Development of a Biorelevant, Material-Sparing Membrane Flux Test

Apr 25, 2017 - To overcome these limitations, we have designed a practical in vitro membrane flux test, that requires minimal active pharmaceutical in...
1 downloads 10 Views 4MB Size
This is an open access article published under an ACS AuthorChoice License, which permits copying and redistribution of the article or any adaptations for non-commercial purposes.

Article pubs.acs.org/molecularpharmaceutics

Development of a Biorelevant, Material-Sparing Membrane Flux Test for Rapid Screening of Bioavailability-Enhancing Drug Product Formulations Aaron M. Stewart, Michael E. Grass,* Deanna M. Mudie, Michael M. Morgen, Dwayne T. Friesen, and David T. Vodak Global Research and Development, Pharmaceutical Science, Capsugel, Bend, Oregon 97701, United States S Supporting Information *

ABSTRACT: Bioavailability-enhancing formulations are often used to overcome challenges of poor gastrointestinal solubility for drug substances developed for oral administration. Conventional in vitro dissolution tests often do not properly compare such formulations due to the many different drug species that may exist in solution. To overcome these limitations, we have designed a practical in vitro membrane flux test, that requires minimal active pharmaceutical ingredient (API) and is capable of rapidly screening many drug product intermediates. This test can be used to quickly compare performance of bioavailability-enhancing formulations with fundamental knowledge of the rate-limiting step(s) to membrane flux. Using this system, we demonstrate that the flux of amorphous itraconazole (logD = 5.7) is limited by aqueous boundary layer (ABL) diffusion and can be increased by adding drug-solubilizing micelles or drug-rich colloids. Conversely, the flux of crystalline ketoconazole at pH 5 (logD = 2.2) is membrane-limited, and adding solubilizing micelles does not increase flux. Under certain circumstances, the flux of ketoconazole may also be limited by dissolution rate. These cases highlight how a well-designed in vitro assay can provide critical insight for oral formulation development. Knowing whether flux is limited by membrane diffusion, ABL diffusion, or dissolution rate can help drive formulation development decisions. It may also be useful in predicting in vivo performance, dose linearity, food effects, and regional-dependent flux along the length of the gastrointestinal tract. KEYWORDS: flux, dissolution rate, drug/polymer colloids, unstirred water layer, itraconazole, ketoconazole, spray-dried dispersion, diffusion, membrane



INTRODUCTION

and to the market. In vitro dissolution tests are designed to compare the relative rates of dissolution and extent of active pharmaceutical ingredient (API) concentration achieved using different formulations. The results of such tests, however, depend on several factors, including (a) apparatus type, (b) dose-to-volume ratio, (c) type of agitation and (d) dissolution medium/media composition(s), and (e) how the media is sampled. Depending on the dissolution methodology employed, the physicochemical properties of the API, and the

Many current small molecule drug candidates have very low solubility. By some estimates, as many as 90% of current small molecule drugs in the development pipeline are BCS class II or IV (low solubility).1 Many such compounds require modifications for bioavailability enhancement, such as salt forms,2−4 cocrystals,5,6 particle size reduction,7 inclusion complexes,8−10 lipid formulations,11,12 and amorphous solid dispersions.13−15 These formulation approaches are developed in order to increase dissolution rate, solubility, and/or increase “carriers” such as micelles, vesicles, complexes, or colloids in the intestine. An understanding of the mechanism of bioavailability enhancement of these drug products is crucial for selecting and optimizing formulations to carry forward to clinical trials © XXXX American Chemical Society

Received: Revised: Accepted: Published: A

February April 20, April 25, April 25,

16, 2017 2017 2017 2017 DOI: 10.1021/acs.molpharmaceut.7b00121 Mol. Pharmaceutics XXXX, XXX, XXX−XXX

Article

Molecular Pharmaceutics

Figure 1. Vertical membrane flux cell integrated into the μDiss Profiler.

across membranes have been described in the literature to assess crystalline API as well as bioavailability-enhancing formulations.17,21−23 An advantage of membrane-based assays is that the concentration is determined indirectly by measuring the flux across a membrane that is impermeable to micelles, colloids, and other bound drug species. The type of membrane used is typically a lipid-filled hydrophobic membrane similar to that used for PAMPA assays,22,24 a hydrophilic regenerated cellulose membrane such as those used in dialysis,21,25,26 or cellbased membranes like Caco-2 or MDCK cell monolayers.17,27,28 For PAMPA-like membranes and cell monolayers, there is typically a large partitioning of drug from the donor compartment into the membrane for lipophilic API. For membrane flux tests, either the aqueous boundary layer (ABL) adjacent to the membrane surface or membrane permeability may be rate-limiting depending on the solubility and distribution coefficient of the API, as well as the presence of other rapidly diffusing drug species such as micelles or inclusion complexes. In both ABL and membrane-limited cases, dissolution rate can be limiting as well. We have developed a practical and easy to use membrane flux assay utilizing a custom designed flux vessel with a lipidfilled membrane and a large surface area to volume ratio similar to that in the upper small intestine.18,29−31 This design allows rapid assessment of drug product intermediates using minimal API, while assessing the relative importance of dissolution rate, ABL diffusion, and membrane diffusion on drug flux. This flux test vessel is built on the Pion μDiss Profiler22,24 system, but differs from the commercially available flux vessel (μFlux) in orientation, geometry, and surface area to volume ratio. We demonstrate the utility of this in vitro membrane test by showing the role of additional apparent concentration (via solubilizing micelles or colloids) on the flux of amorphous itraconazole and crystalline ketoconazole. These cases provide examples for ABL-limited flux, membrane-limited flux, and dissolution rate-limited flux. These systems demonstrate how a formulation scientist may use flux data in the context of determining the potential ratelimiting step for absorption of a compound early in a development program, and how we use these data to guide formulation development. In all cases, the data can be explained by using a steady-state dissolution and diffusion model.

mechanism of solubility enhancement, choosing the formulation providing the fastest rate of release and highest total drug concentration in an in vitro test may or may not lead to selection of the formulation providing optimum in vivo performance. Successful formulation selection and in vitro comparison of bioavailability-enhancing formulations requires an understanding of the rate-limiting steps to absorption for an API, the mechanism of solubility enhancement of the formulations, and the basic physics and assumptions of the in vitro assays being used for assessment. For example, while cyclodextrin formulations increase the apparent solubility assessed in vitro, they do not increase the activity of the drug.16 These complexes can act as rapidly dissociating sources of unbound drug but can also lower the available concentration of unbound drug depending on the dose of the drug and concentration of cyclodextrin. Regardless of the type of formulation, if absorption of a drug is limited by permeability across the epithelial membrane, the critical attribute to measure in vitro is the concentration of unbound drug in solution.17,18 However, assessing the concentration of unbound drug in solution can be difficult in the presence of several different drug species in the medium. In many dissolution tests, the concentration is determined by direct UV absorbance or via separation using a filter or centrifugation followed by sampling of the filtrate or supernatant. The measured concentration may therefore include all, or some subset of, unbound drug, micelle-bound drug, inclusion complexes, and/or drug in colloids from a few nanometers up to a few microns in size. In some cases, however, bound drug species are critical for in vivo performance in order to provide sufficient dissolution rate or to contribute to diffusion across the unstirred water layer (UWL) of the small intestine.19 For example, the best performing amorphous solid dispersions (ASDs) of very low solubility compounds form small drug/polymer colloids that are a rapidly dissolving, highly active source of unbound drug in the lumen and UWL of the upper intestine.20 In these cases it is critical to compare the apparent drug concentration (i.e., unbound, micelle-bound, and drug/polymer colloids) reached for each potential formulation. Employing a membrane flux test in conjunction with in vitro dissolution is a means of overcoming challenges in assessing and rank-ordering formulations with different mechanisms of solubility enhancement. Several in vitro assays based on flux B

DOI: 10.1021/acs.molpharmaceut.7b00121 Mol. Pharmaceutics XXXX, XXX, XXX−XXX

Article

Molecular Pharmaceutics



MATERIALS AND METHODS Flux Apparatus. The vertical membrane flux cell consists of a donor compartment and a receiver compartment, separated by an Accurel PP 1E (55% porous, 100 μm thickness) polypropylene membrane (3M, Maplewood, MN) (Figure 1).32 The membrane is impregnated with 50 μL of Pion GIT-0 lipid solution consisting of 20% w/w phospholipid dissolved into dodecane (Pion Inc., Billerica, MA) and attached to the receiver vessel. Both the donor and receiver compartments are agitated by magnetic stirring. The receiver compartment contains a plastic spacer and grating to elevate the stir bar above the membrane. Samples are introduced to the donor vessel by preweighing directly into the donor vessel and subsequently adding dissolution medium. Once the dissolution medium has been added to the donor vessel, the receiver vessel is inserted into the donor vessel and suspended vertically 5 mm above the donor compartment by a plastic sleeve. For the experiments herein, the donor vessel contained 5 mL of dissolution medium, buffered at pH 5, 6.5, or 8 as indicated. The receiver vessel contained 10 mL of buffer at the same pH as the donor solution with 2% sodium lauryl sulfate (w/w). The surface area of the membrane is 4.90 cm2. The temperature for all experiments was maintained at 37 °C by circulating water through a heating block mounted to a μDiss Profiler (Pion Inc.). UV probes (10 mm path length) connected to a Rainbow UV spectrometer (Pion Inc.) system were used to determine the apparent drug concentration in the receiver vessels. Samples of the donor compartment were removed with a disposable pipet for centrifugation followed by HPLC and DLS analysis of the supernatant. Advantages of Modified Membrane Flux Assay. The vertical membrane flux apparatus described in Figure 1 has several key attributes: (a) low volume to minimize API required (b) high surface area to volume ( A ) similar to the small

Table 1. Comparison of the Vertical Membrane Flux Cell Used in This Study to Commercially Available in Vitro Flux Assays vertical membrane flux cell

property

PAMPA

Pion μFlux

geometry

microtiter plate “sandwich”

side-by-side horizontal

volume membrane surface area A/V ABL thickness A/(VhABL)

350 μL 0.32 cm2

20 mL 1.3 cm2

0.87 cm−1a 2000−4000 μm33

0.98 cm−1 ∼24 μmb

up to 15 h

0.07 cm−1 not measured not measured up to 24 h

96

4

8

typical experiment duration sample capacity

0.22−0.44 cm−2

vertical with receiver vessel inside donor vessel 5 mL 4.9 cm2

408 cm−2 1−3 h

a Calculated based on standard 96-well plate dimensions. bCalculated at 100 rpm for amorphous itraconazole.

schematically shown in Figure 2. By only considering the initial flux of drug across the membrane (i.e., when the concentration

V

intestine as a smooth tube (c) robust mixing in the donor and receiver vessels (d) rapid flux and high sample capacity for quick screening and reduced experiment duration (e) ability to manually sample the donor solution for orthogonal analytical characterization (f) ease of use (g) minimal use of organic solvents (h) compatible with commercial multichannel UV−vis detection (Pion Rainbow and μDiss Profiler) A comparison of the vertical membrane flux cell to commercially available systems is highlighted in Table 1, with the key differences being an increased surface area to volume ratio ( A ), a decreased aqueous boundary layer thickness (hABL), V and increased throughput due to a larger sample capacity and decreased experimental duration. Application of a Steady-State Model to Membrane Flux in the Presence of Micelle-Bound Drug and/or Drug-Rich Colloids. Physical Model Description. There are four resistances to the diffusion of drug from solid particles in the bulk of the donor compartment to being dissolved in the bulk of the receiver compartment. Drug must first (1) dissolve and then (2) diffuse across the aqueous boundary layer (ABL) of the donor compartment, then (3) diffuse through the lipidfilled porous membrane, and finally (4) diffuse across the ABL of the receiver compartment. The last three resistances are

Figure 2. Diagram of the steady-state model with key parameters depicting ABL-limited flux.

in the receiver phase is much less than the solubility in the receiver phase) and ensuring that the solubility of the drug in the receiver phase is much greater than in the donor phase, the resistance across the receiver ABL is minimal and can be ignored. The flux of dissolved drug from the donor to the acceptor compartment can be limited by diffusion across the ABL, the membrane, or a combination of the two. Additionally, the flux of drug from a solid drug (or drug formulation) particle may be limited by dissolution rate. In many cases, the dissolution rate is determined by diffusion of drug molecules at the particle surface (PBL). The flux across each diffusional resistance is a product of the permeability coefficient and the concentration gradients. Equations 1−3 are the effective permeability coefficients for the case of ABL-limited flux, membrane-limited flux, and dissolution rate-limited flux, respectively: Deff PABL = hABL (1) C

DOI: 10.1021/acs.molpharmaceut.7b00121 Mol. Pharmaceutics XXXX, XXX, XXX−XXX

Article

Molecular Pharmaceutics

Figure 3. Description of drug species and notation used in the steady-state diffusion model.

PMKM / D =

PPBL =

DeffM hm

KM / D

8. Only neutral unbound drug (c0) partitions into the lipidfilled pores of the membrane. 9. KM/D is unaffected by the presence of micelles but is pH dependent for ionizable molecules. 10. The concentration of unbound drug in the receiver phase is much less than the solubility in the receiver phase (initial flux condition). 11. The concentration of unbound drug in the receiver phase is much less than the concentration of unbound drug in the donor phase (initial flux condition). 12. Micelle-bound drug is in equilibrium with unbound drug. 13. Colloidal drug is in equilibrium with unbound drug. 14. The membrane composition is constant over the duration of the measurement (no degradation of the membrane). Drug Species Description and Notation. In the pharmaceutical literature, and particularly with respect to solubility enhancing formulations, the terms “concentration” and “solubility” are often used to refer to any number of different drug species (or phases) depending on the assay used. In the following, very specific terms are used to refer to the concentration of drug in different species. Figure 3 summarizes the various drug species considered and the notation used herein. The concentration of unbound drug (cu) plus micellebound drug (cm) is determined by an assay of the supernatant after ultracentrifugation (ca. 300,000g, ca. 8 min.) and is therefore given the notation cu,m = cu + cm. To avoid additional notation, the term cu,m is used for both the theoretical concentration and the experimentally determined value of cu + cm. In a similar manner, the concentration of rapidly diffusing drug species, cu + cm + drug in colloids (cc), is determined by an assay of the supernatant after microcentrifugation (ca. 15,800g, ca. 1 min.) and is given the notation cu,m,c = cu + cm + cc. This is also referred to as the “apparent concentration” because it is the concentration most commonly measured in routine dissolution tests. It is assumed that the micelle-bound drug and drug in colloids disproportionate to form unbound drug on a time scale much faster than drug transport across the ABL. (i.e., dissolution of drug colloids is considered to be infinitely fast). All drug concentrations referred to in this article are those in the bulk medium (i.e., directly measurable concentrations)

(2)

Deff hPBL

(3)

In this work, KM/D is defined as the distribution coefficient of unbound drug between the membrane and the donor compartment (cu = c0 + c±). As a result, KM/D is pH dependent and decreases as the drug becomes increasingly ionized. The derivation of these equations has been previously described34 and applied to in vitro membrane flux.33,35 The initial steadystate flux of drug from the donor to the receiver compartment is solved by using a steady-state model where the flux across the donor ABL, the membrane, and the receiver ABL are all equal, and the flux across each layer is solved based on Fick’s law of diffusion. A full derivation is provided in the Supporting Information with the key results presented here with simplified notation. The key model assumptions are as follows: 1. Drug transport in the ABL and the membrane is described by Fick’s second law. 2. The diffusion coefficient in each medium is not concentration dependent and aqueous diffusion coefficients of ionized and unionized drug are equal. 3. Drug transport via convection within the lipid-filled membrane is minimal and can be neglected. 4. The ABL thickness is an artificial construct representing the resistance to diffusion of drug from the bulk to the surface of the membrane. 5. Net flux of drug occurs in one direction across each barrier in series from the well-mixed, bulk aqueous donor medium to the receiver medium. 6. The instantaneous concentration profile within each diffusion layer resembles a steady-state (pseudosteadystate approximation). 7. The thickness of each diffusion layer is constant with time and across samples (this is not strictly true for the ABL thickness). D

DOI: 10.1021/acs.molpharmaceut.7b00121 Mol. Pharmaceutics XXXX, XXX, XXX−XXX

Article

Molecular Pharmaceutics

the bulk concentration of drug in the donor compartment will be less than the solubility, and the flux will depend on the dissolution rate of the solids in the donor compartment. In cases when dissolution rate is much less than flux across the membrane at saturation, the steady-state concentration of unbound drug can be solved using a pseudosteady-state approximation where the rate of drug dissolving is equal to the measured flux:

unless otherwise specified. Additional notation is introduced in the Supporting Information relating to concentrations in the membrane and at the membrane surface. Aqueous Boundary Layer (ABL) Limited Flux (PABL ≪ PMKM/D and PPBL). When dissolution rate of drug in the bulk donor solution is sufficiently fast and the partitioning of drug from the donor compartment into the lipid membrane (KM/D) is very high, the driving force for diffusion of drug into the bulk medium and through the membrane is much greater than that across the ABL. In this case, the highest resistance to flux is across the ABL. For an ABL-limited system, the initial steady-state flux is given by j = PABLcu , m , c =

Deff hABL

cu , m , c

V j = kdiss (csat − cu) (8) A In the following sections, these models are used to assess the rate-limiting step for flux in several different systems. Materials. Itraconazole (>98% purity), ketoconazole (>98% purity), and sodium lauryl sulfate were purchased from Spectrum Chemical MFG Corp (New Brunswick, New Jersey). Sporanox was purchased from Drug World Pharmacies (New City, NY). The structures of itraconazole and ketoconazole are shown in Figure 4, and physicochemical properties and

(4)

In this work, the thickness of the ABL is assumed to be a constant, hABL = 24 μm, determined from measuring the flux of Sporanox at a known concentration (see Supporting Information). The effective diffusion coefficient, Def f, is the weighted average of the diffusion coefficients of unbound drug, micelle-bound drug, and drug in colloids: Deff = Du fu + Dm fm + Dc fc

(5)

where Di and f i are the diffusion coefficient and fraction of drug for each species discussed above. In the following sections, Du is calculated from ADMET modeling software (Hayduk−Laudie equation),36 while Dm and Dc are determined using the measured size by DLS and the Stokes−Einstein equation (the size distribution is not considered). Membrane-Limited Flux (PMKM/D ≪ PABL and PPBL). Membrane-limited flux studies have been extensively performed and described in pharmaceutical research.22,21,17,37 Under this condition, flux is proportional to the concentration of neutral unbound drug (c0) to a first approximation. Any increase in apparent concentration at constant cu will not increase flux. Because PMKM/D ≪ PABL and PPBL, the concentration of unbound drug in the donor medium at the membrane surface (cD@M) is only slightly less than the concentration in the bulk. Making the approximation that cD@M = cu (bulk) and that the concentration in the receiver compartment is negligible, the flux can be expressed as j = PMKM / Dcu =

DeffM hM

KM / Dcu

Figure 4. Chemical structures of itraconazole and ketoconazole.

solubility values are shown in Table 2. Hydroxypropylmethyl cellulose acetate succinate (Affinisol) L-HP (ca. 7 wt % acetyl, 16 wt % succinoyl; low viscosity grade) and H-HP (ca. 12 wt % acetyl, 6 wt % succinoyl; low viscosity grade) were provided by DOW Chemical Company (Midland, MI). Buffer components (KH2PO4, Na2HPO4, NaCl, NH4OAc) were purchased from Sigma-Aldrich Co. LLC (St. Louis, MO). Gastrointestinal tract (GIT-0) lipid was purchased from Pion Inc. (Billerica, MA). Itraconazole Spray Dried Dispersion Manufacture and Characterization. Two spray dried dispersions of itraconazole were prepared to compare to Sporanox. Spray solutions were prepared by dissolving itraconazole and polymer (either HPMCAS-L HP or HPMCAS-H HP) at a 1:3 wt ratio in a 95:5 mixture of tetrahydrofuran and water (13% total solids by weight). Solutions were spray dried on a customized spray dryer using a Steinen A75 spray nozzle (GEA, Düsseldorf, Germany). Inlet temperatures were within a range of 105−116 °C, outlet temperature was constant at 45 °C, solution flow rate was 192−194 mL min−1, drying gas flow rate was 1836−1854 mL min−1, and atomization pressure was 643−659 psi. Batch sizes were 400 g with a wet yield of 77−85%. The particle size, reported as D10, D50, and D90, was 7, 25, 56 μm and 6, 25, 55 μm for the HPMCAS-L HP and HPMCAS-H HP SDDs, respectively. Residual solvent was removed by tray drying at 40 °C/15%RH for 8 h. Both SDDs were confirmed to be amorphous via X-ray diffraction and each exhibited a single glass transition temperature of 95 ± 1 °C via modulated differential scanning calorimetry (data not shown). Membrane Flux Experimental Studies. The solubility (unbound plus micelle-bound drug) and experimental conditions of five experimental flux studies are described in Table 3. Two different compounds were selected for this investigation

(6)

As discussed above, as ionization increases due to pH change in the presence of excess solids, cu increases while KM/D decreases. The unbound drug concentration is not typically measured directly, but can be readily determined. In practice, it is determined from the measured concentration of unbound plus micelle-bound drug, cu,m, and the micelle partition coefficient (measured separately):34 cu =

1 cu , m 1 + VmK m

(7)

where Vm is the volume fraction of micelles in the donor compartment and Km is the partition coefficient between micelles and the donor buffer. Dissolution Rate-Limited Flux (PPBL ≪ PABL and PMKM/D). If the dissolution rate is slower than or comparable to the flux measured when the donor compartment is saturated with drug, E

DOI: 10.1021/acs.molpharmaceut.7b00121 Mol. Pharmaceutics XXXX, XXX, XXX−XXX

Article

Molecular Pharmaceutics Table 2. Physicochemical Properties and Solubility of Itraconazole and Ketoconazole solubility, cu,m = cu + cm (μg mL‑1)

a

drug

molecular weight (g/mol)

basic pKa

clogD

pH

0 or 0.025% SIF

0.5% SIF

1.0% SIF

2.0% SIF

receiver mediumc

itraconazole (amorphous) ketoconazole (crystalline)

705.6 531.4

3.738 2.9,6.539

5.738 3.7a 3.5a 2.2a

6.5 8 6.5 5

0.1 4.5 8 79

6 38 ND 310

NDb 94 ND 680

20 210 ND 1160

1060 5900 >6000 6800

ACD Laboratories. bND = not determined. c2% SLS (w/w) at specified pH.

Table 3. Formulation, Dose Information, and Experimental Conditions for Flux Studies dose concentration (cdose) (μg mL−1)

study

drug

formulation

pH

summary of experimental conditions

1 2 3a 3b 4 5

itraconazole itraconazole ketoconazole

Sporanox amorphous dispersions crystalline

1000 1000 5000

6.5 6.5 5

ketoconazole ketoconazole

crystalline crystalline

5000 100−5000

8 6.5

0, 0.5, and 2.0% SIF, in duplicate Sporanox and two itraconazole ASDs, no SIF micelles, in duplicate 0.025, 0.5, 1.0, and 2.0% SIF with excess solids, in duplicate 0.0, 0.224, 0.5, and 2.0% SIF without excess solids, in duplicate 0.025, 0.5, 1.0, and 2.0% SIF with excess solids, in duplicate six dose concentrations, no SIF, single measurement

amorphous itraconazole and hydroxypropyl methylcellulose (HPMC) on inert sugar cores. Sporanox beads were dosed into 67 mM phosphate buffered saline (PBS) at pH 6.5 with three concentrations of SIF powder (biorelevant.com, London, U.K.): no SIF, 0.5% SIF, or 2.0% SIF (w/w). The total dose concentration in the donor medium added initially at the onset of the experiment, cdose = 1000 μg mL−1, was constant for all samples. Of critical importance to this work, Sporanox dissolves rapidly to the amorphous solubility of itraconazole (dissolution rate of drug in the bulk donor solution is not limiting) so that cu = csat throughout the experiment and should not be affected by the presence of micelles at saturation since the chemical potential is constant (cu is constant). The concentration of unbound drug plus micelle-bound drug, however, increases with increasing micelle concentration. Also of importance is that Sporanox does not form any drug-rich colloids (cc = 0), such that cu,m,c = cu,m (data not shown). The flux of drug appearing in the receiver phase was determined from the slope of the receiver concentration vs time plot at steady-state (30− 60 min) along with the volume of the receiver solution and the area of the membrane. Samples were analyzed in duplicate. Study 2, ABL-Limited Flux: Flux Increases with Increasing Colloidal Drug Concentration of Itraconazole from Amorphous Formulations. In this study, the flux of itraconazole from three formulations was compared using 67 mM PBS at pH 6.5 (no micelles) as the donor medium. The formulations were Sporanox (described above), 25% itraconazole/75% HPMCAS-H HP SDD, and 25% itraconazole/75% HPMCAS-L HP SDD. For all three formulations, cu ≈ 0.1 μg mL−1, but dissolution of the HPMCAS-based SDDs leads to formation of drug-rich colloids that are approximately 200 nm in diameter (determined by DLS, see Supporting Information) that rapidly dissolve to source unbound drug13(dissolution rate of drug in the bulk donor solution is not limiting). The apparent concentration of unbound drug and colloidal drug (cu,m,c) was determined by centrifuging an aliquot from the donor vessel for 1 min at 15,800g and analyzing the supernatant by HPLC.13 The flux of drug was determined from 30 to 60 min, and samples were analyzed in duplicate. Study 3a, Membrane-Limited Flux: Flux Remains Constant with Increasing Micelle-Bound Concentration of Ketoconazole with Excess Solids (pH 5). Crystalline

in order to demonstrate different rate-limiting steps in the membrane flux test. The flux of amorphous itraconazole is expected to be ABL-limited due to very high extent of partitioning into the lipid-filled membrane (logP = 5.7, >98% neutral at pH 6.5). Crystalline ketoconazole, however, is less lipophilic (logP = 3.7) and the logD decreases as the pH decreases below pH 8 (highest basic pKa = 6.5). Thus, it is expected to be membrane-limited at pH 5 (>95% ionized), and possibly ABL-limited at pH 8 (95% ionized40 and clogD = 2.2, so that the unbound drug concentration is much higher than the neutral unbound drug concentration, cu ≈ 18c0. Flux was determined from 30 to 60 min as a function of SIF concentration in the range of 0.025 to 2.0% SIF, and samples were analyzed in duplicate. Study 3b, Membrane-Limited Flux: Flux Decreases with Increasing Micelle-Bound Concentration of Ketoconzaole below Saturation (cu,m Held Constant, pH 5). To further illustrate the effect of micelle-bound drug on flux for a membrane-limited case, the flux of ketoconazole at pH 5 was investigated at a constant cu,m (unbound + micelle-bound drug) equal to the saturated solubility of unbound drug, csat = 79 μg mL−1. To perform this experiment, ketoconazole was saturated in 10 mM ammonium acetate at pH 5 and excess solids were removed via syringe filtration with a 0.45 μm PTFE filter. Subsequently, 0.224, 0.5, and 2.0% SIF micelles were added to sequester varying fractions of unbound drug, thereby decreasing the unbound drug concentration with an increase in the micelle concentration. The flux of drug was determined from 30 to 60 min, and samples were analyzed in duplicate. Study 4, Combined ABL and Membrane Resistance: Ketoconazole (pH 8). Ketoconazole was tested at a high dose concentration relative to its saturated solubility (high excess surface area from undissolved solids) and well above its pKa (5000 μg mL−1 in 67 mM PBS at pH 8) in order to ensure that dissolution rate of drug in the bulk donor solution is not limiting and that the majority of ketoconazole is neutral. At pH 8, ketoconazole is