Critical Excipient Properties for the Dissolution Enhancement of Phenytoin.

Solubility-enhancing amorphous solid dispersions can aid in the oral delivery of hydrophobic, poorly soluble drugs. Effective solid dispersion excipients enable high supersaturation drug concentrations over biologically relevant time scales. The critical characteristics of an excipient that allow it to work well in a solid dispersion system are not well understood. We prepared poly(N-isopropylacrylamide), poly(N,N-dimethylacrylamide), and poly(N-hydroxyethylacrylamide) excipients of varying molar mass and examined their ability to improve the aqueous solubility of phenytoin, a Biopharmaceutical Class System Class II drug. Binary and ternary solid dispersions of phenytoin and these excipients, along with hydroxypropyl methylcellulose acetate succinate and hydroxypropyl methylcellulose, were prepared at 10 wt % drug loading. Dissolution behavior was studied at early time points (<1 min) and over the course of 6 h. Performance of the ternary solid dispersions was largely a function of the concentration of poly(N-isopropylacrylamide) present in micellar structures and the concentration of PNiPAm micelles in the dissolution media. We present several systems that achieved significant improvement of phenytoin solubility over a wide composition range at enhancement factors among the highest seen to date for phenytoin.

Introduction

Low aqueous solubility of emerging drugs poses an immense challenge to the pharmaceutical industry. A pharmaceutical active’s bioavailability is often captured in two broad-stroke properties: permeability and solubility.[1,2] The Food and Drug Administration and the pharmaceutical industry as a whole classify most drugs by these two characteristics in a four-quadrant Biopharmaceutical Classification System (BCS).[1] Currently, ‘ideal’ drugs that exhibit high solubility and high permeability (Class I) make up roughly 35% of currently marketed drugs; however, only 10% of future drug candidates in the pipeline are in Class I.[3] In contrast, the vast majority (∼80%) of future drug candidates is expected to have low aqueous solubility (Classes II and IV),[3,4] and this has driven extensive efforts toward formulation strategies for improving their aqueous solubility.[5]

Formulation of poorly soluble drugs into amorphous solid dispersions has been identified as a general solubility-enhancement method that does not compromise the intestinal permeability of the pharmaceutical.[6,7] Dispersing the active pharmaceutical within a polymeric solid dispersion excipient provides several advantages for increasing drug solubility.[8,9] Polymer excipients are often designed to be hydrophilic, and as such, enhance the aqueous dispersibility of the poorly soluble drug.[10] Furthermore, by introducing the amorphous form of the drug into aqueous media upon dissolution, solid dispersions deliver a higher-energy form of the drug with a higher solubility than the crystalline solubility of the drug.[11−13]

Unfortunately, the higher free energy of the amorphous compound relative to its crystalline counterpart thermodynamically favors the crystallization of the drug. As such, the choice of a suitable excipient to stabilize the drug in both the solid and dissolved states is critical to the success of this solubilizing method.[14] While suitable excipient-drug pairs can be discovered through screening processes, our aim is to expand the foundational understanding of effective pharmaceutical excipients through systematic variations of key molecular and physical properties of the excipient.

As model drug-excipient pairs, this work focuses on phenytoin, a BCS Class II (low solubility, high permeability) antiepileptic, and poly(acrylamide)-based polymers. Previous work has shown several highly effective poly(acrylamide)-based excipients for achieving and sustaining high supersaturation levels of phenytoin (up to a 22-fold increase over crystalline solubility).[15−19] Specifically, low-molar-mass, micelle-forming poly(N-isopropylacrylamide) (PNiPAm) homopolymer has performed well, both on its own[18] and when blended with hydroxypropyl methylcellulose acetate succinate (HPMCAS).[17] Poly(N,N-dimethylacrylamide-co-N-isopropylacrylamide) (PDMAm-co-PNiPAm) containing ∼70 mol % N-isopropylacrylamide has also performed well, both as free chains and as in the corona of self-assembled micelles.[16,19]

This work probes the transferability of the conclusions of previous PNiPAm-phenytoin systems. Specifically, this work expanded on the understanding afforded by the study of low-molar-mass (<10 kDa) PNiPAm homopolymers that formed dispersed micelles in the aqueous solution.[18] The current understanding and explanation for the effective solubilization behavior is that dissolved phenytoin is stabilized within local environments of high NiPAm content.[17−19] In the case of low-molar-mass reversible addition–fragmentation chain-transfer (RAFT)-synthesized homopolymers, micellization of the PNiPAm homopolymer (NiPAm corona, alkyl chain hydrophobic core) increases the local PNiPAm concentration within the corona environment as compared to free chains.[17−19]

Stabilization of phenytoin in PNiPAm-rich environments may occur because PNiPAm offers a suitable hydrophilic/hydrophobic balance that encourages phenytoin partitioning in the corona.[19] PNiPAm may also offer favorable hydrophobic interactions between PNiPAm isopropyl groups and phenytoin aromatic rings[17,19] and/or specific secondary amide-secondary amide interactions that disrupt the fastest hydrogen-bond ribbon growth axis of phenytoin.[20] Indeed, Moghadam and Larson developed all-atom molecular dynamics simulations of NiPAm-based polymers in water in the presence of phenytoin.[21] They argue that an advantageous balance of hydrophilic/hydrophobic character can be accessed both in low-molar-mass PNiPAm homopolymers and in PNiPAm co-polymers with other hydrophilic co-monomers, and that an ideal hydrophilic/hydrophobic balance can maximize stabilizing excipient-phenytoin contacts.

We previously evaluated PNiPAm homopolymers of varying molar mass and end group for supersaturation achievement and maintenance of phenytoin.[18] By keeping many of the polymeric molecular details the same, i.e., the controlled synthesis by reversible addition–fragmentation chain-transfer polymerization, the chain-transfer agent (CTA), and the molar mass range, in this work, we test the generalizability of previous conclusions. These findings probed the role of the NiPAm repeat unit specifically. By investigating binary and ternary solid dispersions of phenytoin formulated with two structurally similar homopolymers and two cellulosic polymers, we have increased our understanding relating to the importance of NiPAm in the enhancement of phenytoin’s solubility in aqueous media.

Experimental Section

Materials

All chemicals were used as received from the manufacturer: N-isopropylacrylamide (Sigma, >99%), N-hydroxyethylacrylamide (Sigma, 97%), N,N-dimethylacrylamide (Sigma, 99%), 4,4′-azobis(4-cyanovaleric acid) (Sigma, >98%), 4-cyano-4-(dodecylsulfanylthiocarbonyl)sulfanylpentanoic acid (Strem Chemical, >97%), phenytoin (Sigma, >99%), dioxane (Sigma-Aldrich), simulated intestinal fluid (SIF) powder (Biorelevant), and dimethyl sulfoxide-d6 (DMSO-d6, Cambridge Isotope Laboratories, 99.9% D). The cellulosic HPMCAS and hydroxypropyl methylcellulose (HPMC) samples used were AFFINISOL HPMCAS 912 G and METHOCEL E3 LV, respectively, and were provided by the Dow Chemical Company. Cellulosic materials are often characterized by the average number of substituent groups per glucose repeat unit, defined as the functional group degree of substitution. The HPMCAS sample had methoxy, hydroxypropyl, acetate, and succinate degrees of substitution of 1.94, 0.25, 0.57, and 0.28, respectively. This HPMCAS grade was chosen because it was independently determined by the Dow Chemical Company to outperform all other commercial HPMCAS grades as a solid dispersion excipient for phenytoin at a loading of 10 wt % over 6 h. The HPMC sample had methoxy and hydroxypropyl degrees of substitution of 1.91 and 0.25, respectively, and was chosen for its low viscosity in water when dissolved and its intermediate hydrophobic/hydrophilic balance. Substitution data for HPMC and HPMCAS was provided by The Dow Chemical Company.

Synthesis of PNiPAm, PDMAm, and Poly(N-hydroxyethylacrylamide) (PHEAm)

The syntheses of PNiPAm, PDMAm, and PHEAm were all performed via similar methods. A sample polymerization is given here: NiPAm (1.275 g, 0.01127 mol), 4-cyano-4-[(octadecylsulfanylthiocarbonyl) sulfanyl]pentanoic acid (0.500 g, 0.00102 mol), and 4,4′-azobis(4-cyanopentanoic acid) (0.0286 g, 0.000102 mol) were added to a round-bottom flask. Dioxane (25 mL) was added. The reaction mixture was sparged with nitrogen gas for 45 min and then placed in an oil bath at 72 °C. After 5 h, the reaction was exposed to air. The resultant polymer was isolated and purified by precipitation (2 times) in cooled acetone, methanol, ethanol, or a mixture thereof. The polymer was dried under vacuum at 40 °C overnight. Solubility of the polymers in phosphate-buffered saline (PBS, 82 mM sodium chloride, 20 mM sodium phosphate dibasic, 47 mM potassium phosphate monobasic, pH 6.5) was tested by visually examining solutions at 9 mg mL–1.

Molecular Characterization

NMR spectroscopy was performed on a 500 MHz Varian INOVA-500 spectrometer in DMSO-d6 with a d1 relaxation time of 10 s.

Size exclusion chromatography (SEC) was used to determine molar masses and molar mass distributions, using LS detectors. Full traces (Figure S1) were plotted from refractive index signals, relative to polystyrene standards. PNiPAm and PDMAm samples were analyzed using THF as the mobile phase, PHEAm samples were analyzed using dimethylformamide as the mobile phase, and HPMC was analyzed using a mobile phase of 0.1 M Na2SO4 aqueous solution supplemented with 1% acetic acid. Molar mass data for HPMCAS were provided by The Dow Chemical Company.

Thermal Analysis

Differential scanning calorimetry (DSC) was performed on a Discovery DSC (TA Instruments, New Castle, DE) equipped with a refrigerated cooling system. A dry N2 purge flowed through the cell at 50 mL min–1, and samples were run in hermetically sealed TZero aluminum pans. Polymer samples were dried under vacuum before DSC analysis; samples (2–10 mg) were heat–cool–heat cycled at 10 °C min–1 between −85 and 150 °C. Tg values were taken from the second heating scan. Spray-dried dispersions were dried under vacuum before DSC analysis; samples (2–6 mg) were ramped at 5 °C min–1 from −85 to 140 °C; Tg values were taken from the first heating scan.

Turbidity Analysis of PNiPAm Thermoresponsiveness

The cloud points of PNiPAm-(1.3) and PNiPAm-(4.4) in the PBS solution (w/SIF, pH 6.5) were determined by transmittance measurements. Solutions were prepared in glass ampules at 9 mg mL–1 and placed in the holding cell of a heating block (±1 °C). A 10 mW helium–neon laser (λ = 633 nm) passed through a neutral density filter and then through the sample ampule. The transmitted light was focused by a lens, and the signal was collected by a photodiode detector. The heating rate was kept constant at 2.5 °C min–1. The cloud point was taken as the 50% transmittance point (50% T).

Spray-Dried Dispersion Preparation

Spray-dried dispersions were prepared at the lab scale using a mini spray dryer (Bend Research, Bend, OR). Solid dispersions were made either as binary blends of a homopolymer and phenytoin or ternary solid dispersions of two homopolymers and phenytoin. For dispersion preparation, a total-solid loading of 2.0 wt % drug and polymer were dissolved in 90:10 v/v MeOH/H2O. Samples were sprayed at a constant inlet temperature (75 °C), N2 flow rate (28.6 sL min–1), and solution flow rate (1.3 mL min–1). Spray-dried dispersions were collected and immediately dried under vacuum overnight and further stored at room temperature in a vacuum desiccator.

Solid dispersions were all made at 10 wt % phenytoin (90 wt % excipient). Binary solid dispersions were 10 wt % phenytoin and 90 wt % of a homopolymer. Ternary solid dispersions were 10 wt % phenytoin and 90 wt % excipient, where the composition of the excipient was varied across 80:20, 70:30, 60:40, 40:60, and 20:80 wt:wt PNiPAm:diluent, where the diluent was PDMAm-(2.0), PHEAm-(1.7), HPMCAS, or HPMC. Samples that are labeled as either “100:0” or “0:100” are the binary blends of phenytoin with a single excipient, either PNiPAm or the diluent.

Standard Dissolution Tests

Dissolution tests were performed in the phosphate-buffered saline solution (PBS, 82 mM sodium chloride, 20 mM sodium phosphate dibasic, 47 mM potassium phosphate monobasic, preheated to 37 °C) with 0.5 wt % simulated intestinal fluid (SIF) powder (FaSSIF, Biorelevant) at 37 °C. In a 1.5 mL microcentrifuge tube, appropriate amounts of spray-dried dispersion and PBS were added to target a total drug concentration of 1000 μg mL–1. Samples were mixed on a vortex mixer (Scientific Industries Vortex-Genie 2 on a setting of 10) for 1 min to suspend all solids and then placed (t = 0) in an isothermal sample holder held at 37 °C. At set time points (4, 10, 20, 40, 90, 180, 360 min), the samples were centrifuged at 13 000 g for 1 min, a 50 μL aliquot was removed from the supernatant, and the aliquot was diluted with 300 μL of high performance liquid chromatography (HPLC)-grade methanol. The samples were then vortexed for 30 s to resuspend the centrifuged solids, and the tubes were returned to the isothermal sample holder. The drug concentration was determined by reversed-phase high-performance liquid chromatography in 60:40 (v/v) H2O/MeCN. Samples were analyzed on an Agilent 1260 liquid chromatograph with a multi-wavelength UV–vis detector, 1260 MWD, at 225 nm. Separation occurred on an Agilent Poroshell 120 EC-C18 column with 120 Å pores. A calibration curve from 12.5 to 500 μg mL–1 was prepared, with an R2 value of 0.9986.

Early-Time-Point Dissolution Tests

Early-time-point dissolution tests were performed similarly to standard dissolution tests. Rather than the first instance of vortex mixing being for 1 min, the solution was vortexed for 15, 30, or 60 s, and the sample was then immediately centrifuged at 13 000 g for 1 min. An aliquot was taken, and the concentration of phenytoin was determined as described above. Because the vortexer was on the benchtop, dissolution was performed at 22 °C. However, significant heat loss from the PBS was likely limited in the time frames studied.

Dynamic Light Scattering (DLS)

Dynamic light scattering (DLS) measurements were conducted using a Brookhaven Instruments BI-200SM light scattering system equipped with a 637 nm laser. Samples were prepared at a concentration of 9 mg mL–1 of the polymer in phosphate-buffered saline (PBS). Samples were passed through a 0.2 μm filter to remove dust, loaded into glass tubes, and placed in a temperature-controlled, index-matching bath. The temperature was held at 37 °C during measurements. Intensity autocorrelation functions (Figure S6) spanning 1 μs to 1 s were collected at five scattering angles of 60, 75, 90, 105, and 120°, with an acquisition time of 10 min per angle. For each sample, a REPES analysis was performed on the autocorrelation function from 90°. The autocorrelation functions for samples that showed a single population by REPES were fitted to a second-order cumulant model and for samples that showed two populations by REPES were fitted to a double exponential decay model. Residuals (Figure S7) between the intensity correlation functions and the selected fits were plotted to determine the quality of the chosen fit. The mutual diffusion coefficient, Dm, was extracted from linear fits of the mean decay rate (Γ) vs q2 (Figure S8). Mean hydrodynamic radii (Rh) were obtained via the Stokes–Einstein relationship. Dispersities for the samples fit to a second-order cumulant model are given as the average μ2/Γ2 value across all five angles.

Results

Synthesis and Characterization of Polymeric Excipients

The polymers highlighted here were studied as solid dispersion excipients in binary and ternary blends with phenytoin (Scheme ). Synthetic solid dispersion excipients were prepared from three different acrylamide monomers via reversible addition–fragmentation chain-transfer (RAFT) polymerization. Based on the previous work demonstrating the effectiveness of N-isopropylacrylamide as a monomer unit in several polymeric solid dispersion excipient systems with phenytoin, the acrylamide backbone structure was maintained in all of the synthetic polymers studied here.[15−19] Homopolymers of poly(N-isopropylacrylamide) (PNiPAm), poly(N,N-dimethylacrylamide) (PDMAm), and poly(N-hydroxyethylacrylamide) (PHEAm) all inherited a starting RAFT chain-transfer agent (CTA) that possessed both a carboxylic acid terminus and a C12-alkyl chain terminus. This specific CTA was chosen for its ability to drive micellization of low-molar-mass polymers, as previously demonstrated with low-molar-mass PNiPAm.[18] Additionally, two common commercially available cellulosic excipients were studied: hydroxypropyl methylcellulose acetate succinate (HPMCAS) and hydroxypropyl methylcellulose (HPMC) (see the Experimental Section for more detail).

Scheme 1

Chemical Structures of Polymeric Excipients and Phenytoin

The cellulosic structures shown are representative, where the actual placement of all substituents is without regioselective control, and the degree of substitution of each functionality is independently varied.

Phenytoin (Scheme ) has a melting temperature (Tm) of 296 °C, a moderately lipophilic octanol–water partition coefficient (log P) value of 1.92, and a solubility for the crystalline form in DI water of 27.1 μg mL–1.[22,23] The hydantoin ring of phenytoin offers two hydrogen-bond donors and two hydrogen-bond acceptors, and the two phenyl rings confer hydrophobic character to the molecule.

The molecular and thermal characteristics of all polymer samples are summarized in Table . Three series of low-molar-mass homopolymers were prepared based on each acrylamide monomer. Polymers are named by their polymer abbreviation, followed by their Mn value determined by NMR spectroscopy, in kDa, assuming one end group per chain, for example, PNiPAm-(1.3) indicates a PNiPAm homopolymer with an Mn value of 1.3 kDa. Mn values as determined by SEC were equivalent or larger than Mn values by NMR spectroscopy. Dispersities for PDMAm samples were larger than typically seen for RAFT polymerizations. DSC was used to determine the glass transition temperatures (Tg’s) of the polymers. Within each homopolymer series, Tg values generally increased with increasing molar mass. Tg values for the synthetic polymers ranged from 52 to 94 °C. The cellulosic polymers possessed Tg values over 120 °C.

Table 1

Molecular and Thermal Characteristics of Polymer Samples

sample	Mn NMRa (kg mol–1)	Mn SECb (kg mol–1)	Mw SECb (kg mol–1)	Đ SECb	Tg (°C)c
PNiPAm-(1.3)	1.3	1.8	1.8	(1.0)	71
PNiPAm-(4.4)	4.4	4.9	5.0	1.02	94
PDMAm-(2.0)	2.0	2.3	4.4	2.0	52
PDMAm-(3.3)	3.3	3.4	14	4.1	58
PDMAm-(5.6)	5.6	11	30	2.8	61
PDMAm-(6.5)	6.5	10	51	5.1	67
PDMAm-(11.3)	11.3	22	102	4.7	78
PHEAm-(1.7)	1.7				61
PHEAm-(2.9)	2.9				59
PHEAm-(4.3)	4.3				61
PHEAm-(5.6)	5.6				63
PHEAm-(8.8)	8.8				63
HPMCAS		60	142	2.38	121
HPMC		11.8	15.7	1.33	132

Mn NMR determined by 1H NMR spectroscopy in DMSO-d6, 64 scans, d1 = 10 s.

Mn SEC, Mw SEC, and Đ determined by SEC using a light scattering detector.

Glass transition temperatures (Tg’s) as determined by DSC in the second heating cycle at a heating rate of 10 °C min–1.

To probe solubility, each polymer sample was dissolved at 9 mg mL–1 in phosphate-buffered saline (PBS, pH 6.5) at 37 °C. All of the samples were dissolved at 9 mg mL–1 such that no polymer particles were visually apparent (Figure S2). All PNiPAm and PDMAm samples gave clear solutions at this concentration. HPMCAS and HPMC also formed clear solutions at this loading and temperature. The lowest-molar-mass PHEAm-(1.7) sample solution was clear, though all samples of higher molar mass appeared hazy, indicating the presence of larger solution structures.

The thermal phase transitions of the two PNiPAm samples were studied by turbidimetry.[24−26] Cloud points of PNiPAm-(1.3) and PNiPAm-(4.4) at 9 mg mL–1 in PBS-based dissolution media (pH 6.5) with 0.05 wt % simulated intestinal fluid powder (SIF) were determined by transmittance (T) measurements. The cloud point (50% T) of PNiPAm-(4.4) was 38 °C. No cloud point was detected for PNiPAm-(1.3), up to 50 °C (Figure S3). Thus, while thermoresponsive behavior may contribute to the performance of the PNiPAm-(4.4) sample during dissolution, no thermoresponsive behavior is expected to contribute to the performance of the PNiPAm-(1.3) sample.

Dynamic light scattering (DLS) was used to evaluate the presence, size, and size distribution of solution structures that the excipients formed in PBS (pH 6.5, w/o SIF, 37 °C). No SIF powder was added to the PBS used for DLS to limit scattering complexities arising from SIF lipids. Apparent size distributions are shown in Figure . Most samples showed random residuals of the autocorrelation functions to the fits (Figure S7), suggesting reliability of the fitting models chosen. A systematic error was observed for the cellulosic polymers and the PHEAm-(4.3), -(5.6), and -(8.8) samples, indicating that the double exponential model used does not perfectly describe the size distributions within the solution. Plots of the mean decay rates (Γ) vs q2 are linear with near-zero intercepts (Figure S8), which is consistent with the relaxation processes being diffusive. Key results from the DLS studies in PBS are summarized in Table .

Figure 1

Apparent micellar distributions of polymer samples dissolved in PBS (pH 6.5, w/o SIF, 37 °C) at 9 mg mL–1. The scattering angle is 90°.

Table 2

Hydrodynamic Radius (Rh) Values Determined by DLS in PBS Buffer (pH 6.5, w/o SIF)

polymer	Rh (nm)	μ2/Γ2a
PNiPAm-(1.3)b	5, 28
PNiPAm-(4.4)b	12, 49
PDMAm-(2.0)c	4	0.12
PDMAm-(3.3)c	6	0.09
PDMAm-(5.6)c	6	0.16
PDMAm-(6.5)c	7	0.06
PDMAm-(11.3)c	9	0.05
PHEAm-(1.7)c	5	0.16
PHEAm-(2.9)b	6, 66
PHEAm-(4.3)b	8, 68
PHEAm-(5.6)b	9, 85
PHEAm-(8.8)b	7, 69
HPMCASb	9, 157
HPMCb	6	0.45

Calculated from the average μ2/Γ2 values across five angles.

Determined by fitting with a double exponential decay function.

Determined by fitting with a second-order cumulant expansion.

Both PNiPAm samples showed two populations by REPES. The first population was indicative of micelles, and the second population indicated the presence of larger aggregates. All PDMAm samples showed a single peak, suggesting only the presence of micelles, without larger aggregates. The lowest-molar-mass PHEAm sample showed a single peak, though the larger-molar-mass samples also showed a second population at larger sizes of about 100 nm, which could explain the haziness seen in those samples. HPMCAS appeared as two broad distributions around 9 and 157 nm and HPMC as a single broad distribution around 6 nm.

The lowest molar masses of each synthetic polymer, PNiPAm-(1.3), PDMAm-(2.0), and PHEAm-(1.7), form structures that have Rh values of 5, 4, and 5 nm, respectively. Making the assumption that the micelle cores are dense water-free spheres, aggregation numbers (Nagg) can be roughly estimated from the respective Rh values, the total Mn found by NMR spectroscopy, and a bulk density. Estimated values of Nagg for the PNiPAm-(1.3), PDMAm-(2.0), and PHEAm-(1.7) micelles were found to be 200, 100, and 200, respectively (sample calculation included in the Supporting Information).

Spray-Dried Solid Dispersions with Phenytoin

Solid dispersions were formulated as binary and ternary blends of phenytoin with either one or two polymeric excipients, respectively. The solid-state properties of the solid dispersions were analyzed by DSC. DSC is routinely used to quantify the crystallinity of solid dispersions; however, the melting temperature of phenytoin is 296 °C, which is above the degradation temperature of these polymer samples. Thus, it is not possible to directly probe the crystallinity of these solid dispersions by DSC. DSC can also be used to probe the miscibility of the drug within the polymer. A single glass transition temperature (Tg) was seen for all of the solid dispersions, consistent with the drug being molecularly dispersed within the excipient, with Tg values in the range of 54–99 °C (Table S1). Long-term stability of these solid dispersions would benefit from higher solid dispersion Tg’s, typically imparted by higher amounts of the cellulosic polymers relative to the lower-Tg polyacrylamides.

During dissolution testing, the solid dispersions studied were introduced to PBS with SIF powder to mimic intestinal conditions. Under in vitro conditions, the concentration of phenytoin loaded was kept constant at 1000 μg mL–1 by loading the SDD at 10 000 μg mL–1 (polymer loading at 9 mg mL–1). The loaded drug concentration is significantly above the aqueous solubility of crystalline phenytoin (27.1 μg mL–1) and below its estimated amorphous solubility (1280 μg mL–1).[27] This loading concentration was chosen to push the limits of supersaturation without approaching the amorphous solubility, thus avoiding liquid–liquid phase separation phenomena described by Taylor and others.[28] Sink dissolution conditions are met when the volume of dissolution medium is at least 3–10 times the volume that would give the drug’s saturation concentration at the given drug loading; at the determined loading of 1000 μg mL–1 loading, the target drug concentration is relatively close to its estimated amorphous solubility, and, thus, the conditions we use are considered nonsink by the above definition.[29]

Under routine dissolution conditions, dissolution vials were loaded with solid dispersion; PBS (w/SIF) was introduced (the addition of PBS marked t = 0) and the vials were immediately mixed using a vortex mixer for 60 s at room temperature. The samples were placed in an isothermal holder maintained at 37 °C. The concentration of dissolved phenytoin was then evaluated at 4, 10, 20, 40, 90, 180, and 360 min after initial placement in the isothermal holder. At each time point, the concentration of phenytoin was determined by centrifuging down any suspended solids, taking an aliquot of the supernatant, evaluating the concentration via reversed-phase HPLC (post-dissolution in methanol), and resuspending any solids by mixing on the vortex mixer for 60 s. The dissolution vial was then placed back into the isothermal holder until the next time point. The withdrawn medium was not replaced with a new medium after sampling so as to not disrupt the structures in the solution.

Early-time-point data was added to the routine dissolution profiles to increase the temporal resolution of the dissolution experiments. In a similar manner to the routine dissolution conditions, dissolution vials were loaded with solid dispersion. PBS (w/SIF) was added to the dissolution vials (this addition marked t = 0). The vials were then mixed using a vortex mixer for either 15, 30, or 60 s at room temperature (∼22 °C). The vials were then immediately centrifuged, and the concentration of phenytoin was determined as above. Each early-time-point sample was a singular time point from a separate sample, with no resuspension or later time points taken.

Dissolution of Binary Solid Dispersions with Micelle-Forming Homopolymers as Excipients

Binary solid dispersions of phenytoin with micelle-forming homopolymers as the solid dispersion excipients were studied to elucidate the effect of monomer type and molar mass on the dissolution performance of phenytoin. To enable a comparison against the previous work, dissolution studies using PNiPAm-(1.3) and PNiPAm-(4.4) are shown in Figure a, with added early-time-point resolution in Figure S4.

Figure 2

Dissolution of solid dispersions of 10 wt % phenytoin with 90 wt % (a) PNiPAm, (b) PDMAm, (c) PHEAm homopolymers, and (d) HPMCAS and HPMC. In all plots, the dissolution profile of crystalline phenytoin is also included as a reference (open circles). The loaded concentration of phenytoin was 1000 μg mL–1 (indicated with the dashed line). The dissolution was run in triplicate. The data shown are the mean values, and the error bars represent the range of data. (a) Adapted from ref (18) with permission (Copyright 2019 American Chemical Society).

The PDMAm homopolymer series was designed to probe the general ability of a micelle-forming homopolymer with an amide functionality to achieve and maintain a high supersaturation of phenytoin. Dissolution of PDMAm samples (Figure b) showed a limited increase in apparent solubility of phenytoin over its crystalline form. These results suggest that <500 μg mL–1 phenytoin was released into the dissolution media. However, early-time-point data for the PDMAm samples revealed a full release of phenytoin (Figure ) for the two lowest-molar-mass PDMAm samples at the shortest vortex times. Decreased phenytoin concentrations were seen at longer vortex times, indicating that crystallization is occurring rapidly, even during the vortexing process. At higher PDMAm molar masses, the effect of the hydrophobic alkyl tail is minimized, and the resultant polymer becomes more hydrophilic. Thus, while the higher molar mass samples did not show the same high initial concentrations of phenytoin as the lowest-molar-mass samples did, it is likely that these also fully released the phenytoin, but that phenytoin crystallized even before these early aliquots were taken.

Figure 3

Early-time-point dissolution (open circles) and routine dissolution (filled circles) of solid dispersions of 10 wt % phenytoin and 90 wt % PDMAm: (A) PDMAm-(2.0), (B) PDMAm-(3.3), (C) PDMAm-(5.6), (D) PDMAm-(6.5), (E) PDMAm-(11.3). All x-axes are presented in the log scale to make the initial time points easier to visualize. The dissolution profile of crystalline phenytoin was also included as a reference (black open circles). The loaded concentration of phenytoin was 1000 μg mL–1 (indicated with the dashed line).

When compared with the PNiPAm systems previously studied,[18] all PNiPAm and PDMAm samples contained amide functionality in the monomer structure were fully dispersed at 9 mg mL–1, formed micelles, and were able to achieve a full release of loaded phenytoin upon dissolution at short times. However, the PDMAm samples were not able to maintain the high level of phenytoin concentration. Thus, the attributes listed alone do not lead to generalized excipient features that can maintain high supersaturation of phenytoin, and PNiPAm remains an anomaly.

The PHEAm homopolymer series was designed to probe the generality of aqueous-dispersed, low-molar-mass excipients that formed micelles and possessed hydrogen-bond donating abilities at the amide functionality (unlike PDMAm). This secondary amide character matched the secondary isopropylacrylamide of the best-performing excipient and had the potential to probe the role of the combined amide carbonyl and hydrogen-bond donors. Again, this series probed the potential effect across molar masses, ranging from 2 to 9 kDa.

Dissolution of the PHEAm-based spray-dried samples also revealed minimal solubility enhancement of phenytoin over its crystalline solubility (Figure c). Early-time-point data (Figure S5) was less revealing of early time dissolution than seen for the PDMAm series. Our efforts have revealed that the elements of small micelles, polymer solubility, hydrogen-bond donors, and secondary amide functionality are not generalizable to excipients beyond PNiPAm for maintaining supersaturation of phenytoin.

Phenytoin was also formulated into solid dispersions with HPMCAS and HPMC for comparative purposes. Together, routine dissolution and early-time-point studies (Figure ) of these systems revealed peak dissolution within minutes and subsequent crystallization over 6 h. Notably, the complete dissolution seen in early time points for HPMC was not detected in the routine dissolution studies (Figure d). This finding stresses the importance of capturing the full release profile when designing and drawing conclusions from dissolution experiments, especially early time data for hydrophilic excipients and excipient blends.

Figure 4

Early-time-point dissolution (open circles) of solid dispersions of 10 wt % phenytoin and 90 wt % (A) HPMCAS and (B) HPMC. All x-axes are presented in the log scale to make the initial time points easier to visualize. In all plots, the routine dissolution profiles are included as filled in circles. The dissolution profile of crystalline phenytoin was also included as a reference. The loaded concentration of phenytoin was 1000 μg mL–1.

Area-under-the-curve (AUC) integrations are commonly used to quantitatively compare the effectiveness of a solid dispersion to increase the solubility of a drug. Enhancement factors (EFs) here are defined as a simple ratio of AUC over 6 h (AUC6h) of the solid dispersion to the AUC6h of pure phenytoin. Based on the target of 1000 μg mL–1 of phenytoin for 6 h, the maximum EF possible is 22.1. The attained EFs can be related to the fraction of the target, expressed as FET, or a percentage, expressed as FET• 100%. The EFs and FET• 100% values for all binary blends are included in Figure . Only three samples achieved EF values above 5: PNiPAm-(1.3), PNiPAm-(4.4), and HPMCAS. In the case of PNiPAm-(1.3), the FET• 100% reached 97%.

Figure 5

Enhancement factors and FET• 100 (%) values for all binary blends. The target of 1000 μg mL–1 of phenytoin corresponds to a maximum EF possible of 22.1, which is represented by the top dotted line. The bottom dotted line represents an EF value of 1, as set by the solubility of crystalline phenytoin over 6 h.

Dissolution of Ternary Solid Dispersions: Phenytoin with Blends of Micelle-Forming Homopolymers

Given that PNiPAm-(1.3) was the highest performing material in the samples studied, ternary solid dispersions were formulated with PNiPAm-(1.3) and various diluent excipients to probe the importance of PNiPAm content within small micelles for the dissolution enhancement seen in the PNiPAm/phenytoin binary system. By blending phenytoin with PNiPAm and a diluent solid dispersion excipient, we were able to evaluate how the dissolution enhancement relates to the presence and density of NiPAm chains in the micelle. The first two diluents used were PDMAm-(2.0) and PHEAm-(1.7). For these low-molar-mass synthetic polymers, the fully stretched length of the resultant micelle corona chains is roughly equivalent, given their similar degrees of polymerization. Also, since PNiPAm-(1.3), PDMAm-(2.0), and PHEAm-(1.7) are all synthesized from the same RAFT chain-transfer agent, the micellization of these homopolymers and mixtures of these homopolymers are assumed to be similarly driven by the association of the common C12 alkyl tail.

For blends of A–B and B–C block polymers, where B is considered insoluble in the solution and A and C are soluble, phase diagrams can describe and predict the thermodynamics favoring either mixed micelles with intermixed A and C coronas or mixtures of ordinary micelles with pure A and C coronas.[30] In these systems, we take A and C to be the repeating monomer units and B to be the common C12 alkyl core. At these low molar masses, we posit that χN between A (PNiPAm) and C (either PDMAm or PHEAm as the diluent) is sufficiently low that A/C mixing is favorable, and that mixed micelles are formed. With the added compatiblizing effect of water, the enthalpic driving force for A/C segregation of these short blocks is likely insufficient to override the entropic driving force to mix. Thus, dissolved samples of PNiPAm:PDMAm and PNiPAm:PHEAm blends are presumed to exist as mixed micelles with a common C12 hydrophobic core and a mixed corona of PNiPAm and PDMAm or PHEAm. Thus, the amount of NiPAm within the corona is expected to decrease for the mixed PNiPAm:PDMAm and PNiPAm:PHEAm systems.

To probe the solution structures formed in the absence of phenytoin, blends of the polymers at 100:0, 80:20, 70:30, 60:40, 40:60, 20:80, and 0:100 wt:wt PNiPAm:diluent were dissolved in PBS (w/o SIF, pH 6.5, 37 °C) (Figure ) and evaluated by DLS. For both PNiPAm:PDMAm (Figure A) and PNiPAm:PHEAm blends (Figure B), a single structure with an Rh of 4 to 5 nm remained evident in all samples. The larger structure seen for PNiPAm alone remained in all blends, and the size increased over ∼30 nm with increasing diluent present (Rh of 28 nm for pure PNiPAm, Rh of 57 nm for PNiPAm:PDMAm 20:80, Rh of 57 for PNiPAm:PHEAm 20:80). DLS results are summarized in Table .

Figure 6

Apparent micellar distributions of polymer blends of (A) PNiPAm-(1.3) and PDMAm-(2.0), (B) PNiPAm-(1.3) and PHEAm-(1.7), co-dissolved in PBS (pH 6.5) at a total polymer loading of 9 mg mL–1. The scattering angle is 90°. Samples are given as wt:wt PNiPAm:diluent, where 100:0 and 0:100 samples are a single polymer. Data was acquired at 37 °C.

Table 3

Hydrodynamic Radius (Rh) Values Determined by DLS in PBS Buffer (pH 6.5)

excipient blends	(wt/wt)	Rh (nm)a
PNiPAm-(1.3)/PDMAm-(2.0)	80/20	5, 33
	70/30	4, 38
	60/40	4, 40
	40/60	4, 52
	20/80	4, 57
PNiPAm-(1.3)/PHEAm-(1.7)	80/20	5, 31
	70/30	5, 39
	60/40	5, 39
	40/60	5, 50
	20/80	5, 56
PNiPAm-(1.3)/HPMCAS	80/20	5, 49
	70/30	5, 55
	60/40	6, 77
	40/60	6, 100
	20/80	7, 121
PNiPAm-(1.3)/HPMC	80/20	5, 28
	70/30	5, 29
	60/40	5, 27
	40/60	5, 32
	20/80	5, 26

Determined by fitting with a double exponential decay function.

Dissolution studies of the PNiPAm:PDMAm and PNiPAm:PHEAm ternary solid dispersions revealed that these two blend series showed nearly identical release profiles (Figure A,B). The similar abilities of these systems to achieve and maintain supersaturations of phenytoin over 6 h suggests similar mechanisms for phenytoin stability in the solution. The performance was essentially mirrored for the same wt:wt values across both systems. Thus, the performance of these blends was likely related to the fraction of PNiPAm present in the micelles and essentially independent of the diluent chains. When lowering the PNiPAm content relative to the diluent, crystallization was evident for both systems by the 6 h time point for the 60:40 PNiPAm:diluent composition. At all PNiPAm contents higher than 60:40 wt:wt PNiPAm:diluent, AUCs above 15 (FET• 100 (%) values above 68%) were reached.

Figure 7

Dissolution of solid dispersions of 10 wt % phenytoin and blends of (A) PNiPAm-(1.3) and PDMAm-(2.0), (B) PNiPAm-(1.3) and PHEAm-(1.7). In all plots, the dissolution profile of crystalline phenytoin was also included as a reference (open circles). The loaded concentration of phenytoin was 1000 μg mL–1. The data shown are the mean values, and the error bars represent the range of data.

Since both PDMAm and PHEAm are considerably more hydrophilic than PNiPAm, we hypothesize that the micelle coronae become too hydrophilic to encourage the partitioning of phenytoin within them. Thus, the stabilizing effect and solubilization enhancement of PNiPAm are decreased, regardless of whether that stabilizing effect is simply due to a hydrophobic/hydrophilic balanced environment that does not promote phenytoin crystallization or if it is due to specific, crystallization-disrupting phenytoin–PNiPAm interactions.

Dissolution of Ternary Solid Dispersions: Phenytoin with Micelle-Forming PNiPAm and Cellulosic Polymers

Ternary mixtures of phenytoin with PNiPAm and either HPMCAS or HPMC were also prepared. REPES analysis of the PNiPAm:HPMCAS (Figure A) and PNiPAm:HPMC (Figure B) blends showed two discrete populations contributed from each polymer. For the PNiPAm:HPMCAS blends, three species were apparent by REPES. The intermediate-sized population seen for the blends likely captured both the larger PNiPAm aggregates and the smaller HPMCAS aggregates. As the smallest PNiPAm population remains constant in size during the addition of HPMCAS, it is readily posited that mixing of HPMCAS into the PNiPAm micelle structure does not occur.

Figure 8

Apparent micellar distributions of polymer blends of (A) PNiPAm-(1.3) and HPMCAS, and (B) PNiPAm-(1.3) and HPMC, co-dissolved in PBS (pH 6.5) at a total polymer loading of 9 mg mL–1. The scattering angle is 90°. Samples are given as wt:wt PNiPAm:diluent, where 100:0 and 0:100 samples are a single polymer.

For the PNiPAm:HPMC blends, two populations from PNiPAm were seen, and the smaller one overlapped in size with the single peak seen for HPMC. With decreasing PNiPAm content, the presence of the larger PNiPAm-only aggregates decreased relative to the smaller, shared PNiPAm-and-HPMC peak. Because of the overlapping populations seen for PNiPAm and HPMC, definitive conclusions are difficult to make about the mixing of these species. Summarized DLS results are included in Table . The autocorrelation functions of all blends were fit with a double exponential decay fitting. For the samples with three populations, the systematic error seen in the residuals suggests that the double exponential fitting was not completely appropriate. Scattering intensity scales with Rh6, such that we expect the larger populations were present in a very small amount relative to the smaller populations.[31]

The PNiPAm:HPMCAS and PNiPAm:HPMC ternary systems allowed us to probe how the dilution of PNiPAm micelles, in general, not necessarily in a micelle corona as above, affected the ability of the ternary solid dispersions to achieve and maintain phenytoin supersaturation. Because HPMCAS and HPMC are water-soluble polymers that each achieve full, near-immediate release of phenytoin in binary blends (Figure A,B, respectively), their presence should not limit the ability of a PNiPAm-based ternary solid dispersion to achieve high initial supersaturations. Dissolution studies of PNiPAm:HPMCAS and PNiPAm:HPMC are shown in Figure A,B. The PNiPAm:diluent cellulosic ternary systems were able to maintain supersaturation of phenytoin to a lower PNiPAm composition than seen for the mixed micelle ternary systems. EFs above 15 were seen down to a PNiPAm content of 40:60 PNiPAm:diluent. Even just 20:80 PNiPAm:HPMC offered significant solubility enhancement and delayed crystallization over pure HPMC.

Figure 9

Dissolution of solid dispersions of 10 wt % phenytoin and blends of (A) PNiPAm-(1.3) and HPMCAS, (B) PNiPAm-(1.3) and HPMC. In all plots, the dissolution profile of crystalline phenytoin was also included as a reference (open circles). The loaded concentration of phenytoin was 1000 μg mL–1. The data shown are the mean values, and the error bars represent the range of data.

Diluting the relative amount of PNiPAm micelles with HPMCAS and HPMC did less to invoke crystallization over 6 h than directly diluting NiPAm within the micelle coronas with PDMAm and PHEAm. The difference in observed crystallization in these systems suggests that it is the concentration of PNiPAm specifically within the micelles that is critical to the supersaturation sustainment seen for these systems.

Solubility Enhancement across Series

The EFs for all ternary blends are included in Figure . Within each ternary series, four or five blends provided solubility enhancement factors over 15, among the highest seen to date for phenytoin. While the 80:20, 70:30, and 60:40 PNiPAm:diluent blends have near identical EFs across each diluent, differences can be seen at the 40:60 and 20:80 ratios.

Figure 10

Enhancement factors (EF) and FET• 100 (%) values for ternary blends of PNiPAm-(1.3) and a diluent of (A) PDMAm-(2.0), (B) PHEAm-(1.7), (C) HPMCAS, and (D) HPMC. Compositions are denoted by their respective PNiPAm/diluent wt/wt ratio. The engineering target of 1000 μg mL–1 of phenytoin corresponds to a maximum EF possible of 22.1.

Further illuminating the role of PNiPAm relative to the diluent, Figure relates the mol of NiPAm per mol of phenytoin in the solid dispersion to the EFs seen for all ternary blends. All samples outperform a simple average mixing behavior (with the exception of 20:80 PNiPAm:HPMCAS, which showed minimal crystallization retardation relative to a binary solid dispersion with HPMCAS). Although the EFs were generally above a simple averaging, it is unlikely that these effects were due to synergistic behavior between PNiPAm and the diluent. Furthermore, since the performance of the PNiPAm:PDMAm and PNIPAm:PHEAm systems was nearly identical, the results seem to be independent of the diluent polymer, which also argues against synergy. Rather, there seems to be both a critical density of PNiPAm within micelles and a critical amount of PNiPAm total, both are necessary to maintain high supersaturations of phenytoin in the solution.

Figure 11

Enhancement factor of the blends as a function to the molar ratio of NiPAm to phenytoin in ternary solid dispersions of phenytoin, PNiPAm, and (A) PDMAm, (B) PHEAm, (C) HPMCAS, and (D) HPMC. The dotted horizontal line marks the maximum EF, if all phenytoin was fully dissolved for the 6 h.

Discussion

The solubility of a solute, such as a drug molecule, in water largely depends on two independent characteristics: the crystallinity of the solute and the ability of the solute to favorably interact with water.[32] These two molecular characteristics are often captured in two physical properties, Tm and log P. Although these two physical properties can vary independently of each other, in combination they can be used to predict the solubility of the solid solute in water (SWsolid) through an empirically derived general solubility equation[33]where Tm is the melting temperature in degrees Celsius.

This general solubility equation allows for the relative contribution of the crystallinity of the solute and its ability to interact with water to be weighed against each other with respect to overall solubility. Samples with high Tm values are often highly crystalline, and samples with high log P values strongly prefer nonpolar environments over aqueous environments. These two qualities are, in fact, independent of each other, though, particularly in pharmaceuticals, there are often groupings of bioactive molecules within subspaces based on similar structural components.

Poorly soluble molecules with Tm values above 200 °C and with log P values below 2 are said to have “solid-state-limited solubility”, because in their solid state, these molecules form dense crystal lattices with strong, tightly associated intermolecular interactions.[34,22] The poor solubility of these molecules is mostly a function of the crystallinity of the compound, rather than a significant hydrophobic character. In this way, a moderately hydrophobic compound can have extremely poor aqueous solubility.

In contrast, poorly soluble molecules, which are more lipophilic (log P > 3) and have lower-Tm values, often show solvation-limited solubility, where poor interactions of the compound with water means that hydration of the compound is unfavorable. The higher lipophilicity of these compounds promotes self-aggregation over mixing into water, also contributing to their poor water solubility.[34,35]

Unfortunately, while solubility-enhancement studies often look at poorly soluble compounds, we believe that there should be a better separation between mechanisms of observed solubility enhancement based on these more underlying solubility-defining physical properties. While the terms “poorly soluble” in water and “hydrophobic” are often interchanged, these two qualities are not precisely synonymous. It is reasonable to posit that during the development of solubility-enhancing methods, compounds with solid-state solubility challenges would behave substantially differently from those with poor solvation-state solubility. In support of this, notably more solvation-limited-solubility drugs have been formulated and brought to the commercial market than solid-state-limited-solubility drugs.[36]

As mentioned, solid dispersions are known to offer several strategic advantages toward improving the solubility of a poorly soluble drug. These advantages include: (1) introducing the more-soluble amorphous form of the drug upon dissolution and (2) improving wettability/solvation by using hydrophilic excipient. Between these two advantages, it has been noted that strong crystallizer compounds (sometimes referred to as solid-state-limited solubility compounds) benefit more from predissolution amorphization than improved wettability.[34] In contrast, hydrophobic compounds are referred to as being solvation limited, where improved wettability through excipient choice adds more benefit than predissolution amorphization. We posit that the key to dissolution improvement of solid-state-limited-solubility molecules, like phenytoin, involves inhibiting the dissolved drug molecules from reforming intermolecular self-associations. The amorphous character afforded during spray drying inherently limits the crystalline character of the compound; the key is limiting crystallization-driving self-associations in the solution once the drug is dispersed. In this work, solid dispersion excipient systems that offered a threshold amount of PNiPAm-dense micelle coronas were able to delay crystallization of phenytoin over 6 h, giving insight to a possible platform for similar solid-state-limited-solubility molecules in the future.

Conclusions

In conclusion, we studied the relative importance of PNiPAm and its molecular properties for the dissolution enhancement of phenytoin. By specifically varying the chemical moieties of poly(acrylamide)-based homopolymers, we probed the relative importance of different properties with respect to solid dispersions achieving a full release of phenytoin with prolonged supersaturation. We synthesized PDMAm and PHEAm homopolymers that formed aqueous-dispersed micelles in the solution, with the aim of understanding the generality of poly(acrylamide)-based polymers to maintain phenytoin in the solution. Although no further homopolymers were identified as high-performing excipients with respect to maintaining supersaturation of phenytoin, we were able to design ternary systems that fully released phenytoin immediately upon dissolution. Through the formulation of ternary blends composed of phenytoin, PNiPAm, and a diluent of PDMAm, PHEAm, HPMCAS, or HPMC, it was revealed that two critical components were necessary for full sustainment of phenytoin supersaturations: (1) a high density of NiPAm within the micelle corona and (2) a critical concentration of PNiPAm micelles in the dissolution solution. These results will hopefully help guide future development of excipients with stabilizing functionalities in a highly dense formation, potentially for use with other solid-state-limited-solubility drugs.