When does drug stop leaving an amorphous solid dispersion with its polymer? Five polymers compared across drug loadings

Background and motivation

An increasing share of drug candidates is poorly water soluble, and an oral product has to bring enough of the drug into solution in the gut. Making the drug amorphous raises its solubility, but only by a limited factor: the paper cites a typical advantage of 2–20 fold over the crystal (Taylor and Zhang, 2016) R29. By the Noyes–Whitney equation, the dissolution rate gains about as much:

Here $dC/dt$ is the dissolution rate (the change of solution concentration $C$ with time $t$), $k$ is a dissolution rate constant, $C_s$ is the saturation solubility of the solid form being dissolved (crystalline or amorphous) and $C_t$ is the bulk concentration at time $t$ (Eq. 1). In plain terms: the rate is proportional to how far the solution is from saturation, so a 2–20-fold higher solubility gives at most a 2–20-fold faster dissolution, and the highest concentration an amorphous solid can reach is its amorphous solubility. That concentration is also where passive membrane transport is fastest (Raina et al., 2014). For exceptionally insoluble compounds, the authors argue, such a gain is "likely still insufficient" for adequate absorption C1.

An amorphous solid dispersion (ASD), in which the drug is molecularly dispersed in a polymer, changes this picture because the dissolution of each component then depends on the composition (Higuchi et al., 1965; Simonelli et al., 1969). Two regimes are described in the introduction. At high drug loading (DL), diffusion models are "thought to" describe release: drug and polymer start to dissolve independently, the polymer faster, and the drug left behind forms a drug-rich barrier layer at the dissolution front that then controls release according to Eq. 1. At low DL there is too little drug to form an intact drug-rich layer, so release is controlled by the dissolution of the polymer, and drug and polymer leave at the same rate. Corrigan (1985) had already noted the limits of Higuchi's diffusion models for non-disintegrating systems whose components differ greatly in solubility C2. In this congruent regime the drug release rate follows from the polymer release rate and the weight fractions (Eq. 2):

Here $N$ is the weight fraction of each component in the dispersion. In plain terms: when the polymer sets the pace, the drug is simply carried out in the proportion in which it sits in the solid. Since polymers dissolve orders of magnitude faster than poorly soluble drugs, formulating at a loading where release is congruent gives fast and complete release. It has a second benefit: the solution can rise above the amorphous solubility, where it undergoes liquid–liquid phase separation (LLPS) into drug-rich, nano-sized amorphous aggregates. These have been shown to act as a reservoir that keeps the flux across a membrane at its maximum in vitro (Indulkar et al., 2016), and some in vivo studies have linked them to higher bioavailability (Wilson et al., 2018; Stewart et al., 2017; Kesisoglou et al., 2019) C3.

The problem is that congruent release had been observed only at low loadings, at most about 25% w/w, and mostly for dispersions in polyvinylpyrrolidone/vinyl acetate (PVPVA) (Indulkar et al., 2019; Saboo et al., 2019; Bhardwaj et al., 2018; Que et al., 2019). The loading at which release switches from congruent to incongruent, with a steep fall in drug release, had been named the limit of congruency (LoC). For PVPVA dispersions it ranged from 5% DL for ledipasvir to 25% DL for ritonavir and is "thought to depend" on drug properties and drug–polymer interactions C4·R28. A low LoC means a large amount of polymer per dose, which for high-dose drugs leads to tablets that patients find hard to take. Other polymers are used in marketed or developmental ASDs (McKelvey and Kesisoglou, 2019; Newman, 2017), and poor release at high loading had been reported with them too (Davis et al., 2017; Purohit and Taylor, 2017; Jackson et al., 2016), but no correlation between polymer properties and the LoC had been established C5.

The study therefore asks how drug release changes with loading in one drug, felodipine (Fel), dispersed in five polymers of different hydrophobicity and chemistry: polyvinylpyrrolidone (PVP), PVPVA, Eudragit S 100 (EUDS), hydroxypropylmethylcellulose (HPMC) and hydroxypropylmethylcellulose acetate succinate (HPMCAS). Two structural analogues of felodipine, nilvadipine (Nil) and cilnidipine (Cil), were added at one high loading (40 wt%) in HPMCAS to see whether the drug's own hydrophobicity matters. Release of both drug and polymer was measured from compacts of constant surface area, and the solid surface, humid films and drug–polymer hydrogen bonding were examined to explain what was seen A1·A2·A3.

Background concepts

Congruent and incongruent release. Release is congruent when drug and polymer leave the dissolving surface in the same proportion as they are present in the solid, so that their release rates, each divided by its weight fraction, are equal. It is incongruent when one component, normally the polymer, leaves faster; the surface then becomes enriched in the other. The paper reads congruency directly by plotting the normalized drug and polymer rates side by side for every loading (Figure 7).

Limit of congruency and the "cliff". The LoC is the highest loading at which release is still congruent. Above it, drug release does not decline gradually but drops abruptly; the paper, following Indulkar et al. (2019), calls this "falling off the cliff". In this paper each LoC is reported as the last congruent loading tested, with loadings spaced 5 wt% apart for the hydrophilic polymers R15.

Water-induced amorphous–amorphous phase separation (AAPS). A dispersion that is a single amorphous phase when dry can split into a drug-rich and a polymer-rich amorphous phase when it absorbs water. Whether this is favourable depends on the loading, the polymer, the relative affinity of drug and polymer for water, and the amount of water absorbed, which in turn depends on drug–polymer interactions; how fast it happens matters too, because it competes with dissolution (Purohit and Taylor, 2015; Rumondor et al., 2009, 2010, 2011; Marsac et al., 2010) C39. The group had earlier shown that AAPS can turn a congruently releasing dispersion into an incongruently releasing one (Saboo et al., 2019) C40. AAPS is made visible here with prodan, a fluorescent probe that is expected to collect in the drug-rich phase, so that phase separation shows up as bright domains under a confocal microscope C7.

Two measures of polymer hydrophobicity. The paper ranks hydrophobicity in two independent ways. Water sorption measures how much water the bulk polymer takes up as relative humidity (RH) rises; a polymer that takes up less water is more hydrophobic, and the same quantity is what the authors later call hygroscopicity. The static contact angle of a buffer drop on a polymer film measures surface wetting: a larger angle means poorer wetting and a more hydrophobic surface C10·C11.

Neutral and enteric polymers. PVP, PVPVA and HPMC are neutral and dissolve without ionizing. HPMCAS carries acetyl and succinoyl groups and EUDS carries carboxylic acid groups (Figure 1); both dissolve only when these acid groups ionize, EUDS only above pH 7 (Thakral et al., 2013). For this reason all EUDS experiments were run in pH 7.4 buffer and all others in pH 6.8 buffer M6·M8·M9·M10·C30.

Hydrogen bonding read from the NH band. Felodipine has one NH group, a hydrogen-bond donor. In the neat amorphous drug it bonds to the carbonyl of a neighbouring drug molecule, and its infrared stretching band sits at 3342 cm⁻¹. A polymer that bonds more strongly to the NH than the drug itself shifts the band to lower wavenumbers; a weaker partner shifts it to higher wavenumbers (Konno and Taylor, 2006; Kothari et al., 2015; Tang et al., 2002) C24.

Higuchi's critical mixture ratio. In Higuchi's diffusion model for a two-component solid, both components can dissolve congruently only at one particular composition (Simonelli et al., 1969), given by Eq. 4:

Here $N$ is again the weight fraction, $D$ the diffusion coefficient in the boundary layer and $C_s$ the saturation solubility of each component. In plain terms: congruent release requires the drug:polymer ratio in the solid to equal the ratio of their diffusive fluxes at saturation. When the drug is far less soluble than the polymer, this ratio is tiny, so by this model any real dispersion should release incongruently. The paper uses the equation to show that the observed congruent release at 10–15% loading cannot come from this mechanism (see Discussion) E4·C35·C36.

Methods in detail

The study has one central measurement, release of drug and polymer from compacts at increasing loading, surrounded by measurements that rank the polymers and probe why release changes. The overall workflow is shown below.

Materials and dispersions

Felodipine and nilvadipine came from Chemshuttle and cilnidipine from Euroasia Chemicals. The polymers were PVP K29/32 (ISP Technologies), PVPVA 64 (BASF), HPMCAS MF grade (SE Tylose), Eudragit S 100 (Evonik) and HPMC Methocel E5 Premium LV (Dow) M1–M8. The pH 6.8 buffer (100 mM phosphate) was made from 7.057 g monosodium phosphate monohydrate and 6.936 g anhydrous disodium phosphate per litre; the pH 7.4 buffer used for EUDS from 2.872 g and 11.241 g of the same salts M9·M10.

Dispersions were made by dissolving drug and polymer (50 mg/mL total solids) in 1:1 v/v methanol:dichloromethane, removing the solvent on a rotary evaporator with a 45 °C water bath, drying overnight under vacuum, cryo-milling about 1 g at a time (2 min pre-cooling, 1 min grinding at 10 Hz) and keeping the 106–250 µm sieve fraction. Fresh dispersions were amorphous by powder X-ray diffraction and showed a single glass transition by modulated DSC (SI Section S1, Figure S1), which the authors take as a sign that the drug is homogeneously dispersed P5·R26·C6. Loadings were 5, 10, 15, 20 and 30 wt% for PVP, PVPVA and HPMC, and 10, 20, 30, 40 and 50 wt% for HPMCAS and EUDS A2. The glass transition temperatures of the pure components were 46.9 (felodipine), 175.5 (PVP), 106.2 (PVPVA), 154.5 (HPMC), 152.3 (EUDS) and 124.2 (HPMCAS), with no unit printed in SI Table S1; the values fit degrees Celsius M1·M4–M8·X18.

Measuring drug and polymer release separately

Problem. Release from a powder depends on particle size and wetting, which differ between polymers and loadings. To compare polymers fairly, the authors needed a constant dissolving area and separate assays for drug and polymer, so that congruency could be read directly.

How. They used Wood's intrinsic dissolution apparatus (USP <1087>). 100 mg of sieved powder was compressed in an 8 mm die at 1500 psi for 1 min, which exposes a flat face of about 0.5 cm². The die was rotated at 100 rpm in 100 mL of 100 mM phosphate buffer at 37 °C (pH 6.8; pH 7.4 for EUDS), under non-sink conditions with respect to the amorphous solubility. Nine 2 mL samples were taken over 1 h (every 5 min to 30 min, then every 10 min) and replaced with fresh medium. For the drug, 100 µL of each sample was mixed directly with 100 µL acetonitrile and assayed by reversed-phase HPLC; the rest was filtered through a 0.2 µm nylon filter for the polymer assay. PVP and PVPVA were measured by size-exclusion HPLC with UV detection at 205 nm, EUDS by size-exclusion HPLC at 210 nm, and HPMC and HPMCAS by HPLC with an evaporative light scattering detector; each run was done in triplicate P6·P7. From the release curves, each component's surface-normalized rate was calculated with Eq. 3:

Here $A$ is the dissolution rate (mg/min/cm²), $k$ the slope of the concentration–time regression line (mg/mL per min, without forcing it through the origin), $V$ the medium volume (100 mL), $S$ the tablet area (0.5 cm²) and $N$ the weight fraction of the component, used on the assumption that the components have similar densities. In plain terms: the slope is converted into mass released per minute per square centimetre and then divided by the component's share of the solid, so that a congruently releasing dispersion gives the same $A$ for drug and polymer, and each value can be compared with the intrinsic dissolution rate of the pure material E3·C8.

What to keep in mind. The drug samples were not filtered, so the measured drug concentration includes any drug-rich colloids as well as dissolved drug P6·X12. Some high-loading compacts were followed for longer than the stated hour: the 50% DL HPMCAS compact is plotted from 40 to 90 min (Figure 4F), and the 30, 40 and 50% DL EUDS compacts from 25, 80 and 120 min respectively, up to 160 min (Figure S5). Because the regression excludes the origin, the rates for these compacts describe release after a lag, and the regression windows are not stated R7·R11·X9.

Ranking polymer hydrophobicity

Water sorption isotherms of the five polymers and of amorphous felodipine were measured once each on a gravimetric analyzer at 37 °C: about 20 mg of sample was dried at 60 °C and 0% RH for 60 min, then taken from 5 to 95% RH in 10% steps, each held until the mass changed by less than 0.01% in 5 min or for at most 180 min P3. Contact angles were measured with a goniometer on spin-coated films of each polymer (100 µL of a 50 mg/mL solution spun at 3000 rpm for 60 s, then vacuum-dried for 24 h), using a drop of about 1.5 µL of dissolution buffer and three films per polymer P4.

Looking at the dissolving surface and at humid films

To see what happens at the surface, 30% DL tablets were removed after 10 min of dissolution and dried overnight. Micro-computed tomography (micro-CT; 60 kV, 3 µm per pixel) gave cross-sections through the dissolving face, scanning electron microscopy (SEM) showed the surface, and energy dispersive X-ray analysis (EDX) gave the surface chlorine content before and after dissolution. Chlorine is present only in felodipine (Figure 1), so more chlorine means a more drug-rich surface P8·P9·C22. Separately, 30:70 drug:polymer films containing 0.2 wt% prodan were spin-coated on quartz in a glove box at about 18% RH, so that no water entered during preparation, then held at 97% RH for 24 h and imaged by confocal fluorescence microscopy (excitation 407 nm, emission 425–475 nm) P10.

Interactions and the drug-hydrophobicity series

Transmission infrared spectra of 30:70 films on KRS-5 windows (64 scans, 4 cm⁻¹ resolution) gave the position of the felodipine NH band in each polymer P11. For the drug series, the octanol–buffer partition coefficient of nilvadipine and cilnidipine was measured by shaking each drug between octanol and pH 6.5 phosphate buffer for 24 h and analysing both phases by LC-MS (SI Section S2, Eq. S1):

In plain terms: log P is the base-10 logarithm of how much more drug sits in octanol than in water at equilibrium, a standard measure of lipophilicity. The values were 4.06 for nilvadipine and 5.40 for cilnidipine, against 4.46 for felodipine taken from the literature (Van Der Lee et al., 2001); all three are described as predominantly un-ionized across the pH range E5·M1·M2·M3. Water uptake of the three 40% DL HPMCAS dispersions at 100% RH was extrapolated from vapour sorption isotherms measured from 5 to 95% RH at 37 °C (SI Section S3) P15. Differences were tested by Student's t-test or one-way ANOVA with Tukey's comparison at p ≤ 0.05 P12.

Experimental analysis

Felodipine solubility

Crystalline felodipine dissolved to 1.3 ± 0.3 µg/mL and amorphous felodipine to 8 ± 1 µg/mL in pH 6.8 buffer at 37 °C (Table 1, n = 3), an amorphous advantage of about 6-fold, within experimental error of earlier values (Raina et al., 2014). With 1000 µg/mL of each polymer the crystalline solubility stayed between 0.9 ± 0.3 (HPMCAS) and 1.4 ± 0.1 µg/mL (EUDS, pH 7.4), the same within error (Table S2). So the polymers do not solubilize felodipine at equilibrium, and any rise above 8 µg/mL during dissolution is supersaturation, not solubilization R1·R2·C9.

The five polymers rank the same way by water uptake and by wetting

Figure 2 shows the sorption isotherms. PVP took up the most water at every RH, about 71 wt% at 95% RH, followed by PVPVA (about 47 wt%) and HPMC (about 28 wt%); EUDS and HPMCAS took up only about 13 and 10 wt%, and amorphous felodipine about 1.6 wt% at 85% RH, its last point. The contact angles of buffer (Table 2) ran in the same order, from 11.9° on PVP to 46.8° on HPMCAS. The authors rank hydrophobicity as HPMCAS ≥ EUDS > HPMC > PVPVA > PVP from sorption and HPMCAS = EUDS > HPMC > PVPVA > PVP from contact angle; the abstract merges the two into HPMCAS ≃ EUDS > HPMC > PVPVA > PVP, which the measurements "suggested" R3·R4·C10·C11. Each isotherm is a single run, so no uncertainty is given for Figure 2.

Table A. The five polymers. Water uptake read from Figure 2 (single runs, approximate); contact angles from Table 2 (mean ± SD of three films); neat dissolution rates read from Figure 7 (approximate); LoC from the text and Figure 7. Amorphous felodipine dissolves at about 0.0008 mg/min/cm² R3·R4·R13·R15.

Neat polymers dissolve hundreds to thousands of times faster than the drug

Figure 7A compares the intrinsic dissolution rates of the neat materials. Printed on the bars are the ratios of each polymer's rate to that of amorphous felodipine: 4600 for PVP, 4300 for PVPVA, 960 for HPMCAS, 460 for HPMC and 290 for EUDS. PVP and PVPVA dissolve fastest, at about 3.6 and 3.4 mg/min/cm², while EUDS dissolves more than an order of magnitude more slowly. The order does not follow the hydrophobicity ranking: HPMCAS (about 0.77) dissolves about twice as fast as the more hydrophilic HPMC (about 0.37) R13·C15.

Note by the AI writer (not discussed in the paper). The neat-PVPVA rate cannot be reproduced from its own release curve. In Figure S3A, PVPVA alone releases at an almost constant 0.83% per minute, about 50% in 60 min; for a 100 mg compact with 0.5 cm² exposed, Eq. 3 then gives about 1.7 mg/min/cm², half the plotted 3.4. The same check reproduces the plotted rates of neat PVP, HPMC and HPMCAS and of the polymer in the 5% DL PVPVA dispersion. The compact mass for the neat-polymer runs is not stated, so the source of the difference is unknown X10.

Hydrophilic polymers: fast and congruent at low loading, then a cliff

Figures 3 and 5 (PVP) and Figures S3 and S4 (PVPVA, HPMC) show the release profiles, and Figure 7B–D condenses them into rates. With PVP at 5 and 10% DL, drug and polymer profiles overlap; the felodipine concentration passed the amorphous solubility within 5–10 min and reached about 35 and 67 µg/mL at 60 min. From 15% DL the polymer kept dissolving (about 40–51% released in 60 min) while less than 0.5% of the drug was released, and the drug concentration stayed below 1 µg/mL. The authors put the LoC of felodipine–PVP at 10% DL R5·R6·C12. PVPVA behaved the same way with the switch one step later (LoC 15% DL), and HPMC with the switch at the same point as PVP (LoC 10% DL) R9·R10·C14. Table B gives the rates.

Table B. Normalized release rates (mg/min/cm², n = 3) of the hydrophilic-polymer dispersions, all read from Figure 7B–D (approximate; the figure's y-axes are broken, so the small drug rates are the least precise). Amorphous felodipine alone: about 0.0008 R14.a–o.

Three features of Table B matter for the argument. First, below the LoC drug and polymer rates match within the error bars, and they are close to the neat-polymer rate, which the authors read as release controlled by the polymer ("suggesting") C16. Second, one loading step above the LoC, drug release from PVP and PVPVA falls about 200-fold, from about 3.8 to 0.020 (PVP) and from about 1.9 to 0.009 mg/min/cm² (PVPVA). The abstract describes these rates as "more similar to that of the amorphous drug"; in numbers, they remain about 5–25 times above the neat amorphous drug but hundreds of times below the congruent rates. Third, the polymers behave differently once congruency is lost. PVPVA release falls almost as much as the drug (from about 1.9 to 0.14–0.19), PVP release roughly halves (from about 4.1 to 1.7–1.9), and HPMC release does not fall at all (about 0.16–0.20 against 0.17 at 10% DL), while HPMC's drug rate falls in two steps (about 0.14, 0.04 and 0.005 at 10, 15 and 20% DL) R14·X15·X11.

No drug crystallized in solution or on any dissolving surface, by powder X-ray diffraction and/or polarized light microscopy, so the cliff is not a crystallization effect R12.

Hydrophobic polymers: congruent up to 50%, with a slower decline

With HPMCAS (Figures 4, 6, 7E) drug and polymer released congruently at every loading from 10 to 50% DL, and every loading pushed the felodipine concentration above the amorphous solubility, reaching about 21, 55, 76, 35 and 10 µg/mL at 60 min for 10–50% DL. Up to 30% DL the drug rate (about 0.82–0.93 mg/min/cm²) was similar to that of neat HPMCAS (about 0.77); at 40 and 50% DL both components slowed to about 0.31–0.34 and 0.13. At 50% DL release began only after a lag, and data are plotted from 40 min R7·R8·R14.p–t·C13·C17. With EUDS (Figure S5, Figure 7F, pH 7.4) release was also congruent at every loading, but slow throughout: about 0.16, 0.14, 0.15–0.16, 0.12–0.13 and 0.07 mg/min/cm² from 10 to 50% DL against about 0.23 for neat EUDS. The text does not mention that the 30–50% DL EUDS compacts also released only after a lag (Figure S5) R11·R14.u–y·X9.

The authors read the HPMCAS decline as a "penalty": a large amount of hydrophobic drug slows the dissolution of the polymer itself, as reported earlier for HPMCAS (Chen et al., 2016), while drug and polymer still leave together. They place the decline at "≥40% DL" in the results but "notably for ≥30% DL" and "above 30% DL" in the discussion; the 30% DL rates in Figure 7E are still at or above the neat polymer, so the data place it between 30 and 40% DL C17·C29·X4.

The trade-off in one picture

Figure 14 plots felodipine release against loading for PVPVA, HPMCAS and EUDS on a linear axis. PVPVA gives the fastest release up to 15% DL (about 2.45, 2.35 and 1.90 mg/min/cm² at 5, 10 and 15% DL) and almost none from 20%. HPMCAS stays at about 0.82–0.93 to 30% DL and then declines. The curves cross between 15 and 20% DL, and at 20–30% DL HPMCAS releases felodipine about 100–250 times faster than PVPVA. EUDS stays low but flat. Figure 15 sets the neat-polymer dissolution rates beside the LoCs: PVP and PVPVA dissolve fastest and have LoCs of 10–15% DL, whereas the slower HPMCAS and EUDS show no LoC up to 50% DL; HPMC is slow and still has a LoC of 10% DL. The 50% points plotted for HPMCAS and EUDS in Figure 15 mean "no LoC up to the highest loading tested", not a measured LoC of 50% R14·R15·C54·X11.

Table C. Felodipine release (mg/min/cm²) at 10% and 30% DL, read from Figure 7 (approximate); the fold changes are the AI writer's arithmetic on these readings R14.

The table also shows what the authors' "hydrophilic polymers release faster at low loading" means in practice: it holds for PVP and PVPVA, which release felodipine 3–24 times faster than HPMCAS and EUDS at 10% DL, but not for HPMC, which is more hydrophilic than HPMCAS yet releases felodipine more slowly at 10% DL (about 0.14 against 0.82) and still has a LoC R14.b·R14.g·R14.l·R14.p·R14.u.

The surface turns porous and drug-rich only with the hydrophilic polymers

All surface and film measurements were made at 30% DL, above the LoC of the three hydrophilic polymers. After 10 min of dissolution, micro-CT cross-sections of PVP, PVPVA and HPMC tablets show a porous layer at the dissolving face (Figure 9A–C), which the authors call "consistent with" their faster polymer release; HPMCAS and EUDS tablets show no distinct interface and stay homogeneous with the bulk (Figure 9D, E) R19·C20. All initial surfaces were smooth by SEM (Figure S6); after 10 min the PVP and PVPVA surfaces carried round pores or pits, the HPMC surface was rough with voids, and the HPMCAS and EUDS surfaces showed angular particles (Figure 10) R20·R21·C21. EDX (Figure 11, Table D) showed significant increases in surface chlorine for the three incongruent tablets and none for HPMCAS and EUDS R22·C22.

Table D. Surface chlorine by EDX, read from Figure 11 (approximate). Only felodipine contains chlorine. The PVPVA tablet starts with more surface chlorine than the other four at the same loading, which the paper does not discuss R22·X16.

In the humid films (Figure 12), the PVP, PVPVA and HPMC films showed discrete, intensely fluorescent domains after 24 h at 97% RH, "consistent with phase separation", whereas the HPMCAS and EUDS films stayed uniform at both high and low magnification (Figure S7), "suggesting minimal or no phase separation". The films were uniform before exposure (data not shown) R23·C23. The domain sizes differ: the scale bars are 5 µm for PVP, 3 µm for PVPVA and 20 µm for HPMC.

At high loading, the drug's own hydrophobicity slows congruent release

At 40% DL in HPMCAS, all three dihydropyridines released congruently and more slowly than neat HPMCAS, and both rates fell as the drug's log P rose (Figure 8, Table E). All pairwise differences were significant (Table S4). The extrapolated water uptake of the three dispersions at 100% RH fell slightly in the same order (Figure S9) R16·R17·R18·C19.

Table E. 40% DL HPMCAS dispersions. Rates in mg/min/cm² from Table S3 (mean ± SD, n = 3); Tukey-adjusted p < 0.0001 for every drug-rate pair and 0.0004–0.0249 for the polymer-rate pairs (Table S4). Water uptake read from Figure S9 (approximate; overlapping error bars; the axis starts at 4.8 wt%) R16·R17·R18.

The polymers that bond most strongly to felodipine have the lowest LoCs

In the infrared spectra (Figure 13, Table 3), the felodipine NH band moved from 3342 cm⁻¹ in the neat glass to 3292 cm⁻¹ with PVP and 3293 cm⁻¹ with PVPVA, "consistent with" stronger drug–polymer hydrogen bonds. It moved up with EUDS and HPMCAS, "suggesting" weaker interactions than in the neat drug. With HPMC it hardly moved (3344 cm⁻¹); the carbonyl doublet of the neat drug at 1701 and 1682 cm⁻¹ merged into one broader band at 1699 cm⁻¹ (Figure S8), which the authors read as an interaction of similar strength to that in the neat drug ("appears to be the case"). A shoulder at 3342 cm⁻¹ remained in most dispersions, so the extent of drug–polymer bonding could not be quantified. The authors rank hydrogen-bond strength as PVP ≅ PVPVA > HPMC > HPMCAS > EUDS R24·R25·C25.

Table F. Felodipine NH stretch. In Figure 13 the EUDS and HPMCAS maxima appear the other way round from Table 3; the text order (EUDS weakest) agrees with the figure, so the two table entries look swapped R24·X2.

Discussion

Why the polymer has to carry the drug

The authors start from felodipine's numbers. It is a BCS class II drug whose amorphous form is only about 6 times more soluble than its crystal (crystalline solubility about 1 µg/mL), and its intrinsic dissolution rate is extremely low (Figures 3G and 7A). They take this as reinforcing that making an extremely poorly soluble drug amorphous is "unlikely" to give enough dissolution advantage, particularly for high doses. Because neat polymers dissolve several hundred-fold faster than the lipophilic drug, the in vivo advantage of ASDs "most likely" stems in large part from polymer-controlled release, and exploiting it is "likely essential" for drugs with solubility- or dissolution-limited absorption; drug loading decides whether it is achieved C26·C27·R1·R13.

Why the drug:polymer ratio alone cannot explain the LoC

Higuchi's model (Eq. 4) predicts congruent release only at the critical mixture ratio. For felodipine–PVPVA the SI (Section S4) takes the amorphous solubility of felodipine, 8 µg/mL, as $C_{s(\mathrm{d}\mathrm{r}\mathrm{u}\mathrm{g})}$, a literature solubility of PVPVA in the same buffer, 46 mg/mL, as $C_{s(\mathrm{p}\mathrm{o}\mathrm{l}\mathrm{y}\mathrm{m}\mathrm{e}\mathrm{r})}$, and estimates $D_{\mathrm{d}\mathrm{r}\mathrm{u}\mathrm{g}}/D_{\mathrm{p}\mathrm{o}\mathrm{l}\mathrm{y}\mathrm{m}\mathrm{e}\mathrm{r}}=10$ from the Stokes–Einstein equation, assuming similar viscosity for both and sizes of about 1 nm for the drug and 10 nm for the polymer coil. The printed result is $1.7\times {10}^{-6}$; on the main-text side the authors state that the loading of any practical ASD lies above the critical ratio (≥1% DL) R27·E4·E6·C36.

Note by the AI writer (not discussed in the paper). With the stated inputs, $10\times 0.008/46$ (both solubilities in mg/mL) gives $1.7\times {10}^{-3}$, a thousand times larger than printed, as if a unit conversion had been applied twice. The SI also refers to this relation as "equation 3 of the main article"; it is Eq. 4. Either value, about 0.17 wt% or less, is far below the loadings tested, so the argument is unaffected X1.

So by Higuchi's model every dispersion here should release incongruently, start diffusion-controlled and leave a porous drug-rich layer. The authors conclude that the model "may adequately describe" high-loading ASDs but fails for low-loading ones, which release congruently over a range of loadings: up to 10% DL for PVP and HPMC and up to 15% DL for PVPVA in this study C37. An older argument (Corrigan, 1985) holds that at low loading there is simply too little drug to form and support an intact porous drug-rich layer. The authors find that this "doesn't seem to hold true" as the sole explanation, for two reasons: the LoC of PVPVA dispersions varies from 5 to 25% depending on the drug, and the HPMCAS and EUDS dispersions stay congruent up to 50% DL. Evaluating only the drug:polymer ratio is therefore "clearly an over simplification", and the role of water in inducing phase separation at the solid–liquid interface should be considered C35·C38·R28.

The proposed mechanism: a race between phase separation and dissolution

The authors' explanation combines the earlier observation that water-induced AAPS turns congruent into incongruent release with the idea that AAPS and dissolution compete. Competition between them "may be key" in setting the regime C40·C41. The chain they describe is:

At low loading, the authors write, there is too little drug to form a contiguous drug-rich layer, and "there may also be" a lower tendency for AAPS or slower phase separation; these factors "likely help maintain" congruent release. At higher loading a greater tendency or rate of AAPS gives surface drug enrichment that both blocks polymer release from the layers beneath and reduces drug release. Their conclusion: "Water-induced AAPS is likely the cause" of the LoC and the loss of performance of high-loading PVP, PVPVA and HPMC dispersions; in the abstract the decline "was attributed to" water-induced phase separation C42·C43. The tablet images are offered as support: micro-CT, SEM and EDX "supported" AAPS followed by a porous drug-rich layer from preferential loss of polymer C46. The authors rely on earlier work for the step from spin-coated films to powdered dispersions (Purohit et al., 2017; Mugheirbi et al., 2017) C45.

Hygroscopicity as the switch

Since AAPS depends on how much water is absorbed, the authors suggest that polymer hygroscopicity "likely also plays a role". The congruent release of high-loading HPMCAS and EUDS dispersions "conceivably can be explained" by their lower water uptake (Figure 2), too low to induce AAPS; the uniform HPMCAS and EUDS films after storage at high RH are the "support for the conjecture" C44. A related point concerns ionization: the polymers that show a distinct LoC are hydrophilic and neutral, whereas HPMCAS and EUDS are hydrophobic and need to ionize to dissolve, so their interactions with water and drug differ between the dry and the hydrated state, where the pH is far above their pKa. The paper notes this without testing it C30.

Drug–polymer interactions

Earlier work had suggested that strong drug–polymer interactions "may" make dispersions resist water-induced AAPS, especially at high RH (Rumondor et al., 2009, 2011) C47. Here the opposite pattern appears: the polymers that bond most strongly to felodipine, PVP and PVPVA, have the lowest LoCs, while HPMCAS and EUDS, which "appear to form weaker interactions", give the most consistent release. The authors "observe no positive correlation" between hydrogen-bond strength and the LoC and conclude that the role of drug–polymer interactions, "if any", "clearly needs further evaluation" C48.

The drug's own hydrophobicity

For congruently releasing systems at high loading, the authors note that drug had been reported to slow HPMCAS dissolution, attributed to binding and swelling, strong interactions or greater hydrophobicity of the ASD (Corrigan, 1985; Chen et al., 2016; Merkel, 1983) C31. Choosing three analogues with different log P but, in their words, "similar ability to form intermolecular interactions with the polymer", they find that the fall in both rates with log P "substantiates the supposition" that drug hydrophobicity influences release, "at least" for congruent systems at high loading. Because an earlier study found no log P effect at low loading (Saboo et al., 2019), "it seems likely" that there is a cutover loading above which drug properties start to dominate, as the overall hydrophobicity of the system becomes dominated by the drug C32·C33·C34.

What it suggests for formulation

The authors draw three practical points. For PVP and PVPVA dispersions it is "crucial" to keep the loading at or below the LoC to avoid severely compromised dissolution and bioavailability C51. This pattern "may underlie" human pharmacokinetic results for grazoprevir–PVPVA tablets (Marota et al., 2016): the tablet made with a 30% DL dispersion gave the highest bioavailability, raising the loading to 40% halved it, and the neat amorphous drug gave less than a quarter of it; the source attributed the trend to a decline in dissolution rate with loading (McKelvey and Kesisoglou, 2019) C52·R30. Finally, more hydrophilic polymers "may be a better choice" at low loading, below the LoC, because they dissolve faster, while a more hydrophobic polymer "may allow" better release at somewhat higher loading; the choice depends on the dissolution profile needed for exposure and on the dose, and Figure 15 summarizes it as a trade-off C54·C55.

Note by the AI writer (not discussed in the paper). For a formulator the numbers in Table C translate into a simple rule for this drug: if the dose allows 15% loading or less, PVPVA gives the fastest release of the polymers tested; if the dose forces 20–30% loading, HPMCAS releases about 100–250 times more drug per minute than PVPVA, at the cost of being about three times slower than PVPVA at low loading. The LoCs are known only to the nearest 5% step, so a formulation intended to sit just under a LoC would need finer steps around it.

Limitations

Limitations stated by the authors. The LoC measured for a dispersion compact "may not coincide exactly" with the LoC of the same dispersion in a dosage form: the effect of additional excipients on the LoC "has not been widely investigated", surfactants have been shown to improve it (Que et al., 2019; Lang et al., 2016; Harmon et al., 2016), disintegrating formulations "may behave differently" from the compacts studied, and inorganic salts have been reported to improve release at high loading, partly by reducing polymer gelling (Takano et al., 2019; Xi et al., 2020) C53. The role of drug–polymer interactions in the LoC is left open ("if any") C48. The infrared data cannot give the extent of drug–polymer hydrogen bonding because residual drug–drug bonding remains C25. The water sorption profiles are single runs (Figure 2 caption) R3.

The AI writer's own judgement (1): how far the mechanism is shown. The release patterns are measured directly and are clear; the mechanism is not. Phase separation was imaged only in films of one loading (30%) after 24 h at 97% RH, a much slower and drier exposure than the 10 min the tablets spent in buffer, and the tablets were examined only at the same 30% loading, after drying. No film or tablet at or below a LoC was imaged, and nothing was observed inside a dissolving compact, so the claim that AAPS sets the LoC is not tested at the switch itself X5. The porous, chlorine-enriched surfaces are also what one would expect if the polymer simply dissolved faster than the drug, as in the Higuchi–Corrigan picture, whether or not AAPS came first X6. Below the LoC the concentrations exceed the amorphous solubility, which the authors describe as "leading to LLPS", but no drug-rich colloids were measured, and the unfiltered drug samples would include any that formed X12.

The AI writer's own judgement (2): hygroscopicity and interactions as explanations. Water uptake separates the two groups of polymers, but it does not order the LoCs within the hydrophilic group: HPMC takes up much less water than PVPVA (about 28 against 47 wt% at 95% RH) yet has the lower LoC (10 against 15% DL). Water uptake was measured on the neat polymers, once each, not on the dispersions X7. In this polymer set, hydrogen-bond strength, hydrophilicity and the neutral-versus-ionizing split all line up, so the data cannot separate their effects; the absence of a positive correlation between bonding and LoC is not evidence that bonding has no effect X14. The analogues in the log P series are assumed to interact similarly with HPMCAS, but their interactions were not measured, and their extrapolated water uptakes differ by less than their error bars X13. Finally, the "barrier to polymer release" is clear only for PVPVA; PVP release roughly halves above the LoC and HPMC release does not fall at all X15.

The AI writer's own judgement (3): points to check in the source. Several printed values or statements do not agree with each other, though none changes the main conclusions. The neat-PVPVA dissolution rate (about 3.4 mg/min/cm², and its 4300-fold ratio) is not reproduced by its release profile in Figure S3A, which gives about 1.7 X10. Table 3 and Figure 13 disagree on which of EUDS and HPMCAS shows the larger NH shift X2. The loading at which HPMCAS release declines is given as ≥40% and as ≥30% DL X4. The Higuchi critical ratio is misprinted by a factor of 1000 X1. The discussion names "HPMCAS" where the fluorescence images show phase separation for HPMC X3. The SI cites "Figure S8" for sorption isotherms that are not shown and refers the confocal images to "Fig. 13" instead of Figure 12 X17. EUDS was run at pH 7.4 rather than 6.8 X8, the high-loading HPMCAS and EUDS rates exclude a lag phase whose length the text does not discuss X9, and the plotted LoCs of 50% for HPMCAS and EUDS mean "none found up to the highest loading tested" X11.

Conclusion

Within one drug, felodipine, and non-disintegrating compacts at 37 °C, the paper shows two clearly different ways in which release depends on loading. The hydrophilic, neutral polymers PVP, PVPVA and HPMC carry felodipine out at the polymer's pace only up to 10% (PVP, HPMC) or 15% DL (PVPVA); one step higher, drug release collapses while the polymer keeps dissolving. The hydrophobic, enteric polymers HPMCAS and Eudragit S 100 keep drug and polymer leaving together up to 50% DL, more slowly, with a decline above 30–40% DL that also depends on the drug's lipophilicity. At 30% loading only the hydrophilic polymers leave a porous, drug-rich surface and phase-separate in humid films. The authors attribute the cliff "likely" to water-induced phase separation and the steadiness of HPMCAS and EUDS "conceivably" to their low water uptake, and they find no protective effect of stronger drug–polymer hydrogen bonding. The practical message they draw is a trade-off: hydrophilic polymers for fast release at low loading, hydrophobic polymers for release that holds up at higher loading C43·C44·C48·C49·C55.