using these approaches as shown in Fig. 4. For these compounds, the
amorphous solubility was measured by creating a highly supersaturated
solution and allowing an amorphous phase to evolve through LLPS. In
order for this approach to be useful, crystallization during the genera-
tion of the supersaturated solution needs to be avoided, and sensitive
analytical methods that enable the formation of the new phase to be
rapidly detected are necessary.

3.2. Methods to detect and characterize LLPS

3.2.1. Light scattering techniques

Fig. 5. Application of UV spectroscopy to evaluate liquid–liquid demixing in a
Light scattering is one of the most common methods by which the supersaturated solution of nisoldipine at 25 °C [53]. The extinction at 450 nm was
phase behavior of systems undergoing LLPS is characterized; the forma- measured as a function of added drug concentration. The concentration at which a new
tion of a new phase following liquid–liquid demixing results in the phase is formed is estimated from the intersection of the two red lines.
128 L.S. Taylor, G.G.Z. Zhang / Advanced Drug Delivery Reviews 101 (2016) 122–142
determined, and a wavelength where the drug shows no absorption
was selected as the detection wavelength. The UV signal that is mea-
sured by the spectrometer has contributions from both absorbance
and scattering. By detecting at a non-absorbing wavelength, a sudden
increase in light scattering due to the formation of a new phase is readily
detected, as shown in Fig. 5. This approach is simple, rapid and uses
readily available equipment. However, it does have some obvious limi-
tations which include overshooting the phase separation concentration
due to the dynamic nature of the drug addition, no differentiation be-
tween phase separation to a crystalline versus an amorphous phase,
and reliance on the new phase scattering enough light to be detected
at a concentration close to the phase separation concentration. An alter-
native method for ionizable drugs is to use pH to solubilize the drug and
then create a supersaturated solution by changing the pH to where the
unionized form is favored. The pH where light scattering is first ob-
served can then be used together with the added concentration to calcu-
late the concentration of unionized drug at the point of phase separation
[43,58]; this approach avoids the use of organic solvent. Other light scat-
tering methods that have been used to evaluate LLPS include focused
beam reflectance measurements [34], dynamic light scattering (DLS)
[35,43], static light scattering [59] and nanoparticle tracking analysis
(NTA) [39,43]. DLS and NTA have been used to provide information
about the size and the size evolution of the drug-rich phase formed
via LLPS. Several studies have indicated that the new amorphous
phase formed in highly supersaturated solutions is initially very small
in size, typically in the range of 100–500 nm [29,39,40,43]. The small
size results in the rapid redissolution of the amorphous phase when
the concentration drops below the amorphous solubility [29,40,58].
3.2.2. Fluorescence spectroscopy
Environment-sensitive fluorescence probes have been widely
used for the determination of the critical micelle concentration (CMC)
[60–62], to study polymer-surfactant interactions [63], and for the eval-
uation of micro-emulsions [64] and lipid membranes [65]. These appli-
cations originate from the sensitivity of certain fluorophores such as
pyrene to the polarity of their local environment. For example, when a
surfactant micelle is formed in aqueous solution, the emission spectrum
of pyrene changes to reflect a less polar environment due to probe
partitioning into the micelle. Environment-sensitive fluorophores have
also been used to characterize the drug-rich phase that evolves when
the amorphous solubility is exceeded and LLPS occurs [28,35,53]. The
fluorophore is added to the aqueous phase and the emission spectrum
is monitored as the drug concentration is increased. Because most
polarity-sensitive fluorophores are somewhat hydrophobic, when a
drug-rich phase is formed, they preferentially interact with the new
amorphous phase, and their emission spectrum changes. By monitoring
the emission spectrum as a function of added drug, the concentration
where LLPS occurs can be evaluated. In addition, the subsequent crystal-
lization of the drug can be detected; when the drug crystallizes, the
probe is expelled from the drug-rich phase into the aqueous phase
and the emission spectrum registers a more polar environment. An ex-
ample of the use of pyrene to detect LLPS and the subsequent crystalli-
zation of the solution is shown in Fig. 6 for a supersaturated aqueous
solution of nifedipine. For pyrene, the ratio of the intensity of the first
and third emission peaks is used to provide an indication of the probe
environment; a higher value indicates a more polar environment [61].
Initially, the solution is homogeneous and hence the probe registers a
polar environment as it is dissolved in the aqueous phase. When a cer-
tain drug concentration is exceeded (higher than the amorphous solu-
bility), LLPS occurs and a drug-rich phase is formed. The probe, being
hydrophobic, mixes with the drug-rich phase, which is liquid-like and
can therefore mix with other compatible solutes, and registers a more
non-polar environment. Addition of more drug to the solution results
in crystallization. The probe is not incorporated into the crystal lattice,
and hence is expelled from the drug-rich phase into the aqueous phase
and a more polar environment is registered. Environment-sensitive
Fig. 6. Peak ratio (I1/I3) of pyrene emission spectrum as a function of nifedipine
concentration. The left arrow indicates the formation of a drug-rich nifedipine phase
based on a decrease in the polarity of the probe environment, marking the onset of LLPS.
The right arrow at a concentration of around 200 μg/mL shows the crystallization of
nifedipine and an increase in the polarity of the probe environment as it is expelled into
the aqueous phase with the loss of the drug-rich phase due to crystallization [53].
fluorophores thus provide an additional means of characterizing super-
saturated solutions, although it must be ensured that any autofluores-
cence from the drug does not interfere with the probe spectrum, and it
must be ensured that the fluorophore does not alter the phase
behavior of the system.
3.2.3. Other characterization approaches
Ultracentrifugation provides an approach to separate the two phases
which can then be analyzed individually. This is particularly useful to
confirm the amorphous solubility, since techniques which measure
the concentration where the new phase is first observed, following con-
tinuous or sequential addition of drug, may overshoot the amorphous
solubility, since the system is not in equilibrium and may supersaturate
with respect to the amorphous solubility. In contrast, initiating liquid–
liquid demixing by exceeding the amorphous solubility, and letting
the system phase separate to reach the liquid–liquid equilibrium,
followed by ultracentrifugation and analysis of each phase should en-
able the co-existence concentrations (binodal points, Fig. 1) to be deter-
mined [35]. In practice, it has been observed that good agreement is
usually seen between the amorphous solubility values determined
with different methods, including ultracentrifugation, suggesting that
most systems do not undergo much supersaturation with respect to
the amorphous solubility before phase separation occurs [35,66].
However, kinetic effects should always be explored when evaluating
the amorphous solubility using the methods described above. For ultra-
centrifugation to be an effective approach, the system should not
undergo crystallization during centrifugation, and efficient separation
between the two phases needs to be achieved. Crystallization could be
problematic for many compounds since efficient separation may require
centrifugation for periods of around 30–40 min. In such instances, adding
a small amount of a polymer to inhibit crystallization may be useful.
Nuclear magnetic resonance (NMR) spectroscopy has been used to
evaluate phase separated solutions, and can potentially provide infor-
mation about the concentration of the drug in the water-rich phase, as
well as insight into the characteristics of drug-rich phase. [29,67] NMR
spectroscopy has also been used to evaluate the composition of each
phase when LLPS occurred in the presence of polymers, correlating the
effectiveness of various polymers as crystallization inhibitors based on
their distribution between the aqueous and drug-rich phases [49].
 L.S. Taylor, G.G.Z. Zhang / Advanced Drug Delivery Reviews 101 (2016) 122–142
The drug-rich phase can be characterized with various techniques.
Cryo-transmission and scanning electron microscopy [40,41] have
been used to image the colloidal species. Differential scanning calorim-
etry can provide information about the Tg of the water-saturated amor-
phous material, and hence to determine if the drug-rich phase is a
supercooled liquid or a glass at the temperature of interest [29,35,43].
The crystallization of two phase supersaturated solutions and the
influence of polymers on the crystallization kinetics has been studied
using synchrotron radiation [49].
3.3. The amorphous solubility advantage
Eq. 10 indicates that the amorphous solubility is dictated by the crys-
tal solubility, the free energy difference between the crystal and the
amorphous material, and the amount of water associated with the
amorphous material. The latter two factors impact the amorphous-to-
crystalline solubility ratio, or the “amorphous solubility advantage.”
The amorphous-to-crystalline solubility ratio is as equally important
from a drug delivery perspective as the actual magnitude of the amor-
phous solubility, since it provides an indication of how much the bio-
availability might be improved by formulating the amorphous form of
the drug. From the theoretical section above, it is apparent that com-
pounds with very high melting points and enthalpies of fusion will
give rise to a larger chemical potential difference between the pure crys-
talline and amorphous forms at biologically relevant temperatures and
hence greater amorphous solubility advantages. However, the water
content of the amorphous form may be non-negligible, so both of
these factors need to be considered. The amorphous solubility advan-
tage for 23 compounds (the same compounds shown in Fig. 4) is
compared in Fig. 7 which shows the predicted versus experimental
amorphous-to-crystalline solubility (a/c) ratios.
The range of amorphous-to-crystalline solubility ratios observed ex-
perimentally is quite revealing. At the low end, ibuprofen has a ratio
only 4 [28], which can be accounted for by the low melting point
of the crystalline form (~ 77 °C). Therefore, for compounds with
solubility-limited bioavailability, it would be anticipated that formulat-
ing a compound, such as ibuprofen, as the amorphous form might not
lead to substantial improvements in bioavailability. At the high end,
amorphous atazanavir has a solubility that is a factor of ~ 65 greater
than the crystalline form, which has a melting point of ~ 208 °C. The
Fig. 7. Predicted versus experimentally determined amorphous to crystalline (a/c)
solubility ratio for the 23 compounds shown in Fig. 4. A line with a slope of 1 was added
for reference. The increased scatter relative to that of Fig. 4 most likely arises from the
difficulty in accurately measuring the crystalline solubility of some of these compounds
which have solubility values of b100 ng/mL; small errors in the determination of the
crystalline solubility lead to large errors in the value of the predicted amorphous solubility.
129
Fig. 8. Variation in the amorphous solubility of ritonavir as a function of buffer ionic
strength. The 0 ionic strength point is pure water. Data taken from Ref. [35].
amorphous-to-crystalline solubility ratio for two compounds that
have been formulated as amorphous solid dispersions for the purpose
of bioavailability enhancement, ritonavir and telaprevir, have been re-
ported to be around 15 [35] and 20 [29] respectively. Thus this level of
amorphous solubility advantage clearly translates into a tangible bio-
availability advantage based on the commercial amorphous formulation
strategy pursued for these two compounds.
3.4. Additional factors impacting the experimentally observed amorphous
solubility
3.4.1. Ionic strength
The ionic strength of the medium will impact both the crystalline
and amorphous solubility with higher ionic strength solutions being ex-
pected to reduce solubility. It may be difficult to detect the effect of ionic
strength on the crystalline solubility for very poorly water soluble com-
pounds due to small changes to a small number, but, as shown in Fig. 8,
the ionic strength of the buffer used in amorphous solubility measure-
ments can impact the observed values quite considerably and should
therefore be taken into account when comparing values [35].
3.4.2. pH
It is well established that ionizable compounds show pH dependent
solubility. The pH-solubility profile of the ionizable crystalline drug is
typically described well by the Henderson–Hasselbalch equation,
whereby for a weak base, the solubility increases with a decrease of
pH until the salt form crystallizes. The solubility increase is due to an in-
crease in the fraction of ionized compound with decreasing pH, where-
by the unionized concentration of the drug remains constant. pH
impacts the amorphous solubility in an analogous manner whereby,
for a weak base, the overall solution concentration increases with a de-
crease in pH, but the unionized drug concentration remains constant at
the amorphous solubility [43]. The change in the amorphous solubility
as a function of pH was also found to follow the Henderson–Hasselbach
relationship, with the amorphous solubility of the free base used as the
input value. Thus the ratio of the amorphous and crystalline solubilities
remains constant as a function of pH, although the absolute values of
both change due to ionization [43].
3.4.3. Additional solutes
The presence of other poorly water soluble compounds added to the
medium may reduce the amorphous solubility if they mix with the
drug-rich phase. The origin of this effect stems from the reduced activity
130 L.S. Taylor, G.G.Z. Zhang / Advanced Drug Delivery Reviews 101 (2016) 122–142
of the drug in the drug-rich phase when mixed with another compo-
nent, which in turn reduces the amorphous solubility as shown by
Eqs. 11 and 12 and discussed above for water. It has been demonstrated
that the amorphous solubility of ritonavir is decreased in the presence
of lopinavir and vice versa, because these two compounds are miscible
in the amorphous phase [68]. The relationship between the amorphous
solubility and the mole fraction of the drug in the drug-rich phase is il-
lustrated in Fig. 9. For the ritonavir-lopinavir system, the decrease in the
amorphous solubility can be estimated by assuming ideal mixing be-
tween the two compounds in the drug-rich phase. At a mole fraction
of 0.5, the amorphous solubility of each compound is reduced by a factor
of 2. In contrast, it was noted that diclofenac anion has no impact on the
amorphous solubility of ritonavir, presumably because this ionized
species does not mix with amorphous ritonavir.
The effect of polymers on the amorphous solubility has also been
studied. For danazol, polyvinyl pyrrolidone (PVP), hydroxypropyl meth-
yl cellulose (HPMC) and hydroxypropyl methyl cellulose acetate succi-
nate (HPMCAS) were found to have no detectable impact on the
amorphous solubility, suggesting that these polymers do not substan-
tially mix with the danazol drug-rich phase in aqueous solution [69].
In a study by Raina et al., NMR spectroscopy was used to determine
the amount of polymer associated with the drug-rich phase [49].
The concentration of polyvinyl pyrrolidone-vinyl acetate copolymer
(copovidone or PVPVA) and HPMCAS in the drug-rich phase was both
found to be less than 7 wt.% in both instances, confirming that the poly-
mers do not mix substantially with the drug-rich phase, and thus do not
have a discernible impact on the amorphous solubility.
3.5. Membrane transport and supersaturated solutions
Low solubility compounds can result in reduced oral bioavailability
not only by reducing dissolution rates, but also because of the low abso-
lute amount of drug that is present in the intestinal fluids, which will re-
sult in a correspondingly low transport rate across the enterocytes.
Therefore, we can differentiate dissolution-limited bioavailability from
solubility-limited bioavailability. The former situation can be addressed
through formulation strategies that improve the dissolution rate, while
for solubility-limited bioavailability, it is necessary to increase the num-
ber of molecules in the solution beyond that achieved at saturation, i.e.
to generate a supersaturated solution. In the area of transdermal drug
Fig. 9. Relationship between the amorphous solubility (concentration in aqueous phase)
and the mole fraction of ritonavir, Xritonavir, in the drug-rich phase. The points represent
experimental data while the lines are calculated values for a simple model that assumes
ideal mixing between ritonavir and lopinavir in the drug-rich phase. The figure was
adapted from Ref. [68].
Fig. 10. Membrane transport rate (flux) as a function of danazol (DNZ concentration). The
supersaturated danazol solutions were created using antisolvent addition (black data
points), or by adding an amorphous solid dispersion (ASD) formulated with PVP, HPMC
or HPMCAS, containing 90 wt.% polymer. The flux for a crystalline suspension is shown
for reference. Data taken from Ref. [69].
delivery, it has long been recognized that there is a linear relationship
between the degree of saturation of the donor solution and the rate of
transport across the membrane (flux) for systems where the membrane
is the major transport barrier [70–74]. Thus, supersaturated solutions
can improve the rate of delivery of drug across a membrane; this has
been shown in the work of Miller et al. where the increased flux of su-
persaturated progesterone solutions was demonstrated using a rat jeju-
nal perfusion assay [75]. In the context of oral delivery, supersaturation
in the intestinal fluid is therefore considered an important mechanism
to improve bioavailability of poorly soluble compounds [76].
It is important to consider that concentration enhancements per se
may not be predictive of changes in flux values. Thus Twist and Zatz
demonstrated a nearly constant flux for solutions of vastly different con-
centrations (varying by approximately two orders of magnitude) that
contained a solute at a constant thermodynamic activity, achieved by
ensuring that the solutions were saturated with respect to the crystal-
line compound [70]. As explained in detail by Yalkowsky [71,77],
there is a reciprocal relationship between apparent solubility and the
apparent membrane-solution partition coefficient, and since both of
these factors impact flux, drug concentration alone cannot be used to
predict membrane transport rate. Thus if the solubility is increased by
using, for example, a surfactant, co-solvent or complexing agent, the
flux will not increase since the partition coefficient between the mem-
brane and the solution is decreased in a manner that is directly propor-
tional to the increase in solubility. Consequently, it has been suggested
by Higuchi that it is more useful to consider membrane flux in terms
of the solute thermodynamic activity [78]. Thus a higher solute activity
translates to an increased driving force for transit from the intestinal
fluids into and out of the enterocytes. From a physical chemistry view-
point, this is a convenient approach when considering the effect of
supersaturation because supersaturation can be defined in terms of
thermodynamic activity, and hence it is possible to predict a relation-
ship between the degree of supersaturation (S) and the membrane
flux [39,79]:
D
J∝S ð14Þ
hγ m
where J is the membrane flux (mass transferred per unit area of
membrane per unit time), D is the solute diffusion coefficient, h is the
membrane thickness, and γm is the solute activity coefficient in the
 L.S. Taylor, G.G.Z. Zhang / Advanced Drug Delivery Reviews 101 (2016) 122–142
membrane. D, h and γm are constants, hence for a known experimental
system, the observed flux is directly proportional to the supersaturation.
Flux data for supersaturated danazol solutions is shown in Fig. 10,
with a saturated danazol solution providing a reference point; the solu-
bility of crystalline danazol is about 1 μg/mL in buffer [39]. As anticipat-
ed from Eq. 14, the flux increases linearly with increasing danazol
concentration until the concentration reaches approximately 8 μg/mL.
For higher concentrations of added danazol, the flux remains constant.
The plateau cannot be explained by crystallization of danazol, because
the flux is about an order of magnitude higher than that of the crystal-
line drug. Instead, as described in detail by Raina et al. [80], the plateau
can be explained based on the amorphous solubility of danazol (approx-
imately 8 μg/mL, [39]), and the liquid–liquid phase separation of the
system at concentrations higher than the amorphous solubility. When
the solution concentration exceeds the amorphous solubility, a drug-
rich phase is formed, and, in the two phase system, the thermodynamic
activity of the drug in each phase is the same, remaining constant as
long as both phases are present and crystallization is avoided.
Addition of more drug simply increases the amount of the drug-rich
phase, but not the thermodynamic activity of the solute in the system.
Hence the supersaturation is also constant. Since flux depends on solute
thermodynamic activity, and correspondingly the supersaturation, this
parameter also becomes constant, and consequently a plateau is ob-
served. The formation of a plateau in the flux versus concentration pro-
file when the amorphous solubility has been exceeded has been noted
for several compounds [29,68,69,80]. From Fig. 10 it can be seen that
the method of creating the supersaturated solution has little impact
on the value of the flux obtained whereby solutions derived from the
dissolution of amorphous solid dispersions give rise to a similar flux as
for the solution created via antisolvent addition. Flux measurements
clearly provide a way to determine the activity-based supersaturation
of a complex solutions as will be discussed in more detail below.
3.6. The amorphous solubility as the upper limit of supersaturation
Flux studies on supersaturated solutions of poorly water soluble
compounds of the type shown in Fig. 10 demonstrate that there is an
upper limit of supersaturation that can be achieved and that this limit
depends on the amorphous solubility of a compound. This is an impor-
tant conjecture, since the obvious implication is that there is a limit in
the maximum increase in membrane transport, and hence the extent
of bioavailability enhancement that can be achieved based on the use
of a supersaturating dosage form, (of course there are many other
mechanisms that need to be taken into account which could impact bio-
availability, which are discussed in more detail below). Some additional
consequences of this finding also need to be discussed. Most important
of these is the concept that we are not just referring to the upper limit of
supersaturation that is generated by an amorphous formulation. Instead,
we are making the much broader statement that the upper limit of su-
persaturation achievable for any formulation approach is dictated by
the amorphous solubility. Validation of this statement, of course, requires
proper characterization of the activity-based supersaturation. At this
juncture, it is of interest to consider what types of formulation or exper-
imental procedures could result in a solution where the amorphous sol-
ubility is exceeded leading to LLPS and formation of a drug-rich phase. A
number of possible routes to LLPS are summarized in Fig. 11. Out of these
potential routes, temperature change [35], antisolvent addition [35,53],
pH swing [58], amorphous solid dispersion dissolution [39,40] and salt
dissolution [66] have all been demonstrated to result in the formation
of an amorphous, drug-rich phase. There is some evidence in the litera-
ture that lipolysis of a self-microemulsifying drug delivery system may
lead to the formation of an amorphous precipitate [81]. Both prodrug
conversion [82] and dissolution of cocrystals [83] can generate supersat-
urated solutions, hence these are unexplored potential routes that could
result in LLPS.
131
Fig. 11. Potential routes to achieve liquid–liquid or glass-liquid phase separation of poorly
water soluble drugs in aqueous media.
The observation that different routes of generating supersaturation
for a given compound lead to LLPS at the same concentration [35,66]
strongly supports the contention that the amorphous solubility
(or strictly speaking, the spinodal decomposition point, see Fig. 1) of a
compound represents the upper limit of achievable supersaturation.
Therefore, knowledge of the amorphous-to-crystalline solubility ratio
provides an important indication of potential improvements in the rate
of passive transport across the intestinal membrane. This parameter
probably should be considered during lead optimization and candidate
selection, and monitored for structural series that exhibit poor solubility.
3.7. Solubilization
As described above, the concentration of a solution can be increased
beyond the saturation concentration, by creating a supersaturated solu-
tion. The solution concentration can also be increased by adding compo-
nents that solubilize the drug. The fundamental difference between a
supersaturated solution and a solubilized solution is the thermodynamic
activity of the drug in the solution; in a supersaturated solution, by defi-
nition, it exceeds the thermodynamic activity of the crystalline drug,
while in a solubilized solution, the thermodynamic activity of the drug
remains at or below that of the crystalline form. Solubilization can be
achieved in many different ways, but some of the most pharmaceutically
relevant solubilized solutions contain surfactant micelles, cosolvents or
complexing agents such as cyclodextrins. Given the widespread presence
of surfactants in both simulated and real intestinal fluids as well as in for-
mulations, solubilization by surfactants will be considered briefly.
Micellar solubilization has been described in detail previously [19],
and only those aspects pertinent to this review will be briefly discussed
herein. Surfactant micelles increase the solubility of poorly water
soluble compounds by providing a microphase for the drug to interact
with. Surfactant micelles are aggregates of amphiphilic surfactant mol-
ecules that form in solution when the critical micelle concentration is
exceeded. The polar segment of the molecule interacts with the aqueous
phase, while the non-polar segment is self-associated to form the micel-
lar core. Solutes are solubilized by micelles via incorporation into the
micelle, whereby the location of solubilization depends on the proper-
ties of the solute as well as the micelle structure. The solubilization
capacity of a given surfactant for the solute of interest is typically deter-
mined by measuring the increase in solubility of the crystalline com-
pound as a function of micellar surfactant concentration. The micellar
partition coefficient provides a measure of the affinity of the drug for
the micelle relative to its affinity for the aqueous phase and is given by:
Cm
Km ¼ ð15Þ
Cw
132 L.S. Taylor, G.G.Z. Zhang / Advanced Drug Delivery Reviews 101 (2016) 122–142
where Km is the micellar partition coefficient for a given concentration
of micelle, Cm is the solute concentration in the micelle and Cw is the sol-
ute concentration in the aqueous phase. Commonly, Km is determined at
the saturation solubility. The total solubility (Stot) in the presence of a
given concentration of micellar surfactant can then described by [19]:
Stot ¼ Sw þ K m Sw ð16Þ
where Sw is the solubility of the crystalline drug in water. From Eq. 16,
two populations of molecules are implied; free drug molecules in the
bulk aqueous solution (determined by the crystal solubility) and
surfactant-associated molecules (determined by the affinity of the
drug for the micelle). This model of two populations of molecules, free
and surfactant-associated, has been adopted widely to better under-
stand the properties of solutions containing surfactants in particular in
the context of permeability assessment [84,85]. In this context, it is rec-
ognized that only free drug molecules provide the driving force for ab-
sorption, and that solubilization by micelles will lead to a decrease in
the absorption kinetics [84]. Consequently, it has been pointed out
that erroneous estimates of drug permeability will result if micelles
are present, and calculations are performed using total instead of free
drug concentration [85].
3.8. Effect of solubilizing additives on amorphous solubility, supersaturation
and flux
Solubilizing additives, in particular surfactants are frequently
present in the formulation, dissolution medium and are present in the
intestinal fluids in the form of bile salts and phospholipids. Therefore,
when assessing the performance of supersaturating dosage forms, it is
necessary to gain some insight into the impact of solubilizing additives
on the amorphous solubility, supersaturation, and membrane transport
rate. Solubilizing additives increase the crystalline solubility as
discussed above, and this needs to be taken into account when assessing
supersaturation. Since solubilizing additives increase the crystalline sol-
ubility, they would also be expected to increase the amorphous solubil-
ity. This has been found to be the case. For example, the solubility of
crystalline felodipine was found to increase from around 1 μg/mL in
buffer to 40 μg/mL in the presence of Vitamin E TPGS micelles while
the amorphous solubility changed from around 10 to 177 μg/mL [79].
The solubility is increased in both instances by association of the drug
with the micelles. However it is apparent that, in the example provided,
the crystalline solubility increases to a greater extent than the amor-
phous solubility, for a given concentration of micellar surfactant. The
proposed origin and consequences of this observation will be discussed
subsequently.
Adding a solubilizing excipient to a supersaturated solution will alter
the degree of supersaturation of the solution. Using nifedipine and
felodipine as model compounds, it was noted that at a given solute
concentration above the equilibrium solubility, solubilizing additives,
including surfactants, proteins and cyclodextrins, reduced the solute
membrane flux [79]. The decrease in flux indicated that the extent of su-
persaturation was reduced by the solubilizing additives. The flux could
be increased back to the same maximum value (which was determined
by the concentration where liquid–liquid phase separation (LLPS)
occurred) by increasing the total solute concentration to above the
amorphous solubility in the presence of the additive.
Commonly, it is assumed that Eq. 4 holds, and hence that supersatu-
ration in the presence of a solubilizing additive can be assessed by mea-
suring the crystalline solubility in the medium of interest, evaluating
the solution concentration generated by the formulation, and then
determining the supersaturation from the ratio of the two quantities.
However, it has been shown recently that Eq. 4 under predicts the su-
persaturation in solutions of felodipine and nifedipine that contain sol-
ubilizing components including surfactant micelles, cyclodextrins and
proteins [79]. Using flux measurements to determine the activity-
based supersaturation, it was noted that the supersaturation increased
more steeply in the presence of vitamin E TPGS surfactant micelles
and in fasted state intestinal fluid (FaSSIF) than would be expected
based on Eq. 4. However, the maximum flux value that could be obtain-
ed was the same in the presence and absence of the surfactants, indicat-
ing that the amorphous solubility still dictates the maximum possible
supersaturation. Due to the solubilization effect of the surfactant
micelles, the maximum in the flux was not observed until much higher
solute concentrations, consistent with an increased amorphous solubil-
ity, as described above. Because of the breakdown of Eq. 4, the
amorphous-to-crystalline solubility changed from ~9 in buffer to ~4.5
in the presence of vitamin E TPGS. If only the concentration ratios in
the different media were considered, it would appear that the amor-
phous solubility advantage is lowered in the presence of a surfactant;
this is not supported by the flux measurements which show that the
same maximum flux can be achieved as long as the solubilization effect
is counteracted by increasing the solution concentration to the amor-
phous solubility. The breakdown of Eq. 4 for evaluating supersaturation
in complex solutions can be readily explained by consideration of the
surfactant micelle-water partition coefficient (Eq. 15). This is typically
measured at the crystalline solubility and is assumed to remain con-
stant; if this parameter was indeed a constant, then Eq. 4 would hold.
However, there is ample evidence in the literature that Km is highly de-
pendent on the solute concentration, normally decreasing with increas-
ing solution concentration [86–88]. A decrease in Km with increasing
solute concentration as the solution becomes more supersaturated
provides the most plausible explanation for the breakdown of Eq. 4 in
the presence of surfactant micelles. It can be envisaged that the
binding/solubilization sites on the micelle approach saturation as the
solute concentration increases beyond the crystal solubility, hence the
amount of “free” drug increases more rapidly with concentration than
would be predicted by assuming a constant value of Km. It is therefore
apparent that care should be taken to evaluate the supersaturation in
complex solutions in order to enable better predictions of membrane
transport rates to be made. This is analogous to the issues with
determining drug permeability in the presence of surfactants,
described above.
4. Biological absorption of insoluble drugs from the ASD: A mass
transport process
When solid dosage forms are administered orally, the drug needs to
dissolve in the human gastrointestinal fluids before the drug molecules
can be absorbed into systemic circulation and subsequently transported
to the site(s) of action. The human gastrointestinal fluid, however, is not
a simple homogeneous system. Besides water, it contains naturally
secreted fluids which typically contain micelle-forming surfactants. On
the wall of the intestine, there are layers of mucus. In addition, it is high-
ly dynamic. For example, the amount of fluid secretion varies depending
on food intake; the mucus layer is constantly moving down the gastro-
intestinal tract (GIT). In addition, it periodically contains food sub-
stances. Pharmaceutical dosage forms also always bring in some
excipients into the system. The highly complex and dynamic human
GIT is considered as an in vivo “dissolution vessel.” Its complexity and
dynamic nature present significant difficulties in understanding the dis-
solution and subsequent in vivo performance of pharmaceutical oral
dosage forms.
Conceptually, however, we can simplify the system to aid under-
standing. As shown in Fig. 12, we can view the human GIT as a collection
of interconnected “compartments.” The central compartment is the
aqueous compartment into which drug dissolves. This is also the com-
partment that is in direct contact with intestinal membranes. Within
this compartment, an equilibrium can exist between ionized and union-
ized drug depending on the pKa of the molecule and the local pH. Drug
molecules in this compartment partition into the intestinal membrane
to begin the passive absorption process. However, only the unionized
 L.S. Taylor, G.G.Z. Zhang / Advanced Drug Delivery Reviews 101 (2016) 122–142
Fig. 12. Schematic representation of possible drug microenvironments in the gastrointestinal tract. Adapted from Ref. [79].
drug is thought to undergo passive absorption [89]. Peripheral to the
central aqueous compartment, there are many other compartments or
microenvironments that the drug can reside in. These compartments in-
teract with the central compartment and can act as reservoirs whereby
drug will move from these compartments into the central compartment
at a given rate as drug is absorbed by passive transport. Drug molecules
can transport directly to these compartments from the administered
dosage forms, they can also first dissolve in the central aqueous com-
partment and redistribute to these compartments. For instance, the
drug can dissolve in the aqueous compartment and then partition into
micelles and lipid vesicles. If dissolution from the dosage forms leads
to supersaturated aqueous solutions, the drug can crystallize or undergo
LLPS (or GLPS, if the Tg of the water saturated amorphous drug is higher
than physiological temperature).
Only the non-ionized drug molecules in the central aqueous com-
partment are able to permeate the GI membrane, and the concentration,
or more specifically, the thermodynamic activity of drug in this com-
partment dictates the kinetics of passive absorption from this compart-
ment, rather than the total concentration of the solubilized drug in all
compartments. The presence of drug in other compartments, can how-
ever, play an important role in feeding the central compartment, which
in turn will also influence the absorption process depending on the rel-
ative kinetics of the various processes. The next section of the review
centers on this idea and discusses the many physical processes occur-
ring during dissolution and how they impact the evolution of drug ther-
modynamic activity in the central aqueous compartment. From these
analyses, formulation approaches can be developed to manipulate the
drug activity in the central aqueous compartment in order to improve
oral absorption.
4.1. Physical processes and considerations that impact dissolution of the
ASD and absorption
While there are many different types of supersaturating dosage
forms, the focus herein will be on amorphous solid dispersions. Upon
oral administration and dissolution of an ASD, many physical processes
occur, in parallel and in series, which could impact the oral absorption
and performance. Characterization of the occurrence or absence of
certain processes forms the basis of ASD formulation development.
Competition among various processes, in particular dissolution rate,
crystallization kinetics, concentration evolution and LLPS, and
absorption rate is critical in determining the thermodynamic activity
of the drug in the central aqueous compartment. Here, we will briefly
133
discuss some of the factors that impact the dissolution of amorphous
solid dispersions and the evolution of the solution concentration in
the central aqueous compartment.
4.1.1. Disintegration and erosion
Depending on the dosage form, tablets or capsules, disintegration
leads to the breakage of the particles into smaller fragments. This pro-
cess increases the contact surface area between the solids and the disso-
lution medium, and facilitates the mass transfer from the solid to the
solution. Whether the dosage forms disintegrates or erodes depends
on the matrix polymer used in the ASD, the presence or absence of a
disintegrant in the formulation, and whether other formulation compo-
nents, such as surfactants, are included. For instance, formulations using
copovidone and HPMC to form the ASD matrix tend to erode when they
come in contact with water, while formulations which contain HPMCAS
as the matrix polymer can be designed to disintegrate.
4.1.2. Driving force for dissolution
In order for solids to dissolve, there must be a thermodynamic driv-
ing force. For simple one component solids, such as crystalline and neat
amorphous drugs, the driving force is the chemical potential gradient
between the dissolving material and the solution, which in turn
depends on the solubility of the solid in the dissolution medium, as
shown by Eq. 1. However, for multi-component solids, the thermody-
namic driving force for dissolution is poorly understood. It appears
that some, but not all, ASDs can dissolve to exceed the amorphous
solubility of drug, undergoing LLPS and achieving the maximum
possible supersaturation, while others have a lower driving force for
dissolution and achieve much lower maximum concentrations [39,90].
For felodipine ASDs formulated with hydrophilic polymers, it was
noted that when the drug:polymer ratio was low, the solids dissolved
rapidly under nonsink conditions with the generation of nanoparticles
indicating that the amorphous solubility was exceeded during the dis-
solution process [44]. However, with increased drug:polymer ratio,
the dissolution rate was not only slower, but no nanoparticles were gen-
erated and the concentration approached, but did not exceed, the amor-
phous solubility. This is clearly an area that warrants more investigation
and it can be anticipated that the properties of individual components
(drug, polymer, and surfactant), their ratios, the type and extent of in-
termolecular interactions, ionization states in the solid state and during
dissolution, etc. all have an impact on the overall dissolution rate and
extent.
134 L.S. Taylor, G.G.Z. Zhang / Advanced Drug Delivery Reviews 101 (2016) 122–142
4.1.3. Kinetics of dissolution
Universal to the dissolution of all solids, wetting of the ASD particles
is a pre-requisite for dissolution into aqueous media. It is well known
that surfactants can improve the wetting of hydrophobic materials
[91]. Although there are surfactants present in the human GIT, these
are predominantly found in the small intestine, hence the composition
of the ASD in terms of the drug loading and types of excipients used in
the formulation will still be important for wetting. Particle size, of
course, is an important factor for dissolution when the dosage form dis-
integrates. Hence amorphous nanoparticles may show improved disso-
lution properties relative to amorphous formulations with larger
particles [92]. The ratio of the drug and polymer in the ASD is also a crit-
ical factor that impacts the dissolution kinetics whereby high drug load-
ing dispersions formed with hydrophilic polymers typically dissolve
more slowly than low drug loading dispersions [9,39,44], often leading
to differences in the subsequent phase behavior including the tendency
to undergo LLPS following dissolution [39]. Congruency during dissolu-
tion, i.e. the relative dissolution rate of the drug versus the polymer, is
also expected to impact the time profile of the dissolution rate, where
the dissolution rate is either sustained throughout the entire dissolution
process or decreases over time. This was observed for sulfathiazole-PVP
ASDs where it was noted that the polymer dissolved faster than the
drug, leading to a reduction in drug dissolution rate with time; this
was more pronounced for dispersions containing a higher
drug:polymer ratio [9]. Faster dissolution of the polymer relative to
the drug dissolution will lead to surface enrichment of the drug,
which will not only reduce the dissolution rate, but could also lead to
crystallization of undissolved solid [93]. Corrigan [94] proposed a
model suggesting that the dissolution rate of the drug will be controlled
by the properties of the drug at moderate to high drug loadings, where-
as the drug release rate will be polymer-controlled release at low drug
loadings.
4.1.4. Phase behavior of the ASD solids during dissolution
When an ASD particle comes in contact with the dissolution medi-
um, it will absorb water. Thus, the properties of the ASD will change ac-
cordingly. Water, being a very effective plasticizer, will lower the Tg of
the ASD to an extent that depends on the amount of water absorbed
[95]. If the “wet” Tg of ASD is below physiological temperature, there
is greatly enhanced mobility which facilitates nucleation [96] and crys-
tal growth within the amorphous phase [97]. Therefore, it is important
that the polymers used to formulate the ASD inhibit crystallization of
the API in the matrix, otherwise the rate and extent of dissolution will
be greatly reduced [98,99]. This was observed for amorphous nifedipine
systems. In the absence of a polymer, amorphous nifedipine crystallized
immediately upon exposure to water and no supersaturation could
be observed [98]. However, when formulated with HPMC, matrix crys-
tallization was inhibited, leading the generation of a supersaturated
solution.
The enhanced mobility resulting from water sorption could also
induce amorphous–amorphous phase separation of ASD [100], a com-
plicated process that is not only determined by the thermodynamics
of multi-component system (including water) [101–104] but also the
kinetics of water diffusion and consequent changes. In this scenario,
the initially miscible system will form drug-rich and a drug-lean do-
mains within the ASD matrix. ASDs formulated with PVP in particular
have been found to be quite susceptible to water-induced amor-
phous–amorphous phase separation [100,103]. The drug-rich regions
are obviously more likely to crystallize due to the depletion of polymer,
hence amorphous–amorphous phase separation is a precursor to crys-
tallization [100]. Stronger intermolecular interactions among the ASD
components and formulation with less hygroscopic polymers tend to
lead to formulations that resist phase separation [104] and this in turn
is thought to be of importance for maintaining a high dissolution rate
throughout the process.
4.1.5. Phase behavior in aqueous phase
As the dissolution progresses, the concentration of drug in aqueous
solution increases, often very rapidly in the case of ASDs with a high
polymer:drug ratio [44]. Meanwhile, the dissolved drug also partitions
into other “microenvironments” that are present in the solution, such
as bile salt micelles that are present in certain regions of the GIT.
When the concentration exceeds the solubility of the crystalline drug,
it is supersaturated, and there exists a driving force for crystal nucle-
ation from the aqueous phase. If nucleation and crystal growth occur
within a timeframe relevant to oral absorption, the drug molecules
that are present in the high energy state, i.e. in the supersaturated solu-
tion, will transform to the lower energy state, i.e. the crystalline phase.
As crystals form, the supersaturation will be depleted, reducing the driv-
ing force for flux across the membrane. Although absorption could still
be enhanced relative to formulations containing the crystalline API,
due to the fact that the re-crystallized particles are small in size and
are wetted by the aqueous medium, the benefit of using an ASD is most-
ly lost. For example, it was observed that felodipine-PVP dispersions dis-
solved rapidly to give high supersaturation, but that the supersaturation
was short-lived due to crystallization of the drug from the solution
phase [44]. Therefore, incorporation of effective crystallization inhibi-
tion polymer(s) into the formulation is essential to prevent crystalliza-
tion from the supersaturated solution and allow continued increases
in concentration and maintenance of supersaturation [44,105]. Assum-
ing the crystallization inhibitor is sufficiently effective, the concentra-
tion of drug in the solution will continue to increase as dissolution
progresses. However, at some point, the homogeneous aqueous solu-
tion will undergo spinodal decomposition to two liquid phases (LLPS),
as described earlier. The system has now reached its highest possible
thermodynamic activity/supersaturation. From this point onwards, fur-
ther dissolution of the ASD increases the amount of the drug-rich phase,
but does not increase the supersaturation. In this complex process, it is
important to have a good estimate of the distribution of drug in the dif-
ferent states: aqueous phase, drug-rich droplets, and in the solubilized
form (micelles or complexes) since this will impact membrane trans-
port rates. Due to the low inherent solubility (i.e. drugs with high dose
number), it is obvious that, at any given time, only a very small fraction
of the administered dose is present in the central aqueous compart-
ment. Therefore a well-designed formulation will rely on transferring
the drug into drug-rich liquid droplets and compartments such as mi-
celles in order to achieve good absorption, since these forms of the
drug will equilibrate rapidly with the central aqueous compartment
and continuously replenish the free drug molecules in this compart-
ment as absorption proceeds. This situation will be illustrated in more
detail subsequently. In contrast, transfer from undissolved solid dosage
form or crystalline drug to the central compartment is expected to be
much slower [58], whereby dissolution is the rate limiting process,
leading a reduced absorption rate.
4.1.6. Physical stability of the drug-rich phase that evolves following LLPS
Typically the liquid droplets and glassy particles that are formed via
spinodal decomposition are small in size, in the nanometer range, when
first formed [40,43]. This is the basis for the fast equilibration with or
dissolution into the central aqueous compartment, and the expected
maintenance of high concentration/activity as absorption proceeds.
However, these drug-rich nanospecies are mainly comprised of hydro-
phobic drug molecules. In addition, the large surface area of the system
is thermodynamically unstable. Over time, these initially nano-sized
drug-rich particles tend to grow into larger droplets in order to reduce
the surface area in contact with the aqueous phase, thus reducing the
specific surface area [40,44]. Similarly they can agglomerate, reducing
their interaction with water, leading to a reduced effective surface
area for mass transfer into the central aqueous compartment. Further-
more, the interfaces present due to the formation of a second phase
provide a site for heterogeneous nucleation, and hence crystallization
may be favored in such solutions.
 L.S. Taylor, G.G.Z. Zhang / Advanced Drug Delivery Reviews 101 (2016) 122–142 135

0.7
Both surfactants and polymers have been shown to reduce/prevent
coalescence and agglomeration of nanodroplets [40,44]. In addition, A
0.6
certain polymers are able to stabilize against crystallization [40], while
surfactants may inhibit or promote crystallization depending on the
0.5
specific drug and surfactant [40,48,106]. Therefore, judicious selection
0.5
of excipients that not only prevent crystallization, but which also

Flux (µg/h.cm2)
0.4 1
promote and maintain drug-rich nanodroplets, is expected to be critical 2
to the performance of the dosage form. 3
0.3
4
4.1.7. Flux through GI membrane 5
0.2 6
This is the last step of the many processes described above, that
7
occur in parallel and in series, and that ultimately lead to absorption
0.1
of the drug, and subsequent entry into the systemic circulation.
Although many biological processes such as gut-wall metabolism and
0.0
efflux can influence the extent of absorption, membrane flux is the ulti-
mate physical process that directly links to passive absorption. All the 0 1 2 3 4 5
other processes discussed thus far, in one way or another, impact this
Time (h)
final step of absorption. As discussed in earlier sections, the flux through
a membrane, whether an artificial membrane or the intact human gas-
1.2
trointestinal membrane, is dictated by the thermodynamic activity of
the diffusant present in the central aqueous compartment, on the B
donor side of the membrane. Thus, it is important to accurately measure 1.0
or estimate the thermodynamic activity of the drug in the central aque-
ous compartment, rather than the overall concentration of drug in solu- 0.8 10
tion, especial when complex structures are present peripheral to the 15
aqueous compartment. For BCS Class IV compounds, i.e. low permeabil- Flux (µg/h.cm2) 20
ity compounds, permeation through the membrane is slow. Thus 0.6
diffusion through the membrane likely becomes the rate determining
step for the absorption process rather than the many other processes 0.4
discussed. In this case, the thermodynamic activity of the drug may
not be the most critical parameter that influences absorption; rather it
is the low permeability. For BCS Class II compounds, i.e. high permeabil- 0.2

ity compounds, however, diffusion through the membrane is fast.

Therefore, it is advantageous to provide a higher thermodynamic activ- 0.0
ity of the drug in the central aqueous compartment via supersaturation 0 2 4 6 8 10 12 14 16 18
in order to drive transport across the membrane. Furthermore, the rate Time (h)
of removal of the drug from the central aqueous compartment may be
sufficiently high as compared to the rate of drug entering the central Fig. 13. Flux as a function of various initial donor cell concentrations, obtained using a flow
aqueous compartment (e.g. by dissolution) that the effective activity through diffusion system. Triangles represent the experimental data. Dotted lines
of drug in the central compartment is significantly impacted. In extreme represent data fit to a mathematical model. A) Experiments carried out for 5 h at
concentrations lower than the amorphous solubility showing a decline in flux at longer
cases, the absorption or flux through the membrane is dictated
time periods. B) Experiments carried out for 16 h at initial concentrations above the
completely by dissolution, i.e. absorption is dissolution rate-limited. amorphous solubility showing sustained flux at a maximum value. The duration of the
This situation can be overcome by ensuring that the drug is present sustained flux depends on the excess drug concentration beyond the amorphous
in a reservoir with a low barrier for entering the central aqueous solubility. Legend shows the initial donor cell concentration in μg/mL. Data adapted
compartment. from Ref. [107].

4.2. The “reservoir” effect of LLPS: An ideal case for absorption
In an ideal situation, the thermodynamic activity of the drug in the
central compartment is maintained at the highest possible value,
which is the amorphous solubility, until the dose is completely
absorbed. For drugs with high dose numbers, this can be achieved by
formulating an ASD to dissolve rapidly and completely, forming a super-
saturated aqueous solution in equilibrium with drug-rich nanodroplets.
In the absence of crystallization, as the drug is absorbed from the central
compartment, the drug-rich phase will replenish absorbed drug,
maintaining the supersaturation at the maximum value.
The fast equilibrium between the aqueous solution and these
drug-rich nanodroplets is expected because these nanodroplets
are pre-wetted by water and have very high specific surface area
[29,35,40,58]. In a recent study, using an automated flow-through dif-
fusion system and clotrimazole as a model compound, the advantage of
using solutions containing drug-rich nanodroplets for sustained mem-
brane flux has been clearly demonstrated [107]. As seen in Fig. 13A,
when a solution of clotrimazole is placed on the donor side, the mole-
cules start to diffuse through the membrane to reach the receiver side.
The rate of mass transfer through the membrane, i.e. the flux, is propor-
tional to the thermodynamic activity of the drug on the donor side [78].
For initial donor concentrations between 0.5 and 7 μg/mL, which are
below the amorphous solubility of 7.5 μg/mL, there is proportionality
between the maximum value of the flux and the initial donor concen-
tration. For a given solution, as the drug molecules diffuse through the
membrane over time, the mass, concentration, and activity of the
drug on the donor side decrease accordingly leading to the progressive
decrease of flux through the membrane over time. This behavior is con-
sistent with the physical chemistry of undersaturated and supersaturat-
ed solutions, and the principles of mass transport [72,78,108].
When the initial concentration of clotrimazole on the donor-side ex-
ceeds its amorphous solubility of 7.5 μg/mL, however, LLPS occurs. As
shown in Fig. 14, 200–250 nm droplets of clotrimazole are initially
present on the donor side. As the initial concentration increases beyond
the amorphous solubility, the number of droplets is also increasing in
accordance with mass balance. The time profiles for the flux of these
supersaturated solutions exhibit different patterns from those that are
single phase solutions at or below the amorphous solubility. As shown
in Fig 13B, upon initial rise to the plateau, the flux remains constant at
136 L.S. Taylor, G.G.Z. Zhang / Advanced Drug Delivery Reviews 101 (2016) 122–142

300 Size 16 on membrane transport, the next logical step is to analyze the impact of
Particle concentration these processes on absorption. This is best done by schematic illustra-
250 tions of different potential scenarios. In this section, we simulate the

Particle concentration
(no. of particles/mL)
12 evolution of drug activity in the central aqueous compartment, the cor-
200 responding diffusive flux through the membrane, percentage of dose
Size(nm)

absorbed, and the percentage of administered dose present in the cen-

150
100
50
0 0
10 15 20
Concentration in donor compartment (µg/mL)
Fig. 14. Size and particle concentration of the nanodroplets for varying drug
concentrations above the amorphous solubility. Data adapted from Ref. [107].
its maximum value for some time, before it eventually decreases. The
length of the plateau roughly correlates with the extent of initial con-
centration over and beyond the amorphous solubility, i.e. the amount
of drug present in the solution in the form of nanodroplets.
This study clearly demonstrates the “reservoir” effect of a system that
has undergone LLPS. As the free drug molecules in the aqueous phase dif-
fuse through the membrane, drug molecules in the nanodroplets rapidly
replenish the aqueous phase in order to maintain the same thermody-
namic activity of the drug in the two phases. As a consequence of this
fast equilibrium, the activity of drug in the aqueous phase is maintained
at its maximum value, and the flux through the membrane is also main-
tained at its maximum value until the nanodroplets are consumed.
4.3. Interplay among the different physical processes during dissolution and
absorption
After considering the many processes that can occur during the dis-
solution of an ASD, and discussing the beneficial reservoir effect of LLPS
x108
8 tral aqueous compartment as a function of time for various scenarios.
Here we assume that the drug of interest has a crystalline and amor-
phous solubility of 1 and 10 μg/mL respectively, thus, the activity of the
4 amorphous phase is 1, and the activity of the crystalline phase is 0.1;
the dose is assumed to be 5 mg; the volume of central aqueous compart-
ment in GIT is 100 mL, which is consistent with recent findings in humans
[109] and, that the drug is absorbed solely via passive absorption. The flux
of the drug (which is related to the permeability) was chosen to represent
a situation where the permeability is high, and the absorption is limited
by the amount of free drug in solution. While these simulations do not
contain sufficient parameters to represent the complex biological situa-
tion, they serve the purpose of illustrating the potential impact of LLPS
and crystallization on absorption, enabling various situations to be com-
pared in terms of the impact on the extent of absorption.
4.3.1. Case 1: Instantaneous achievement of LLPS, whereby the nanodroplets
are physically stable throughout the absorption process
In this case, the administration of the dosage form leads to LLPS rap-
idly, either via fast dissolution from the solid formulation or via dilution/
pH change of a solution, and generates nanodroplets that are in fast
equilibrium with the aqueous phase. Within a short period of time
relative to the absorption process, the administered dose is distributed
between the central aqueous compartment and the LLPS compartment
(20% in aqueous compartment, 80% in LLPS compartment).
As seen in Fig. 15, the absorption, as indicated by the membrane flux,
is constant for 6 h. During this period, the activity of drug is constant at
its maximum value of 1 (defined as the activity of the water-saturated
amorphous material). This is due to the reservoir effect of a solution
that has undergone LLPS. The percentage of the dose absorbed increases
linearly over time, reaches 96% after 6 h and 100% in 8 h. Beyond 6 h, the
nanodroplets are consumed, and the solute activity and flux decrease

A B
2500 100

2000 80

1500 60

1000 40

500 20

0 0
0 1 2 3 4 5 6 7 8 0 1 2 3 4 5 6 7 8

C D
1 20

0.8
15
0.6
10
0.4
5
0.2

0 0
0 1 2 3 4 5 6 7 8 0 1 2 3 4 5 6 7 8

Fig. 15. Time profiles for Case 1. A) Flux through the membrane (arbitrary units); B) percent of drug absorbed; C) activity of drug; D) percent of administered drug in the central aqueous
compartment. The x-axes are time (h).
 L.S. Taylor, G.G.Z. Zhang / Advanced Drug Delivery Reviews 101 (2016) 122–142 137

progressively. Throughout the 6 h period of absorption, the percentage
of the total administered dose that is present in the central aqueous
compartment is constant at 20%. Initially, the rest of the administered
dose, 80%, is present as nanodroplets. As the absorption progresses
over time, the fraction present in the drug-rich nanodroplet compart-
ment decreases progressively until it is completely consumed. Of the
administered dose, 80% of the molecules absorbed originate from the
drug-rich nanodroplet compartment.
We term the dissolution and absorption behavior described in Case 1
as “spring and plateau.” “Spring” reflects the fast dissolution of the solid
and the rapid achievement of concentrations where LLPS/GLPS occurs.
From a physical chemistry point of view, this represents the thermody-
namic limit for supersaturation. “Plateau” refers to the maintenance of
this highest possible supersaturation throughout the absorption process
by replenishment of the drug in the central compartment from the
drug-rich phase, avoiding crystallization through formulation with
appropriate inhibitors. We believe that this should be the goal of
every enabling formulation with high dose number.
4.3.2. Case 2: Instantaneous achievement of LLPS, but system is not
physically stable due to crystallization
Here we consider three sub-types of behavior falling into this cate-
gory. As for Case 1, the administration of the dosage form is considered
to rapidly result in LLPS leading to the generation of nanodroplets that
are in fast equilibrium with the aqueous phase. The administered dose
is initially distributed between the central aqueous compartment and
the LLPS compartment with a 20/80 ratio. However, the system is
physically unstable due to crystallization, and hence changes over
time. The three sub-types will be discussed separately.
4.3.2.1. Case 2a. In this scenario, the drug crystallizes rapidly and starts to
deplete the aqueous concentration due to the formation of crystals after
30 min, reaching the solubility of the crystalline phase after another
30 min. The recrystallized solids are assumed to have a small particle
size, therefore, are able to dissolve fast enough to maintain the concen-
tration in the central aqueous compartment at the intrinsic solubility of
1 μg/mL.
Because of the rapid crystallization, the activity of drug decreases
from the initial value of 1 to 0.1 within an hour. Afterwards, the activity
is constant, but at a low level of 0.1. The flux decrease correspondingly
with the activity. The time profile for the fraction of dose absorbed ex-
hibits a change in slope because of the change of solute activity and
flux. After 8 h, 23% of the administered dose is absorbed. Initially, 20%
of the dose is in the central aqueous compartment while 80% in the
drug-rich nanodroplet compartment. Upon crystallization, the drug in
the nanodroplet compartment transfers to the crystals compartment,
either directly in or on the surface of the liquid droplets or through a
solution-mediated process. At the end of 1 h, the central aqueous com-
partment contains only 2% of the administered dose, less than 12% of the
dose is absorbed, the rest of the administered dose, more than 86%, is
present in the crystals compartment. Over the next 7 h, another 11%
of the dose is absorbed by dissolution of the crystals, which is a much
slower process due to the reduced activity of the solute.
4.3.2.2. Case 2b. In Case 2b, a similar situation to Case 2a is envisaged,
except in this instance, the recrystallized solids have a higher apparent
solubility (3 μg/mL), than the intrinsic solubility of 1 μg/mL. This could
be due to either recrystallization to a metastable polymorph, where
the recrystallized material has low crystallinity or where the recrystal-
lized material is highly defected.
For the first hour, the profile is the same as for Case 1a. Subsequently,
the solute activity is maintained at a 3-fold higher value, and the flux is
also a factor of 3 higher. At 8 h, the percentage of dose absorbed in this
situation is 46%, as compared to 23% in Case 2a. At the end of 1 h, the
drug molecules are distributed in different compartments with the
same ratios as in Case 2a. However, because of the difference in the na-
ture of the crystals compartment in Case 2b, i.e. the residual supersatu-
ration, the absorption over the next 7 h increases to 34% as compared to
11% for Case 2a.
4.3.2.3. Case 2c. The crystallization kinetics of different compounds is
known to vary considerably [45]. After achieving LLPS initially as for
all the Case 2 scenarios, if nucleation is delayed such that depletion of
the aqueous concentration does not occur until 2 h (as opposed to

A B
2500 100

a a
2000 80
b b
1500 60 c
c

1000 40

500 20

0 0
0 1 2 3 4 5 6 7 8 0 1 2 3 4 5 6 7 8

C D
1 20

0.8 a a
15
b b
0.6
c c
10
0.4
5
0.2

0 0
0 1 2 3 4 5 6 7 8 0 1 2 3 4 5 6 7 8

Fig. 16. Time profiles for Case 2. A) Flux through the membrane (arbitrary units); B) percent of drug absorbed; C) activity of drug; D) percent of administered drug in the central aqueous
compartment. The x-axes are time (h).
138 L.S. Taylor, G.G.Z. Zhang / Advanced Drug Delivery Reviews 101 (2016) 122–142

30 min), and crystal growth is slow so that it takes 2 h (not 30 min) to 1

decrease to a final concentration of 3 μg/mL (instead of the intrinsic
solubility of 1 μg/mL), quite different flux and absorption profiles are 0.8
predicted (Fig. 16). This type of behavior might arise from a drug with
a moderate crystallization tendency, or it could result from the formula- With absorption
0.6
tion design where a crystallization inhibitor is incorporated into the
Without absorption
formulation. In this case, the solute activity remains at a high value of
0.4
1 for 2 h, then gradually decreases to 0.3 over the next 2 h, and is main-
tained at 0.3 for the remaining 4 h. The flux through the membrane par-
allels the activity profile. A significant percentage of the dose is absorbed 0.2
(over 32%) during the first two hours when the activity is high. At the
4 h point, over 51% of the dose is absorbed. At the end of 8 h, 71% of 0
the administered dose is absorbed. Compared to Case 2a (23%) and 0 1 2 3 4 5 6 7 8
Case 2b (46%), this is much improved.
This case mirrors the situation frequently referred to as “spring and Fig. 18. Comparison of drug activities in central aqueous compartment with and without
absorption. The x-axes is time (h).
parachute” which describes the behavior of enabling formulations
with superior performance [76]. “Spring” reflects the fast dissolution
of the solids, leading to a high concentration that is supersaturated
(not necessarily reaching LLPS); while “parachute” refers to the slow compartment increases, to a final percentage of ~ 6% after 8 h (see
decrease of the concentration afterwards so as to maximize absorption. Fig. 17).
This situation occurs when the ASD formulation design is sub-
4.3.3. Case 3: ASD is dissolving at a slow rate and LLPS is not achieved optimum, e.g. due to a high drug loading which retards dissolution [9].
In Case 3, we assume that the ASD is dissolving at a slower rate and Although the concentration does not “spring” to a high level, it still
whereby the concentration never exceeds the amorphous solubility and leads to supersaturation and absorption is improved relative to a
hence LLPS does not occur. However, the system is physically stable crystalline formulation. When the dissolution rate is comparable to
with no crystallization occurring. the absorption rate, the activity in the central aqueous compartment
In this case, the activity of the dissolved drug slowly increases, might be significantly impacted by absorption. Fig. 18, where the solute
reaching the activity of the crystalline phase in about an hour, subse- activity in the central aqueous compartment with and without absorp-
quently becoming supersaturated. The rise of solute activity slows tion are compared, demonstrates this point.
down at later time points and has almost reached a plateau with a
final activity value of 0.29. The flux through the membrane parallels
the solute activity. The curve for the percentage of dose absorbed 4.3.4. Case 4: Crystalline suspensions
shows that after 8 h, 30% of the administered dose is absorbed. For completeness of the analysis, we include simulations on crystal-
Compared to the crystalline suspension (b 13%, see below), this is a sig- line suspensions. In this case, we assume the best case scenario where
nificant improvement. Initially, all drug molecules are present in the un- the crystals are able to maintain the concentration at the intrinsic solu-
dissolved ASD. Gradually, the percentage of dose in the central aqueous bility (1 μg/mL) in the central aqueous compartment.

A B
2500 100

2000 80

1500 60

1000 40

500 20

0 0
0 1 2 3 4 5 6 7 8 0 1 2 3 4 5 6 7 8

C D
1 20

0.8
15
0.6
10
0.4
5
0.2

0 0
0 1 2 3 4 5 6 7 8 0 1 2 3 4 5 6 7 8

Fig. 17. Time profiles for Case 3. A) Flux through the membrane (arbitrary units); B) percent of drug absorbed; C) activity of drug; D) percent of administered drug in the central aqueous
compartment. The x-axes are time (h).
 L.S. Taylor, G.G.Z. Zhang / Advanced Drug Delivery Reviews 101 (2016) 122–142 139

As seen in Fig. 19, the absorption is steady for 8 h, but at a much Table 1
lower level. The activity of drug is constant at 0.1. Only 2% of the Comparison of percentages of absorption at residence times of 4 h or 8 h.

administered dose is present in the central aqueous compartment at Residence time (h)
any time. Over the entire 8 h period, only 13% of the dose is absorbed. Case 8 4
Although the simulations and analyses of the above four cases are
1 100 65
not inclusive of all situations that could be encountered in vivo, and
2
they are oversimplified relative to the actual in vivo situation, they 2a 23 17
serve to illustrate that Case 1 represents the desired behavior from a 2b 46 27
dosage form performance and absorption point of view. Thus, in design- 2c 71 52
ing ASD formulations, or any other types of enabling formulation, we 3 30 11
4 13 6
should strive for the “spring and plateau” situation illustrated by Case
1, rather than the “spring and parachute” model, described in Case 2c.
time of the solids in stomach are drastically different. The impact of
4.4. Other considerations related to absorption pH will be very different for acidic, basic, and neutral drugs. The lower
pH in the fasted state will facilitate the dissolution of basic drugs,
The simulations in Section 4.3 are performed assuming an eight hour while discouraging the dissolution of acidic drugs. The longer residence
window in the GIT where absorption can occur, where the permeability time in stomach (not a site of absorption for most drugs) will also play a
of the drug is constant throughout this time period. This residence time role of the occurrence of the different processes described in Section 4.1.
is somewhat long, and the permeability assumption is unlikely to be Formulations based on enteric polymers, such as HPMCAS and
valid for a majority of compounds. For many drugs, there is either an ab- Eudragit® L 100–55, will not dissolve in stomach, while formulations
sorption window, or the permeability gradually decreases in the lower based on neutral polymers, such and PVP and HMPC, or cationic poly-
regions of the GIT, hence the effective residence time is shorter. mers, such as Eudragit® E PO, will begin to dissolve in stomach. Regard-
Table 1 compares the percentages of dose absorbed, based on the simu- less of whether there is dissolution, the dosage form will begin to absorb
lation in Section 4.3, when the residence time is reduced from 8 h to 4 h. water while in stomach. Therefore, the impact of residence time in the
It is obvious that, in all cases, the absorption suffers significantly. stomach will be very different for different formulations.
Besides consideration of the total time for absorption, it is also im-
portant to keep in mind the timescales of the various mass transport
processes listed in Section 4.1 in relation to the residence time in the 5. Concluding remarks
GIT and to each other. Some of these are clearly demonstrated in the
simulations, such as dissolution of the ASD, nucleation and crystal The physical chemistry of supersaturated solutions that evolve fol-
growth rates, and diffusive flux through the membrane. In addition, lowing the dissolution of solubility enhancing formulations is of great
the kinetics of phase separation in the ASD particles upon exposure to potential importance in influencing oral absorption. Factors influencing
dissolution media, in vitro or in vivo, should be considered in relation the properties of supersaturated solutions including the amorphous and
to the dissolution rate. crystalline solubility, liquid–liquid phase separation, and the presence
Another important consideration is fasted state versus fed state. of solubilizing substances need to be considered in order to understand
Comparing the two states, the pH of the gastric fluid and the residence their impact on the membrane transport rate. Consideration of the

A B
2500 100

2000 80

1500 60

1000 40

500 20

0 0
0 1 2 3 4 5 6 7 8 0 1 2 3 4 5 6 7 8

C D
1 20

0.8
15
0.6
10
0.4
5
0.2

0 0
0 1 2 3 4 5 6 7 8 0 1 2 3 4 5 6 7 8

Fig. 19. Time profiles for Case 4. A) Flux through the membrane (arbitrary units); B) percent of drug absorbed; C) activity of drug; D) percent of administered drug in the central aqueous
compartment. The x-axes are time (h).
140 L.S. Taylor, G.G.Z. Zhang / Advanced Drug Delivery Reviews 101 (2016) 122–142
thermodynamic and kinetic properties of supersaturated solutions can
aid in making predictions about mass transfer processes.
With respect to the complex oral absorption process, the evolution
and properties of supersaturated solutions need to be considered in
the context of the gastrointestinal tract environment. In particular,
there are many physical processes that can occur during the dissolution
of solubility-enabling amorphous solid dispersion formulations. These
processes govern the evolution of drug supersaturation in the central
aqueous compartment, which drives the passive absorption. Consider-
ation of the relative rates of transport between drug present in different
microenvironments and the central aqueous compartment in the gas-
trointestinal tract, as well as the relationship between drug thermody-
namic activity and absorption rate, is likely essential to gain a more
comprehensive understanding of the passive absorption process. For
delivery of poorly soluble drugs, we believe that pharmaceutical
scientists should consider the thermodynamic and kinetic benefits of
supersaturating dosage forms that dissolve to undergo LLPS, designing
ASD formulations that lead to the “spring and plateau” effect. Future ef-
forts should be directed towards accurate assessment of supersatura-
tion in complex media and connecting the in vitro observations
described herein, which point towards the importance of solute
thermodynamic activity as well as the reservoir effect, to the in vivo
performance of solubility enhancing formulations.
Disclosure
Purdue University and AbbVie jointly participated in writing,
reviewing, and approving the publication. Lynne S. Taylor is a professor
at Purdue University. She has no additional conflicts of interest to report.
Geoff G. Z. Zhang is an employee of AbbVie and may own AbbVie stock.
Acknowledgements
Helpful discussions with Drs. Yi Gao, Shweta Raina and David Alonzo
and Ms. Anura Indulkar are greatly appreciated.
References
[1] C.A. Lipinski, F. Lombardo, B.W. Dominy, P.J. Feeney, Experimental and computa-
tional approaches to estimate solubility and permeability in drug discovery and
development settings, Adv. Drug Deliv. Rev. 23 (1997) 3–25.
[2] T. Loftsson, M.E. Brewster, Pharmaceutical applications of cyclodextrins: basic
science and product development, J. Pharm. Pharmacol. 62 (2010) 1607–1621.
[3] C.J.H. Porter, C.W. Pouton, J.F. Cuine, W.N. Charman, Enhancing intestinal drug
solubilisation using lipid-based delivery systems, Adv. Drug Deliv. Rev. 60 (2008)
673–691.
[4] B.E. Rabinow, Nanosuspensions in drug delivery, Nat. Rev. Drug Discov. 3 (2004)
785–796.
[5] E. Merisko-Liversidge, G.G. Liversidge, Nanosizing for oral and parenteral drug
delivery: a perspective on formulating poorly-water soluble compounds using
wet media milling technology, Adv. Drug Deliv. Rev. 63 (2011) 427–440.
[6] V. Andronis, G. Zografi, Crystal nucleation and growth of indomethacin poly-
morphs from the amorphous state, J. Non-Cryst. Solids 271 (2000) 236–248.
[7] A. Saleki-Gerhardt, C. Ahlneck, G. Zografi, Assessment of disorder in crystalline
solids, Int. J. Pharm. 101 (1994) 237–247.
[8] C. Ahlneck, G. Zografi, The molecular basis of moisture effects on the physical and
chemical stability of drugs in the solid state, Int. J. Pharm. 62 (1990) 87–95.
[9] A. Simonelli, S. Mehta, W. Higuchi, Dissolution rates of high energy polyvinylpyr-
rolidone (PVP)-sulfathiazole coprecipitates, J. Pharm. Sci. 58 (1969) 538–549.
[10] A.P. Simonelli, S.C. Mehta, W.I. Higuchi, Dissolution rates of high energy
sulfathiazole-povidone coprecipitates II: characterization of form of drug control-
ling its dissolution rate via solubility studies, J. Pharm. Sci. 65 (3) (1976) 355.
[11] J.A. Baird, B. Van Eerdenbrugh, L.S. Taylor, A classification system to assess the
crystallization tendency of organic molecules from undercooled melts, J. Pharm.
Sci. 99 (2010) 3787–3806.
[12] S.B. Murdande, M.J. Pikal, R.M. Shanker, R.H. Bogner, Solubility advantage of amor-
phous pharmaceuticals: I. A thermodynamic analysis, J. Pharm. Sci. 99 (2010)
1254–1264.
[13] P.J. Marsac, S.L. Shamblin, L.S. Taylor, Theoretical and practical approaches for
prediction of drug-polymer miscibility and solubility, Pharm. Res. 23 (2006)
2417–2426.
[14] Y. Sun, J. Tao, G.G.Z. Zhang, L. Yu, Solubilities of crystalline drugs in polymers: an
improved analytical method and comparison of solubilities of indomethacin and
nifedipine in PVP, PVP/VA, and PVAc, J. Pharm. Sci. 99 (2010) 4023–4031.
[15] T. Cai, L. Zhu, L. Yu, Crystallization of organic glasses: effects of polymer additives
on bulk and surface crystal growth in amorphous nifedipine, Pharm. Res. 28
(2011) 2458–2466.
[16] A.D. Kwong, R.S. Kauffman, P. Hurter, P. Mueller, Discovery and development of
telaprevir: an NS3-4 A protease inhibitor for treating genotype 1 chronic hepatitis
C virus, Nat. Biotechnol. 29 (2011) 993–1003.
[17] N. Shah, R.M. Iyer, H.-J. Mair, D.S. Choi, H. Tian, R. Diodone, K. Fähnrich, A. Pabst-
Ravot, K. Tang, E. Scheubel, J.F. Grippo, S.A. Moreira, Z. Go, J. Mouskountakis, T.
Louie, P.N. Ibrahim, H. Sandhu, L. Rubia, H. Chokshi, D. Singhal, W. Malick,
Improved human bioavailability of vemurafenib, a practically insoluble drug,
using an amorphous polymer-stabilized solid dispersion prepared by a solvent-
controlled coprecipitation process, J. Pharm. Sci. 102 (2013) 967–981.
[18] L.S. Taylor, G. Zografi, Spectroscopic characterization of interactions between PVP
and indomethacin in amorphous molecular dispersions, Pharm. Res. 14 (1997)
1691–1698.
[19] S.H. Yalkowsky, Solubility and Solubilization in Aqueous Media, Oxford University
Press, New York, 1999.
[20] S.I. Sandler, Chemical, Biochemical, and Engineering Thermodynamics, John Wiley,
Hoboken, N.J., 2006
[21] J.W. Mullin, Crystallization, fourth ed. Butterworths, London, 2001.
[22] R. Paus, Y. Ji, L. Vahle, G. Sadowski, Predicting the solubility advantage of
amorphous pharmaceuticals: a novel thermodynamic approach, Mol. Pharm. 12
(2015) 2823–2833.
[23] G. Parks, L. Snyder, F. Cattoir, Studies on glass. XI. Some thermodynamic relations
of glassy and alpha-crystalline glucose, J. Chem. Phys. 2 (1934) 595–598.
[24] G. Parks, S. Thomas, D. Light, Studies on glass XII. Some new heat capacity data for
organic glasses. The entropy and free energy of dl-lactic acid, J. Chem. Phys. 4
(1936) 64–69.
[25] M.C. Brick, H.J. Palmer, T.H. Whitesides, Formation of colloidal dispersions of organ-
ic materials in aqueous media by solvent shifting, Langmuir 19 (2003) 6367–6380.
[26] S.B. Murdande, M.J. Pikal, R.M. Shanker, R.H. Bogner, Solubility advantage of
amorphous pharmaceuticals: II. Application of quantitative thermodynamic
relationships for prediction of solubility enhancement in structurally diverse
insoluble pharmaceuticals, Pharm. Res. 27 (2010) 2704–2714.
[27] B.C. Hancock, M. Parks, What is the true solubility advantage for amorphous
pharmaceuticals? Pharm. Res. 17 (2000) 397–404.
[28] L. Almeida e Sousa, S.M. Reutzel-Edens, G.A. Stephenson, L.S. Taylor, Assessment of
the amorphous “solubility” of a group of diverse drugs using new experimental
and theoretical approaches, Mol. Pharm. 12 (2014) 484–495.
[29] L.I. Mosquera-Giraldo, L.S. Taylor, Glass–liquid phase separation in highly supersat-
urated aqueous solutions of telaprevir, Mol. Pharm. 12 (2014) 496–503.
[30] G. Zografi, B. Hancock, Water–solid interactions in pharmaceutical systems, in:
D.J.A. Crommelin, K.K. Midha, T. Nagai (Eds.), Topics in Pharmaceutical Sciences
1993, Proceedings of International Congress on Pharmaceutical Sciences F.I.P.
Medpharm Scientific, Stuttgart, Germany 1994, pp. 405–419.
[31] R. Ceolin, M. Barrio, J.-L. Tamarit, N. Veglio, M.-A. Perrin, P. Espeau, Liquid–liquid
miscibility gaps and hydrate formation in drug–water binary systems: pressure–
temperature phase diagram of lidocaine and pressure–temperature–composition
phase diagram of the lidocaine–water system, J. Pharm. Sci. 99 (2010) 2756–2765.
[32] I.B. Rietveld, M.-A. Perrin, S. Toscani, M. Barrio, B. Nicolai, J.-L. Tamarit, R. Ceolin,
Liquid–liquid miscibility gaps in drug–water binary systems: crystal structure
and thermodynamic properties of prilocaine and the temperature–composition
phase diagram of the prilocaine–water system, Mol. Pharm. 10 (2013) 1332–1339.
[33] A.N. Campbell, A.J.R. Campbell, Concentrations, total and partial vapor pressures,
surface tensions and viscosities, in the systems phenol–water and phenol–
water–4% succinic acid, J. Am. Chem. Soc. 59 (1937) 2481–2488.
[34] E. Deneau, G. Steele, An in-line study of oiling out and crystallization, Org. Process.
Res. Dev. 9 (2005) 943–950.
[35] G.A. Ilevbare, L.S. Taylor, Liquid–liquid phase separation in highly supersaturated
aqueous solutions of poorly water-soluble drugs: implications for solubility
enhancing formulations, Cryst. Growth Des. 13 (2013) 1497–1509.
[36] P.E. Bonnett, K.J. Carpenter, S. Dawson, R.J. Davey, Solution crystallisation via a
submerged liquid–liquid phase boundary: oiling out, Chem. Commun. (2003)
698–699.
[37] L.B. Gower, Biomimetic model systems for investigating the amorphous precursor
pathway and its role in biomineralization, Chem. Rev. 108 (2008) 4551–4627.
[38] H.-H. Tung, E.L. Paul, M. Midler, J.A. McCauley, Crystallization of Organic
Compounds, John Wiley & Sons, Inc., 2008 107.
[39] M.J. Jackson, U.S. Kestur, M.A. Hussain, L.S. Taylor, Dissolution of danazol
amorphous solid dispersions: supersaturation and phase behavior as a function
of drug loading and polymer type, Mol. Pharm. 13 (2016) 223–231.
[40] G.A. Ilevbare, H. Liu, J. Pereira, K.J. Edgar, L.S. Taylor, Influence of additives on the
properties of nanodroplets formed in highly supersaturated aqueous solutions of
ritonavir, Mol. Pharm. 10 (2013) 3392–3403.
[41] A.F.A. Aisha, Z. Ismail, K.M. Abu-salah, A.M.S.A. Majid, Solid dispersions of α-
mangostin improve its aqueous solubility through self-assembly of nanomicelles,
J. Pharm. Sci. 101 (2012) 815–825.
[42] T. Tachibana, A. Nakamura, A method for preparing an aqueous colloidal dispersion
of organic materials by using water-soluble polymers: dispersion of Β-carotene by
polyvinylpyrrolidone, Kolloid Z. Z. Polym. 203 (1965) 130–133.
[43] A.S. Indulkar, K.J. Box, R. Taylor, R. Ruiz, L.S. Taylor, pH-dependent liquid–liquid
phase separation of highly supersaturated solutions of weakly basic drugs, Mol.
Pharm. (2015).
[44] D.E. Alonzo, Y. Gao, D.L. Zhou, H.P. Mo, G.G.Z. Zhang, L.S. Taylor, Dissolution and
precipitation behavior of amorphous solid dispersions, J. Pharm. Sci. 100 (2011)
3316–3331.
 L.S. Taylor, G.G.Z. Zhang / Advanced Drug Delivery Reviews 101 (2016) 122–142
[45] B. Van Eerdenbrugh, S. Raina, Y.-L. Hsieh, P. Augustijns, L. Taylor, Classification of
the crystallization behavior of amorphous active pharmaceutical ingredients in
aqueous environments, Pharm. Res. 31 (2014) 969–982.
[46] S.L. Raghavan, A. Trividic, A.F. Davis, J. Hadgraft, Crystallization of hydrocortisone
acetate: influence of polymers, Int. J. Pharm. 212 (2001) 213–221.
[47] D.E. Alonzo, S. Raina, D. Zhou, Y. Gao, G.G.Z. Zhang, L.S. Taylor, Characterizing the
impact of hydroxypropylmethyl cellulose on the growth and nucleation kinetics
of felodipine from supersaturated solutions, Cryst. Growth Des. 12 (2012)
1538–1547.
[48] J. Chen, L.I. Mosquera-Giraldo, J.D. Ormes, J.D. Higgins, L.S. Taylor, Bile salts as
crystallization inhibitors of supersaturated solutions of poorly water-soluble
compounds, Cryst. Growth Des. 15 (2015) 2593–2597.
[49] S.A. Raina, B.V. Eerdenbrugh, D.E. Alonzo, H. Mo, G.G.Z. Zhang, Y. Gao, L.S. Taylor,
Trends in the precipitation and crystallization behavior of supersaturated aqueous
solutions of poorly water-soluble drugs assessed using synchrotron radiation, J.
Pharm. Sci. 104 (2015) 1981–1992.
[50] S.B. Murdande, M.J. Pikal, R.M. Shanker, R.H. Bogner, Aqueous solubility of crystal-
line and amorphous drugs: challenges in measurement, Pharm. Dev. Technol. 16
(2010) 187–200.
[51] S. Ozaki, I. Kushida, T. Yamashita, T. Hasebe, O. Shirai, K. Kano, Evaluation of drug
supersaturation by thermodynamic and kinetic approaches for the prediction of
oral absorbability in amorphous pharmaceuticals, J. Pharm. Sci. 101 (2012)
4220–4230.
[52] L. Lindfors, S. Forssén, P. Skantze, U. Skantze, A. Zackrisson, U. Olsson, Amorphous
drug nanosuspensions. 2. Experimental determination of bulk monomer concen-
trations, Langmuir 22 (2006) 911–916.
[53] S. Raina, D. Alonzo, G.Z. Zhang, Y. Gao, L. Taylor, Using environment-sensitive
fluorescent probes to characterize liquid–liquid phase separation in supersaturat-
ed solutions of poorly water soluble compounds, Pharm. Res. 32 (2015)
3660–3673.
[54] J.A. Thomson, P. Schurtenberger, G.M. Thurston, G.B. Benedek, Binary liquid phase
separation and critical phenomena in a protein/water solution, Proc. Natl. Acad. Sci.
84 (1987) 7079–7083.
[55] L. Lafferrere, C. Hoff, S. Veesler, Study of liquid–liquid demixing from drug solution,
J. Cryst. Growth 269 (2004) 550–557.
[56] W. Haiss, N.T.K. Thanh, J. Aveyard, D.G. Fernig, Determination of size and concen-
tration of gold nanoparticles from UV–vis spectra, Anal. Chem. 79 (2007)
4215–4221.
[57] M.A. Mougan, A. Coello, A. Jover, F. Meijide, J. Vazquez Tato, Spectrofluorimeters as
light-scattering apparatus: application to polymers molecular weight determina-
tion, J. Chem. Educ. 72 (1995) 284.
[58] Y.L. Hsieh, G.A. Ilevbare, B. Van Eerdenbrugh, K.J. Box, M.V. Sanchez-Felix, L.S.
Taylor, pH-induced precipitation behavior of weakly basic compounds:
determination of extent and duration of supersaturation using potentiometric
titration and correlation to solid state properties, Pharm. Res. 29 (2012)
2738–2753.
[59] J. Wang, E. Matayoshi, Solubility at the molecular level: development of a critical
aggregation concentration (CAC) assay for estimating compound monomer
solubility, Pharm. Res. 29 (2012) 1745–1754.
[60] E.D. Goddard, N.J. Turro, P.L. Kuo, K.P. Ananthapadmanabhan, Fluorescence
probes for critical micelle concentration determination, Langmuir 1 (1985)
352–355.
[61] K. Kalyanasundaram, J.K. Thomas, Environmental effects on vibronic band intensi-
ties in pyrene monomer fluorescence and their application in studies of micellar
systems, J. Am. Chem. Soc. 99 (1977) 2039–2044.
[62] K. Kalyanasundaram, J.K. Thomas, Solvent-dependent fluorescence of pyrene-3-
carboxaldehyde and its applications in the estimation of polarity at micelle-
water interfaces, J. Phys. Chem. 81 (1977) 2176–2180.
[63] P. Chandar, P. Somasundaran, N.J. Turro, Fluorescence probe investigation of
anionic polymer-cationic surfactant interactions, Macromolecules 21 (1988)
950–953.
[64] R. Zana, S. Yiv, C. Strazielle, P. Lianos, Effect of alcohol on the properties of micellar
systems: I. Critical micellization concentration, micelle molecular weight and ion-
ization degree, and solubility of alcohols in micellar solutions, J. Colloid Interface
Sci. 80 (1981) 208–223.
[65] A.P. Demchenko, Y. Mély, G. Duportail, A.S. Klymchenko, Monitoring biophysical
properties of lipid membranes by environment-sensitive fluorescent probes,
Biophys. J. 96 (2009) 3461–3470.
[66] L. Almeida e Sousa, S.M. Reutzel-Edens, G.A. Stephenson, L.S. Taylor, Supersatura-
tion potential of salt, co-crystal, and amorphous forms of a model weak base,
Cryst. Growth Des. 16 (2016) 737–748.
[67] S.R. LaPlante, N. Aubry, G. Bolger, P. Bonneau, R. Carson, R. Coulombe, C. Sturino,
P.L. Beaulieu, Monitoring drug self-aggregation and potential for promiscuity in
off-Target In Vitro pharmacology screens by a practical NMR Strategy, J. Med.
Chem. 56 (2013) 7073–7083.
[68] N.S. Trasi, L.S. Taylor, Thermodynamics of highly supersaturated aqueous solutions
of poorly water-soluble drugs—impact of a second drug on the solution phase
behavior and implications for combination products, J. Pharm. Sci. 104 (2015)
2583–2593.
[69] M.J. Jackson, Dissolution Behavior of Amorphous Solid Dispersions, Purdue
University, 2015.
[70] J. Twist, J. Zatz, Influence of solvents on paraben permeation through idealized skin
model membranes, J. Soc. Cosmet. Chem. 37 (1986) 429–444.
[71] S.H. Yalkowsky, G.L. Flynn, Correlation and prediction of mass transport across
membranes II: influence of vehicle polarity on flux from solutions and suspen-
sions, J. Pharm. Sci. 63 (1974) 1276–1280.
141
[72] N.A. Megrab, A. Williams, B. Barry, Oestradiol permeation through human skin and
silastic membrane: effects of propylene glycol and supersaturation, J. Control.
Release 36 (1995) 277–294.
[73] M. Pellett, S. Castellano, J. Hadgraft, A. Davis, The penetration of supersaturated so-
lutions of piroxicam across silicone membranes and human skin in vitro, J. Control.
Release 46 (1997) 205–214.
[74] K. Moser, K. Kriwet, C. Froehlich, A. Naik, Y.N. Kalia, R.H. Guy, Permeation enhance-
ment of a highly lipophilic drug using supersaturated systems, J. Pharm. Sci. 90
(2001) 607–616.
[75] J.M. Miller, A. Beig, R.A. Carr, J.K. Spence, A. Dahan, A win–win solution in oral
delivery of lipophilic drugs: supersaturation via amorphous solid dispersions in-
creases apparent solubility without sacrifice of intestinal membrane permeability,
Mol. Pharm. 9 (2012) 2009–2016.
[76] J. Brouwers, M.E. Brewster, P. Augustijns, Supersaturating drug delivery systems:
the answer to solubility-limited oral bioavailability? J. Pharm. Sci. 98 (2009)
2549–2572.
[77] S.H. Yalkowsky, Perspective on improving passive human intestinal absorption, J.
Pharm. Sci. 101 (2012) 3047–3050.
[78] T. Higuchi, Physical chemical analysis of percutaneous absorption process from
creams and ointments, J. Soc. Cosmet. Chem. 11 (1960).
[79] S. Raina, G.Z. Zhang, D. Alonzo, J. Wu, D. Zhu, N. Catron, Y. Gao, L. Taylor, Impact of
solubilizing additives on supersaturation and membrane transport of drugs,
Pharm. Res. (2015) 1–15.
[80] S.A. Raina, G.G.Z. Zhang, D.E. Alonzo, J. Wu, D. Zhu, N.D. Catron, Y. Gao, L.S.
Taylor, Enhancements and limits in drug membrane transport using supersat-
urated solutions of poorly Water soluble drugs, J. Pharm. Sci. 103 (2014)
2736–2748.
[81] P.J. Sassene, M.M. Knopp, J.Z. Hesselkilde, V. Koradia, A. Larsen, T. Rades, A.
Müllertz, Precipitation of a poorly soluble model drug during in vitro lipolysis:
characterization and dissolution of the precipitate, J. Pharm. Sci. 99 (2010)
4982–4991.
[82] J. Stappaerts, S. Geboers, J. Snoeys, J. Brouwers, J. Tack, P. Annaert, P. Augustijns,
Rapid conversion of the ester prodrug abiraterone acetate results in intestinal
supersaturation and enhanced absorption of abiraterone: in vitro, rat in situ and
human in vivo studies, Eur. J. Pharm. Biopharm. 90 (2015) 1–7.
[83] S.L. Childs, P. Kandi, S.R. Lingireddy, Formulation of a danazol cocrystal with
controlled supersaturation plays an essential role in improving bioavailability,
Mol. Pharm. 10 (2013) 3112–3127.
[84] F. Poelma, R. Breäs, J. Tukker, D. Crommelin, Intestinal absorption of drugs. The
influence of mixed micelles on the disappearance kinetics of drugs from the
small intestine of the rat, J. Pharm. Pharmacol. 43 (1991) 317–324.
[85] K. Katneni, S.A. Charman, C.J.H. Porter, Permeability assessment of poorly water-
soluble compounds under solubilizing conditions: the reciprocal permeability
approach, J. Pharm. Sci. 95 (2006) 2170–2185.
[86] B.H. Lee, S.D. Christian, E.E. Tucker, J.F. Scamehorn, Solubilization of mono-and
dichlorophenols by hexadecylpyridinium chloride micelles. Effects of substituent
groups, Langmuir 6 (1990) 230–235.
[87] A. Goto, F. Endo, The distribution of alkyl paraben in aqueous sodium lauryl sulfate
micellar solutions: III. The gel filtration of solubilized systems, J. Colloid Interface
Sci. 66 (1978) 26–32.
[88] S.R. Croy, G.S. Kwon, Polysorbate 80 and Cremophor EL micelles deaggregate
and solubilize nystatin at the core–corona interface, J. Pharm. Sci. 94 (2005)
2345–2354.
[89] P.A. Shore, B.B. Brodie, C.A.M. Hogben, The gastric secretion of drugs: a pH partition
hypothesis, J. Pharmacol. Exp. Ther. 119 (1957) 361–369.
[90] T. Xie, L.S. Taylor, Dissolution performance of high drug loading celecoxib
amorphous solid dispersions formulated with polymer combinations, Pharm.
Res. (2015) 1–12.
[91] P.E. Luner, S.R. Babu, S.C. Mehta, Wettability of a hydrophobic drug by surfactant
solutions, Int. J. Pharm. 128 (1996) 29–44.
[92] M. Morgen, C. Bloom, R. Beyerinck, A. Bello, W. Song, K. Wilkinson, R. Steenwyk, S.
Shamblin, Polymeric nanoparticles for increased oral bioavailability and rapid
absorption using celecoxib as a model of a low-solubility, high-permeability
drug, Pharm. Res. 29 (2011) 427–440.
[93] F. Qian, J. Wang, R. Hartley, J. Tao, R. Haddadin, N. Mathias, M. Hussain, Solution
behavior of PVP-VA and HPMC-AS-based amorphous solid dispersions and their
bioavailability implications, Pharm. Res. 29 (2012) 2766–2776.
[94] O.I. Corrigan, Mechanisms of dissolution of fast release solid dispersions, Drug Dev.
Ind. Pharm. 11 (1985) 697–724.
[95] B.C. Hancock, G. Zografi, The relationship between the glass transition temperature
and the water content of amorphous pharmaceutical solids, Pharm. Res. 11 (1994)
471–477.
[96] H. Konno, L.S. Taylor, Ability of different polymers to inhibit the crystallization
of amorphous felodipine in the presence of moisture, Pharm. Res. 25 (2008)
969–978.
[97] A.C.F. Rumondor, M.J. Jackson, L.S. Taylor, Effects of moisture on the growth rate of
felodipine crystals in the presence and absence of polymers, Cryst. Growth Des. 10
(2010) 747–753.
[98] S.A. Raina, D.E. Alonzo, G.G.Z. Zhang, Y. Gao, L.S. Taylor, Impact of polymers on the
crystallization and phase transition kinetics of amorphous nifedipine during
dissolution in aqueous media, Mol. Pharm. 11 (2014) 3565–3576.
[99] T. Xie, L.S. Taylor, Improved release of celexocib from high drug loading amorphous
solid dispersions formulated with polyacrylic acid and cellulose derivatives, Mol.
Pharm. 13 (2016) 873–884.
[100] H.S. Purohit, L.S. Taylor, Phase separation kinetics in amorphous solid dispersions
upon exposure to water, Mol. Pharm. 12 (2015) 1623–1635.
142 L.S. Taylor, G.G.Z. Zhang / Advanced Drug Delivery Reviews 101 (2016) 122–142
[101] A.C.F. Rumondor, L.S. Taylor, Effect of polymer hygroscopicity on the phase
behavior of amorphous solid dispersions in the presence of moisture, Mol.
Pharm. 7 (2010) 477–490.
[102] A.C.F. Rumondor, L.A. Stanford, L.S. Taylor, Effect of polymer type and storage rela-
tive humidity on the kinetics of felodipine crystallization from amorphous solid
dispersions, Pharm. Res. 26 (2009) 2599–2606.
[103] A.C.F. Rumondor, P.J. Marsac, L.A. Stanford, L.S. Taylor, Phase behavior of
poly(vinylpyrrolidone) containing amorphous solid dispersions in the pres-
ence of moisture, Mol. Pharm. 6 (2009) 1492–1505.
[104] A.C.F. Rumondor, H. Wikstrom, B. Van Eerdenbrugh, L.S. Taylor, Understanding the
tendency of amorphous solid dispersions to undergo amorphous–amorphous
phase separation in the presence of absorbed moisture, AAPS PharmSciTech 12
(2011) 1209–1219.
[105] D.E. Alonzo, G.G.Z. Zhang, D.L. Zhou, Y. Gao, L.S. Taylor, Understanding the behavior
of amorphous pharmaceutical systems during dissolution, Pharm. Res. 27 (2010)
608–618.
[106] J. Chen, J.D. Ormes, J.D. Higgins, L.S. Taylor, Impact of surfactants on the crystalliza-
tion of aqueous suspensions of celecoxib amorphous solid dispersion spray dried
particles, Mol. Pharm. 12 (2015) 533–541.
[107] A.S. Indulkar, Y. Gao, S. Raina, G.G.Z. Zhang, L.S. Taylor, Exploiting the phenomenon
of liquid–liquid phase separation to achieve enhanced and sustained membrane
transport of a poorly water soluble drug, Mol. Pharm. (2016) (Submitted).
[108] A.F. Davis, J. Hadgraft, Effect of supersaturation on membrane transport: 1. Hydro-
cortisone acetate, Int. J. Pharm. 76 (1991) 1–8.
[109] D.M. Mudie, K. Murray, C.L. Hoad, S.E. Pritchard, M.C. Garnett, G.L. Amidon, P.A.
Gowland, R.C. Spiller, G.E. Amidon, L. Marciani, Mol. Pharm. 11 (2014) 3039.