Why strong drug–polymer binding lowers the drug-load limit of copovidone dispersions: the polymer follows the drug into the insoluble phase

Background and motivation

Amorphous solid dispersions (ASDs), in which a poorly soluble drug is mixed at the molecular level into a polymer, are used because they dissolve quickly and can create high supersaturation. On dissolution they often shed a second phase of amorphous, drug-rich nanoparticles a few hundred nanometres across, formed by liquid–liquid phase separation (LLPS); these particles equilibrate quickly with the solution and keep it supersaturated, which "is thought to" help absorption (Section 1; refs 10–18) C7. The practical problem is drug load. A tablet with little polymer is smaller, but higher drug loading risks crystallization, and some ASDs that stay amorphous still release poorly at only about 10–20% drug (refs 4, 5, 19–26) C1·R17.

The threshold is called the limit of congruency (LoC). Below it, drug and polymer leave the ASD at the same percentage rate (congruent release): the polymer controls the rate, drug can release completely even in non-sink conditions, and drug-rich nanoparticles form. Above it, the two rates diverge, drug release becomes very slow and incomplete, and no nanoparticles form. The pattern is best known for copovidone (PVPVA, Kollidon VA 64), has also been seen for some hypromellose and HPMCAS dispersions, and among drug–PVPVA systems the LoC ranges from 5 to 40% drug load (refs 4, 5, 24, 27) C1·R17.

Two lines of earlier work frame the paper. First, a drug-rich layer is often seen on the surface of a dissolving ASD above the LoC and "is thought to be largely responsible" for the collapse in release; it was proposed to arise from rapid, water-induced amorphous–amorphous phase separation (AAPS) at the surface. Han and co-workers argued that the hydrophobic phase made this way releases as nanoparticles while it stays below a percolation threshold, but forms a continuous insoluble network above it (Han 2019, ref 32) C8. The authors point out what this picture left open: it assumed that the hydrophobic phase is mostly drug, with little water or polymer, and it offered no way to predict the percolation threshold or the LoC C9. Second, an earlier study by the same group (Que 2021, ref 14) used the four analogues studied here and found that the LoC falls as drug–polymer interactions get stronger: phenolphthalein (PHPH), which hydrogen-bonds to the vinylpyrrolidone carbonyl of PVPVA, had the lowest LoC, and the methylated, non-interacting control (MePHPH) the highest. Saboo and co-workers saw the same with indomethacin and its non-hydrogen-bonding ester (ref 31) C2·R17. Why the interaction should matter was not understood C9.

The paper sets out to find how drug–polymer interactions change where PVPVA goes when the drug separates from water, whether into the insoluble drug-rich phase or into the aqueous phase, and to relate this phase composition to the LoC A1. Three kinds of measurement are used: solution NMR and HPLC of drug nanoparticles in water, PVPVA exchange between suspended ASD particles and polymer solutions, and infrared analysis of the surface of partly dissolved compacts.

Background concepts

Hydrophobic and hydrophilic phases

When water makes a drug–polymer mixture separate, two phases form: an insoluble, drug-rich phase, which the paper calls the hydrophobic phase (the nanoparticles in a dilute suspension, or drug-rich domains inside a wet ASD), and a water-rich phase holding the dissolved polymer, the hydrophilic or aqueous phase. All compositions in the paper are given on a dry basis, as weight percent of PVPVA among drug plus PVPVA, because the water content of the hydrophobic phase was not measured C50. "Drug load" (DL) is the weight percent of drug in the dry ASD.

Ternary phase diagrams, tie lines and the plait point

A drug–PVPVA–water mixture can be drawn in a triangle whose corners are the pure components. The authors expect a type I diagram in the Treybal classification: water–PVPVA and drug–PVPVA are miscible, drug–water is not, so PVPVA acts as a "cosolvent" and the two-phase region forms a dome against the drug–water side ("Typically", refs 54–57) C31. Inside the dome a mixture splits into two coexisting phases whose compositions, the binodes, lie on the binodal curve; the straight line joining them is a tie line, and the overall composition always lies on it. The spinodal lies inside the binodal and marks where a mixture becomes unstable and separates spontaneously (spinodal decomposition). The binodal and spinodal meet at the plait point, also called the critical point, where the two coexisting compositions become identical (Figures 8 and 9).

The paper uses one rule from this geometry, the lever rule, which it names but does not print E5. On a tie line, the phase whose binode lies closer to the overall composition is the larger one. If a mixture separates just inside the dome near the water-rich binode, it makes a little hydrophobic phase; near the drug-rich binode, a lot.

Associative and segregative phase separation

When two solutes separate from water, the second solute can go either way (refs 70, 71). In segregative separation one phase is enriched in solute 1 and depleted in solute 2: for these systems, a drug-rich phase with little polymer, and the polymer in the water. In associative separation one phase is enriched in both solutes and the other depleted in both; the paper notes that such a phase has also been called a complex coacervate C47. On a ternary diagram this decides the tie lines. In an associative system the tie lines have a positive slope, which marks PVPVA accumulating in the hydrophobic phase, and the plait point lies on the left, low-drug side of the diagram; in a segregative system the tie lines have a negative slope and the plait point lies on the right, high-drug side (Figure 9) C48. The paper draws both cases as schematics; it measured no tie line itself X14.

Why two solutes that bind each other can both precipitate

Hydrogen bonding between a solute and water helps it dissolve. But two solutes that bind each other strongly can drop out of water together as one insoluble, amorphous phase, because the mutual bonds replace their contacts with water. Known cases include hydrophobic ion pairing, insoluble complexes of polyacrylic acid with polyvinylpyrrolidone, and insoluble complexes of polyvinylpyrrolidone with many phenols (refs 40, 41, 58–68) C43. An earlier variable-temperature NMR study found that polymer partitioning into a drug-rich phase is driven by entropy, and the authors of that study "postulated" that drug–polymer interactions lower the enthalpy cost of moving the polymer out of water (Ueda 2021, ref 69) C45. Both ideas are used later to read the data.

Methods in detail

The study uses four chemical analogues on the phenolphthalein scaffold (Figure 1) that differ only in their substituents and so in how they bind the vinylpyrrolidone carbonyl of PVPVA, as established spectroscopically in the earlier study: PHPH (two phenolic OH groups, hydrogen bonds), TIPP (3′,3″,5′,5″-tetraiodophenolphthalein, OH groups plus four iodines, hydrogen and halogen bonds), MeTIPP (its methyl ether, halogen bonds only) and MePHPH (the methyl ether of PHPH, no specific interaction) M1–M4·C2. PHPH and TIPP were bought from Sigma-Aldrich; MePHPH and MeTIPP were synthesized as before M1–M4. The only polymer is PVPVA, with a vinylpyrrolidone to vinyl acetate ratio of 1.2 M5. The paper has four parts, three experimental and one argued on paper.

Contribution 1: How much PVPVA ends up in the drug-rich nanoparticles

Problem. The earlier study suggested that the amount of PVPVA in the drug nanoparticles follows the interaction strength, but a clean, equilibrium number was needed for each drug, and equilibrium is hard to prove for a particle that may still be gaining or losing polymer.

How. Nanoparticles were made by solvent shift in two ways (Section 3.1). In method A the drug, dissolved at 10 mg/mL in an organic solvent (DMSO-d6 for NMR samples; ethanol for PHPH and MePHPH and DMSO for TIPP and MeTIPP for HPLC samples), was injected (30.0–38.2 µL) into 3 mL of buffer already containing 0.5 mg/mL PVPVA, stirred at 1000 rpm at 37 or 60 °C. The amount was set to give about 100 µg/mL of drug in the nanoparticle phase on top of what dissolves; for PHPH, for example, 127.5 µg/mL in total because 27.5 µg/mL stays in solution. In method B the drug and PVPVA were dissolved together in the organic solvent and injected into plain buffer to the same totals P1·P2. PHPH, MePHPH and MeTIPP were tested in pH 6.8, 50 mM phosphate buffer; TIPP in pH 2.5, 50 mM phosphate buffer, chosen to keep it unionized P1·C11. The reasoning is that method A should start with polymer-poor particles, since little PVPVA is near the drug when the particles form, and method B with polymer-rich particles; when the two converge, the composition is taken as equilibrium C10.

The aqueous phase was measured by ¹H NMR on an 800 MHz spectrometer with a cryoprobe, directly in water with WATERGATE water suppression, at the equilibration temperature (8 scans; external reference in a coaxial insert; PVPVA calibrated from 0.01 to 20 mg/mL) P8. The PVPVA content of the particles then follows from what is missing from the solution (eq 1):

Here $W_{p,2}$ is the weight percent of polymer in the hydrophobic phase on a dry basis, $C$ a concentration, the subscripts p and d polymer and drug, $T$ the total in the suspension, 1 the aqueous phase and 2 the hydrophobic phase (nanoparticle or ASD particle). In plain terms: whatever polymer and drug are not dissolved must sit in the particles, so their ratio gives the particle composition E1. For MePHPH the dissolved drug was below 5 wt% of the total and was not measured; the literature amorphous solubility at 37 °C was used. TIPP and MeTIPP in solution were below detection and set to zero C12. As a cross-check, pellets were centrifuged out (21,100g, 15 min, 37 °C), blotted, dissolved in DMSO-d6 and analysed directly (eq 2) P9:

In plain terms: the same polymer fraction, now measured on the dissolved solid itself E2. HPLC (a C18 column for drug, size exclusion for PVPVA) was used to confirm the NMR values P10, and pellets were checked for crystals by polarized light microscopy P13. Note by the AI writer (not discussed in the paper): eq 1 can be checked on Figure 4c. With 30 mg of a 28% PHPH ASD in 15 mL of 1 mg/mL PVPVA, the totals are 2.44 mg/mL PVPVA and 0.56 mg/mL PHPH; with about 0.65 mg/mL PVPVA left in solution after 48 h and 27.5 µg/mL PHPH dissolved, eq 1 gives about 77.1%, the same as the plotted value of about 77% E1·R5.e–f.

Amorphous solubility. The amorphous solubility of PHPH and MePHPH in plain buffer was measured in triplicate by infusing an ethanol stock into buffer with a syringe pump and watching for light scattering at 360 nm with an in-situ UV probe; the onset of scattering marks the amorphous solubility. Values in buffer with 1 mg/mL PVPVA came from the earlier study, and TIPP and MeTIPP are too insoluble to measure P6. The printed stock concentrations, 0.3 mg/mL for PHPH and 20 mg/mL for MePHPH, look swapped: a 0.3 mg/mL stock could not reach the 200 µg/mL measured for PHPH without adding about twice the buffer volume of ethanol. The temperature of this measurement is not stated X1·X2.

Contribution 2: Which way PVPVA moves between ASD particles and solution

Problem. Congruent release needs the polymer to leave the ASD. Is there always a thermodynamic push for a water-soluble polymer to do so?

How. ASD powders made by rotary evaporation from ethanol were suspended (30 mg in 15 mL of pH 6.8 buffer, 37 °C, 300 rpm) and PVPVA was followed in the particles and in solution after 1, 24 and 48 h P3·P7. Only loadings above the cited LoC were used, because below it the particles dissolve quickly: MePHPH at 70% and PHPH at 10, 28 and 50%. The buffer usually contained 1 mg/mL PVPVA, so that polymer could move in either direction C21. The 28% PHPH ASD was chosen because 28% is the average drug load that the 10 and 50% particles reached at equilibrium and is close to the composition of PHPH nanoparticles; it was also tested in plain buffer R15·P7.

Why it works. If the equilibrium hydrophobic phase holds less polymer than the ASD, PVPVA should flow out; if it holds more, PVPVA should flow in from the solution. The direction of flux therefore shows on which side of the equilibrium composition each ASD lies.

Contribution 3: What the surface of a dissolving compact becomes

How. Compacts of 100 mg were pressed in a 0.5 cm² die at 1500 psi for 60 s and dissolved in the die (100 mL, 37 °C, 100 rpm) P4·P5. For PHPH compacts the medium was spiked with 40 µg/mL PHPH so that drug on the surface could not dissolve; MePHPH is so insoluble that nothing was added C13. The 40 µg/mL is above the 27.5 µg/mL measured with 1 mg/mL PVPVA but below the 200 µg/mL measured in plain buffer, and the medium held no added polymer except in one test, so some early drug release from the PHPH surfaces is not excluded X20. After 10, 30 or 60 min the compact was blotted, dried in a stream of air for about 15 min, taken out of the die, dried under vacuum overnight and its surface read by ATR-FTIR (128 scans) P12. The logic: below the LoC drug and polymer leave together, so the surface composition should not change; above it drug release is negligible, so any change in surface drug content must come from PVPVA moving out of or into the surface C23.

Surface composition was read from standard curves made from ASDs of known drug load (Figure S7, each point in triplicate). For PHPH ASDs, spectra were normalized to the PHPH band at 1513 cm⁻¹ and the height of the PVPVA band (1492 cm⁻¹ in the text, 1491 cm⁻¹ in the SI caption X9) was used through a quadratic fit:

In plain terms: the drug load DL (%) rises with the inverse $x$ of the normalized PVPVA band height, over standards from 5 to 50% E3·R13. For MePHPH ASDs, spectra were normalized to the PVPVA band at 1674 cm⁻¹ and the MePHPH band at 1509 cm⁻¹ was used:

In plain terms: the normalized MePHPH band height $I$ grows exponentially with drug load, over standards from 30 to 90% E4·R13. Note by the AI writer (not discussed in the paper): running the nominal standards back through these fits reproduces the starting points of Figures 5 and 6. The 15% PHPH standard returns about 13.8%, the 30% and 50% MePHPH standards about 34% and 44%, matching the 0 min values of about 13.5, 34 and 44%. The offsets from the nominal loads therefore come from the fits, and surface values are best compared with their own 0 min value. The MePHPH readings near or above 90% also lie at or beyond the highest standard X8.

Contribution 4: A phase-diagram account of the LoC

Problem. Why would a strongly interacting drug form a release barrier at a lower drug load?

How. The authors reason on a hypothetical type I diagram with horizontal tie lines (Figure 8). As water enters, the polymer matrix takes up solvent with little chain diffusion until the chains disentangle, and the hydrophobic drug stays trapped, so the ASD hydrates at a constant drug/polymer ratio: a straight path toward the water corner C32. Because water enters fast, the mixture is assumed to separate by spinodal decomposition where this path crosses the spinodal, though the conclusion would be the same for nucleation from the metastable region C33. Both phases are assumed to reach their binodal compositions at once C35. By the lever rule, an ASD whose polymer/drug ratio lies above that of the plait point (ASD 1) enters the dome near the hydrophilic binode and makes a small, dispersed hydrophobic phase; one below it (ASD 2) enters near the hydrophobic binode and makes a large, continuous one. Near the plait point a small shift in the entry point changes the hydrophobic volume a lot; far from it, little C34. Plotted against drug load, the hydrophobic-phase volume at the onset of separation is "approximately a step function" with the step at the plait point, so "significant or full continuity" of the hydrophobic phase "is expected" near the plait point (Figure 8d) C36. Because real systems do not separate instantly, the plait point is "only an approximation" of that switch C37.

The authors then add the interaction. In an associative system, where PVPVA joins the drug in the hydrophobic phase, the plait point sits at a low drug content, so the step to a large hydrophobic phase, and with it the LoC, comes at a lower drug load than in a segregative system (Figures 9 and 10) C48·C49. Figure 9 adds a second idea, composition X: the ASD whose polymer/drug ratio equals that of the hydrophobic binode in equilibrium with a dilute solution. An ASD richer in polymer than X should lose PVPVA to the water; one richer in drug than X should take PVPVA up; at X nothing should move C53·C54. This is the reasoning that Contribution 2 tests.

The authors list three assumptions behind the construction (Section 5.1). First, the two phases equilibrate as fast as water arrives, so they stay in pseudo-equilibrium with the local water content; PHPH and MePHPH "may adhere" to this, since their nanoparticles equilibrated in less than 5 min and compact surfaces in less than 10 min C38. Second, discrete hydrophobic domains do not substantially coalesce; this "may be true" for all four systems, since particles at the end of dissolution were about 150–400 nm in the earlier study C41·R16. Third, a continuous structure persists once formed and is not sheared off; the compact surfaces of Contribution 3 are offered as support C42.

Experimental analysis

PVPVA lowers the amorphous solubility of PHPH

In pH 6.8 buffer the amorphous solubility was 200 ± 20 µg/mL for PHPH and 2.81 ± 0.13 µg/mL for MePHPH. With 1 mg/mL PVPVA (values from the earlier study) it was 27.5 ± 0.9 and 4.40 ± 0.04 µg/mL (Table 1). PVPVA thus cut PHPH's value more than seven-fold and raised MePHPH's by about 50% R1·C14. The authors call the PHPH result surprising and read it as insoluble PHPH–PVPVA complexes ("likely due to"), as reported for other compound–polymer pairs; in the discussion they write that the complex is "apparent" from Table 1 C15. No complex was characterized directly. The two columns of Table 1 come from two studies, and the temperature of the buffer-only values is not stated X2.

The nanoparticles of strongly binding drugs are mostly polymer

For PHPH and MePHPH, methods A and B gave the same composition after only 5 min at 37 °C, which the authors take as equilibrium ("indicating") R2·C16. TIPP and MeTIPP equilibrated very slowly at 37 °C: in Figure S1a the two TIPP routes still sat at about 46.5 and 25 wt% PVPVA after about 72 h. At 60 °C the TIPP routes came to about 39 and 32 wt% at 3 h and about 40 and 37 wt% at 4 h, which the authors call equilibration within 3 h. MeTIPP crystallized: crystals are marked at 2 h in Figure S1c and the text gives crystallization after 1 h, so the mean of the two routes at 1 h was taken as an approximate value and, because the polymer content was falling, as an upper bound R9·C17·C18·X12. To compare all four drugs, every suspension was then measured at 60 °C by NMR (Figure 2a).

Table A. PVPVA in the equilibrated drug nanoparticles, wt% on a dry basis (Table 2, n = 3, 5 min at 37 °C; 60 °C values read from Figure 2a, approximate; TIPP in pH 2.5 buffer). Interaction types and LoCs are cited from ref 14 R2·R3·R8.

The ranking is steep. PHPH nanoparticles are about three-quarters polymer, TIPP nanoparticles about one third, and MeTIPP and MePHPH nanoparticles less than one tenth R3·C19. Temperature changed this little: PHPH held about 77 wt% at 60 °C and about 71 wt% at 37 °C, and for MePHPH the error bars were larger than any temperature effect, so the authors expect the 60 °C ranking "to hold true at 37 °C" (Figure 2b) R3.e–h·C20. Method also matters for PHPH: NMR gives about 71 wt% and HPLC about 78 wt% at 37 °C, and Table S1 lists 69.6–77.8 wt% for the nanoparticles R2·R11. Pelleted PHPH and MePHPH nanoparticles showed no crystals in polarized light (Figure S4) R10.c.

Figure 7 then sets these PVPVA contents beside the LoCs of the earlier study, with the drugs ordered by interaction strength: MePHPH (LoC 40%), MeTIPP (30%), TIPP (10%) and PHPH (5%), against about 5, 9, 36.5 and 73 wt% PVPVA R8. The two orders mirror each other exactly. The authors say the PVPVA content "trends with" the interaction, read the large polymer contents of PHPH and TIPP as further evidence of complex formation ("was further demonstrated"), and infer ("it is reasonable to infer") that favourable hydrogen bonding lowers the cost of drawing PVPVA into the insoluble phase, while the methoxy drugs, lacking hydrogen bonds, keep it out C26·C44·C46. In their terms, MePHPH and MeTIPP separate segregatively and PHPH and TIPP associatively, and "the associative systems were reported to have a lower LoC" C47·C67. Note by the AI writer: the PHPH bar in Figure 7 reads about 73 wt%, against about 77 wt% in Figure 2a; Figure 7 gives no temperature X3.

Above composition X the polymer flows into the ASD

Table B. PVPVA exchange between suspended ASD particles and solution (30 mg ASD in 15 mL pH 6.8 buffer, 37 °C). Values at 1–48 h read from Figures 3 and 4, approximate; the 0 h values are the nominal compositions R4·R5.

The 70% MePHPH particles lost almost all their polymer, as expected for a water-soluble polymer; the text gives "∼4%" at equilibrium, Figure 3 about 3.5 wt% at 24 h and 1.5 wt% at 48 h, and Table S1 3.9 ± 2.4 wt% measured directly R4·R11.g·X19. The 10% PHPH particles also lost polymer, but only down to about 75 wt% ("∼75%" in the results, "approximately 72%" in the discussion) R5.a·X4. The 50% PHPH particles did the opposite: they took up PVPVA until they held about 69 wt%, and the solution was "almost completely depleted" R5.c–d·C22. The 28% particles, chosen because 28% is the average drug load of the equilibrated 10 and 50% particles, gained only a little polymer from solution and released almost none into plain buffer R5.e–h·R15.

The authors read the end points as the hydrophobic binode in dilute solution: about 72 wt% PVPVA for PHPH and below 5 wt% for MePHPH, reached because the system lowers its free energy by moving PVPVA toward that composition C52·C53. For PHPH, composition X is about 28% drug; at X there is no driving force for the polymer to dissolve, and the slight uptake at 28% shows that X shifts a little when the total composition changes C54. At 50% the ASD is far richer in drug than X, so PVPVA flows in C55. They call the general point "an important conclusion": a water-soluble polymer does not diffuse out of an ASD whose polymer/drug ratio is below that of the equilibrium hydrophobic phase, which sets "a purely thermodynamic limit" on ASD dissolution, relevant above all to associative systems C56. The reversal itself is clear in the data, but it was shown for one associative system, PHPH, and drug release was not measured in these tests X15.

Converted to drug loads, the dilute hydrophobic phases lie at roughly 30% for PHPH, 65% for TIPP, above 90% for MeTIPP and above 95% for MePHPH, all well above the LoCs of 5, 10, 30 and 40% R14. The thermodynamic limit therefore "appears to occur at higher drug loading than the LoC", which the authors say "reinforces the concept" that only part of the ASD needs to turn into hydrophobic phase at the interface to stop release C57. Note by the AI writer: these rounded values differ slightly from the measured compositions, for example about 22–30% drug for PHPH and 92.5–94.7% by NMR for MePHPH X18.

Above the LoC, compact surfaces drift toward the drug-rich phase

Table C. Drug content of the dried compact surface (%), by ATR-FTIR, read from Figures 5 and 6 (approximate; n not stated) R6·R7.

At or below the LoC the surface stayed where it started: the 5% PHPH surface at about 6% and the 30% MePHPH surface at about 33–35%, apart from a reading of about 43% at 30 min. The authors call this "clearly observed" and explain it by mass balance: drug and polymer leave in the ASD ratio, so that ratio "must persist across the entire dissolution interface" R6.a·R7.a·C24·C58·X10. Above the LoC the surfaces changed. The 15% PHPH surface rose to about 21% drug at 30 min ("approximately 20%"), and the 50% MePHPH surface to about 90–101%, read as PVPVA leaving the surface. The 50% PHPH surface hardly changed in plain buffer, about 49 to 53%, read as PVPVA unable to leave ("indicating"); with 1 mg/mL PVPVA in the medium it fell to about 26% within 10 min, read as PVPVA entering the compact ("consistent with") R6.b–d·R7.b·C25.

The authors conclude that above the LoC the surface composition "evolves to be very similar" to the dilute hydrophobic binode, and that the barrier persisted once formed C59·C42. The match is closest for MePHPH (about 90–101% against a binode above 95% drug). For PHPH the 15% surface had reached about 21% at its last point, still rising, against a binode of about 22–30% drug, and the 50% surface reached the binode only when the medium supplied polymer X7. They resolve the puzzle that a surface should look like an equilibrium phase while the core is not at equilibrium by suggesting that "it is possible" for the outer gel layer, where water activity is close to one, to equilibrate with the solution, while deeper layers sit elsewhere on the diagram C61. A continuous hydrophobic phase that persists would keep its contents, while hydrophilic domains touching the solution dissolve away, which "may explain" how the hydrophobic phase comes to cover the surface C60.

What the controls show

The design holds the scaffold constant and changes only the substituents, so each methylated drug is the control for its parent. Removing the phenolic OH groups, PHPH to MePHPH and TIPP to MeTIPP, cuts the PVPVA content of the nanoparticles from about 77 to about 5 wt% and from about 36.5 to about 9 wt% at 60 °C, while the cited LoC rises from 5 to 40% and from 10 to 30% R3·R8. Within the ASD particle tests, MePHPH at 70% and PHPH at 10% both lose polymer while PHPH at 50% gains it, so the direction of polymer flux depends on drug load as well as on the drug R4·R5. The 28% PHPH test without added PVPVA is the null control for composition X R5.g–h.

Discussion

The authors' account of the LoC

The authors pull the results together in Section 5.3. The LoC "is due to" a hydrophobic barrier on the ASD surface, which "is postulated" to form by phase separation in the gel layer next to the dry core. The barrier forms when the hydrophobic phase outweighs the hydrophilic phase in volume and so becomes an interconnected network, and an ASD whose drug/polymer ratio exceeds that of the plait point "is reasoned" to give the larger hydrophobic phase C62. Strong drug–polymer interactions "are hypothesized" to lower the LoC because they move the plait point to a lower drug/polymer ratio: the hydrophobic binode then carries more polymer, the hydrophobic phase is larger at a given drug load, and an interconnected barrier appears earlier C63. The conclusions keep the hedges: a polymer-rich hydrophobic phase "is suggested" to favour the barrier, associative ASDs "need less drug" than segregative ones to reach the minimum hydrophobic volume, and "it is speculated herein" that barrier formation "may be closely related to the position of the plait point" C63. Separately, the polymer can leave the hydrophobic phase as the gel layer takes up water, but strong interactions "can impose a limit" to this: above a system-dependent drug load the driving force reverses C64.

What carries this account is a chain of three links. The first, that strong binders put much PVPVA into the drug-rich phase, is measured directly for four drugs (Table A). The second, that the phase holding more polymer occupies more volume at a given drug load and so connects earlier, follows from mass balance and the lever rule on a schematic diagram (Figures 8–10). The third, that a connected drug-rich phase is the release barrier, is taken from earlier work and from the surface measurements here, which show what the surface becomes but not how release responds C29·C62·X6.

What it suggests for formulation

The authors draw two practical points. A strongly interacting drug could reach a higher LoC with a polymer that does not partition into the hydrophobic phase, or with an additive that reduces that partitioning ("opens possibilities"; in the conclusions, "may open a path forward") C65. And because the hydrophobic volume depends on the plait point, the equilibrium phase diagram "may allow" an upper-limit, and "possibly" a direct, estimate of the LoC for PVPVA and related polymers C30·C66. Neither was tested here: no other polymer, no additive and no phase diagram appear in the paper.

Note by the AI writer (not discussed in the paper): the interaction that is often sought to stabilize an amorphous drug against crystallization is, for these analogues, the one that lowers the LoC. Two inexpensive measurements in this paper separated associative from segregative behaviour before any release test: the PVPVA content of drug nanoparticles made by solvent shift in polymer solution, and the change in amorphous solubility when PVPVA is added (a fall for PHPH, a rise for MePHPH). Whether either screens other drugs and polymers is untested.

What remains open

The authors flag the speed of equilibration. Their construction assumes that the two phases keep up with the incoming water; for PHPH and MePHPH this "may" hold, but for TIPP and MeTIPP, whose nanoparticles took more than an hour to equilibrate even at 60 °C, it is "unclear", and they suggest that a high-Tg hydrophobic phase "can kinetically trap the polymer" C38·C39·C40. They also note that the water content of the hydrophobic phase is unknown, so all compositions are on a dry basis C50. Beyond what the authors state, the plait-point positions themselves, the morphology of the gel layer and the effect of the barrier on release would all need measuring before the account could be used to predict a LoC X14·X6.

Limitations

Limitations stated by the authors. The phase diagram describes equilibrium, while the morphology of a dissolving interface also depends on kinetics, so the plait point is "only an approximation" of where the hydrophobic phase becomes continuous C30·C37. The local-equilibrium assumption is "unclear" for TIPP and MeTIPP C39. The MeTIPP nanoparticle value is only an upper bound because the drug crystallized C18. TIPP had to be studied at pH 2.5 to keep it unionized C11. The hydrophobic phase's water content is unknown C50.

The AI writer's own judgement (1): what is measured and what is argued. The paper measures polymer partitioning well. PHPH and MePHPH differ by an order of magnitude in nanoparticle PVPVA by two analytical methods, two preparation routes and two temperatures, and the four-drug ranking is the exact reverse of the cited LoC ranking R2·R3·R8. The mechanism that links partitioning to the LoC is argued, not measured: no ternary diagram, tie line or plait point was determined, and Figures 8–10 are hypothetical, one with horizontal tie lines X14. Release was not measured either; every LoC comes from the earlier study, and the compacts were dissolved only to expose their surfaces X6. The correlation rests on four close analogues and one polymer, so it shows a consistent ranking, not a predictive relationship.

The AI writer's own judgement (2): the details behind the numbers. Several values are stated differently in different places: the PHPH binode is about 77 (Figure 2a), about 73 (Figure 7), "nearly 75%" (text), 71.4–71.5 by NMR and 77.6–78.0 by HPLC (Table 2); the 10% PHPH end point is "∼75%" and "approximately 72%"; the 70% MePHPH end point is "∼4%", about 1.5 wt% in Figure 3 and 3.9 ± 2.4 wt% in Table S1 X3·X4·X19. Table S1 does not say which drug load its ASD-particle rows belong to X4. The TIPP and MeTIPP values come from 60 °C and were not fully converged, and the comparison with LoCs measured at 37 °C mixes conditions X12·X13. The surface values carry the offsets of their calibration fits, the MePHPH readings above 90% lie beyond the highest standard, the compacts were dried before analysis, and the 40 µg/mL PHPH spike is below PHPH's 200 µg/mL amorphous solubility in plain buffer X8·X17·X20. The amorphous-solubility stock concentrations look swapped, and the PVPVA infrared band is given as 1492 cm⁻¹ in the text and 1491 cm⁻¹ in the SI X1·X9. None of these changes the direction of any result, but they set how precisely the numbers can be used.

Conclusion

The paper answers part of its question with data and part with an argument. The data show that drug–PVPVA interaction strength sets where the polymer goes when the drug separates from water: about three-quarters of the PHPH nanoparticle phase is PVPVA, a twentieth of the MePHPH phase, with TIPP and MeTIPP in between, in exactly the reverse order of their LoCs R3·R8·C26·C67. For PHPH, PVPVA left low-loading ASD particles and entered high-loading ones, which defines a composition (about 28% drug) above which the polymer has no driving force to dissolve, a limit that lies above the LoC R5·R15·C56·C57. Above the LoC, compact surfaces moved toward the composition of the drug-rich phase R6·R7·C59.

The argument is that a polymer-rich drug phase is larger at a given drug load, so it connects into a surface barrier sooner; on a ternary diagram, strong interactions move the plait point to low drug content. The authors present this as a hypothesis ("are hypothesized", "it is speculated herein"), drawn on schematic diagrams, with release and phase diagrams left to other work C63·X14·X6. Measured ternary diagrams and plait points for these drugs, set against release from the same batches, would be the direct test.