Using a second polymer to keep celecoxib supersaturated after release from high-loading amorphous solid dispersions

Background and motivation

Many new drug candidates dissolve poorly in water, and making them amorphous is one way to raise their apparent solubility and dissolution rate; the higher free energy of the amorphous form "may" give higher apparent solubility, faster dissolution and better absorption, but the same free energy drives the drug back to the crystal and can remove the advantage (Hancock 2000, ref 2; Baird 2010, ref 3) C7. Polymers are mixed with the drug in amorphous solid dispersions (ASDs) to slow crystallization during processing and storage, and they "are thought to" do so by lowering molecular mobility through a higher glass transition temperature (Tg) and by hydrogen bonding with the drug (refs 5–7) C8.

Stability in the dry solid is only half of the problem. The authors point out that dissolution tests are often run under sink conditions, which are unlikely in the gut for many poorly soluble drugs, and that what an ASD achieves in solution depends on the amount added, the drug loading, how fast each component dissolves, the crystal solubility, the supersaturation reached and how fast the drug crystallises C9. Earlier work from the same group (Alonzo 2010, ref 8) had shown two ways in which an amorphous drug can crystallise while it dissolves. Water absorbed into the solid lowers its Tg, so crystals can nucleate and grow on the wetted matrix; in that case only a small supersaturation is ever reached. If the drug dissolves faster than the matrix crystallises, it can instead crystallise later from the supersaturated solution, and the more supersaturated the solution, the more prone it is to do so. Polymers dissolved in the medium "may" inhibit either route for a pure amorphous drug, but how the polymer acts during the dissolution of an ASD was little known C10·C31.

A second line of earlier work showed that a good inhibitor in the solid may be a poor one in solution. Polyacrylic acid strongly inhibited crystal growth of acetaminophen from the supercooled liquid, and so slowed crystallization in ASDs, but did nothing against nucleation or growth from aqueous solution (Trasi 2012, ref 13; Trasi 2015, ref 14); felodipine and other systems behaved similarly (refs 6, 15–17). The authors therefore suggest that combinations of polymers "may be appropriate" to secure both storage stability and dissolution performance (Marks 2014, ref 18) C28·C29. Ternary dispersions had been used to raise dissolution rate (refs 19, 20) or solid-state stability (ref 21), but there was little work on how a pair of polymers affects crystallization during ASD dissolution C30.

The paper sets out to do three things with celecoxib (CEX), a poorly soluble anti-inflammatory drug: measure how readily the pure amorphous form crystallises and by which route; compare polyvinylpyrrolidone (PVP, two grades), hydroxypropylmethyl cellulose (HPMC) and HPMC acetate succinate (HPMCAS), alone and in pairs, for their effect on the route and rate of crystallization and on the release of CEX from ASDs; and test whether polymer pairs help ASDs at high drug loading, which is why every dispersion contains 50% drug A1·A2·A4·A6.

Background concepts

Supersaturation and the two solubilities. A solution is supersaturated when its drug concentration exceeds the solubility of the crystal. For CEX in pH 6.8, 100 mM sodium phosphate buffer (SPB) at 37 °C, the crystal solubility measured here is 1.5 ± 0.1 µg/mL R11.a. The upper limit is set by the amorphous solubility, the highest concentration of free drug an amorphous form can give; the authors take about 22 µg/mL from earlier calculations (Ilevbare 2013, ref 10; Abu-Diak 2011, ref 11) and treat it as the maximum free-drug concentration an ASD can produce C11·M1. The degree of supersaturation $S$ is defined through the chemical potential of the dissolved drug (Eq. 4):

Here $\mu$ is the chemical potential of the solute, ${\mu }^{*}$ its chemical potential in a saturated solution, $R$ the gas constant and $T$ the temperature. In plain terms: supersaturation is how far the dissolved drug sits above the energy of a saturated solution. In dilute solution the authors write it as the concentration ratio $S=c/c^{*}$, where $c$ is the concentration in the supersaturated solution and $c^{*}$ the crystal solubility E4·E5. The paper prints "InS" for $\ln S$ and drops the asterisk from the second $c$; both are misprints X2.

Non-sink dissolution. The dissolution tests use 8.8 mg CEX, alone or in 17.6 mg of a 50% ASD, in 400 mL buffer, so that full release gives 22 µg/mL, the reported amorphous solubility P5. The authors call this at-sink with respect to amorphous solubility and non-sink with respect to crystal solubility: an ASD can release all its drug, but the solution it makes is about 15 times supersaturated with respect to the crystal (22/1.5, a calculation by the AI writer) C17·R11.a·X9. Whether it stays supersaturated depends on crystallization.

Matrix and solution crystallization. The paper keeps two routes apart. Matrix crystallization happens in or on the wetted solid; crystals formed there act as seeds that grow quickly and consume the supersaturation, so it limits how high the concentration can rise. Solution crystallization happens after release, by nucleation and growth in the supersaturated liquid, and it sets how long the supersaturation lasts; once nucleation starts, the concentration falls shortly afterwards (desupersaturation) C31. The authors' framing is that matrix crystallization should ideally be prevented, so that a high supersaturation is reached at all C31.

Induction time and classical nucleation theory. The paper measures solution crystallization through the nucleation induction time, the delay between creating a supersaturated solution and detecting crystals. It is split into two parts (Eq. 1):

Here $t_{ind}$ is the measured induction time, $t_n$ the time for a critical nucleus to form and $t_g$ the time for it to grow to a detectable size. In plain terms: what is measured is the time to form a nucleus plus the time to see it E1. If nucleation reaches steady state quickly and $t_n$ is much longer than $t_g$, the induction time is inversely proportional to the nucleation rate $J$ (Eq. 2) E2·C41:

In plain terms: a longer induction time means slower nucleation. Classical nucleation theory (CNT) gives the nucleation rate as (Eq. 3):

Here $\gamma$ is the interfacial tension between nucleus and solution, $v$ the molecular volume of the crystallizing solute, $k$ the Boltzmann constant, $T$ the temperature and $S$ the supersaturation; the pre-exponential factor $A$ is not defined in the paper E3·X2. In plain terms: nucleation slows very steeply when the interfacial tension rises or the supersaturation falls, and at constant temperature these are the only two handles the theory offers. The authors use this equation to ask whether polymers slow nucleation by changing $S$ or $\gamma$ C42·C43.

Two-step nucleation. Because CNT gives no clear answer (see Discussion), the authors turn to computer simulations in which nucleation proceeds in two steps: a dense, disordered liquid cluster of solute forms first and then rearranges into an ordered nucleus. Additives are "proposed" to interfere with the rearrangement, depending on their affinity for the cluster relative to the solvent and on geometry (Anwar 2009, ref 25) C45.

Methods in detail

The study has three parts that are run side by side and then compared: a solution test that ranks the polymers as nucleation inhibitors; slurry and microscopy tests that show where amorphous CEX crystallises; and non-sink dissolution tests of eight ASDs and of neat CEX. The figure below shows the workflow.

Materials. CEX came from Attix Pharmaceuticals. PVP was used in two grades: K12, with a weight-average molecular weight of 2000–3000 g/mol (BASF), and K29/32, with 58,000 g/mol (ISP Technologies). HPMCAS (type AS-MF) and HPMC (grade 606) came from Shin-Etsu; their molecular weights are not given M1·M2·M3·M4·M5. Figure 1 draws the four structures; the cellulosic polymers are shown with their substituents: methyl and hydroxypropyl groups and, for HPMCAS, also acetyl and succinoyl groups M4·M5.

Making the solids. ASDs were made by dissolving CEX and polymer(s) in 50:50 ethanol:dichloromethane, removing the solvent by rotary evaporation and drying overnight under vacuum. Table I lists eight compositions: four binary ASDs at CEX:polymer 5:5 by weight (PVP-K12, PVP-K29/32, HPMC, HPMCAS) and four ternary ASDs at CEX:PVP:cellulosic polymer 5:4:1, that is, with one fifth of the PVP replaced by HPMC or HPMCAS P1. Pure amorphous CEX was made by melting the crystal at about 180 °C on aluminium foil and quench cooling; HPLC showed no degradation P2·R12. All solids were ground, sieved to 106–250 µm, stored over phosphorus pentoxide and checked as amorphous by powder X-ray diffraction (patterns not shown) P1·R13.

Contribution 1: Ranking the polymers as nucleation inhibitors in solution

Problem. To separate the polymer's effect in solution from its effect in the solid, the authors first need a test in which no solid is present: a supersaturated CEX solution with and without dissolved polymer.

How. 275 µL of a 4 mg/mL CEX solution in methanol was pipetted into 50 mL of pH 6.8, 100 mM SPB at 37 °C, stirred at 300 rpm, giving an initial concentration of 22 µg/mL, about the amorphous solubility P4·C11. The buffer held no polymer, 5 µg/mL of one polymer, or 5 µg/mL PVP-K12 or PVP-K29/32 plus 5 µg/mL HPMCAS P4. A fibre optic UV probe recorded full spectra every minute for 8 h: absorbance at 249 nm gave the concentration, and extinction at 350 nm, where CEX does not absorb, gave light scattering. The onset of crystallization was taken as a sudden rise in scattering together with a rapid fall in absorbance (Chen 2015, ref 12), and every condition was run in triplicate P4.

Why it works. Starting from a clear solution at the amorphous solubility removes the solid matrix from the question; any difference between polymers is a difference in nucleation and growth from solution. The scattering channel detects particles before the concentration drop becomes large. Difference from earlier work. The same group had measured induction times with many polymers (Ilevbare 2013, ref 10); here the test serves as a reference against which the ASD dissolution results are read C48.

Contribution 2: Watching where amorphous CEX crystallises

Problem. Matrix crystallization happens on the wetted solid and is not visible in a concentration curve; it has to be detected in the solid itself.

How. In the slurry test, 200 mg of ASD was stirred in 2 mL of buffer at 37 °C and followed by in situ Raman spectroscopy (785 nm laser) every 15 min for 7 h P3. Figure 3 gives the reference spectra: crystalline CEX has a peak at 1614 cm⁻¹ with a small shoulder at 1596 cm⁻¹, while amorphous CEX has a peak at 1611 cm⁻¹ with a much stronger shoulder near 1598 cm⁻¹ R3. The authors follow the 1598 cm⁻¹ shoulder; its loss marks crystallization C14. Pure amorphous CEX was also slurried, alone and with 22 µg/mL polymer dissolved in the buffer, but those spectra are not shown R4.f·R4.g. In the microscopy test, amorphous CEX particles on a depression slide were covered with 3–4 drops of buffer, with or without 5 µg/mL HPMC or HPMCAS, and imaged under polarized light, where crystals appear bright P7.

Why it works. Raman spectroscopy sees the solid in the slurry, so it reports crystals in or on the particles; polarized light microscopy separates crystals on the particles from crystals in the surrounding liquid. The authors note that Raman alone cannot tell which route the crystals took C38.

Contribution 3: Non-sink dissolution of binary and ternary ASDs

Problem. The practical question is whether an ASD releases its drug fast and keeps it in solution; this needs a test at a dose where full release produces a strongly supersaturated solution.

How. 17.6 mg of ASD, containing 8.8 mg CEX, was added to 400 mL of buffer at 37 °C and 300 rpm, so that full release gives 22 µg/mL; neat crystalline and amorphous CEX were tested at 8.8 mg P5. A fibre optic UV probe recorded spectra every minute for 16 h, and each condition was run in triplicate P5. In two variations the medium held 5 µg/mL HPMCAS or HPMC before the CEX:PVP-K12 ASD or the amorphous CEX was added P5·A5. Crystal solubility was measured separately by shaking excess crystalline CEX for 48 h at 37 °C in buffer with or without 22 µg/mL of each polymer, ultracentrifuging and analysing by HPLC P6.

Why it works. Because the dose matches the amorphous solubility, a curve that reaches 21–22 µg/mL means the whole dose dissolved, and any later fall means crystals have formed R5·C17. Comparing binaries, ternaries and polymer-in-medium tests in one setup isolates what the second polymer does. Difference from earlier work. Earlier ternary dispersions were aimed at dissolution rate or storage stability; here the minor polymer has one job, to stop crystallization after release C30·C6.

Experimental analysis

Result 1: the cellulosic polymers hold a supersaturated solution; PVP only delays nucleation. Without polymer, the 22 µg/mL solution began to crystallise within 5 min; the concentration fell to about 4 µg/mL within an hour and the solution became turbid with visible crystals R1·R2.a. With 5 µg/mL PVP-K12 the onset came at about 1 h, and with PVP-K29/32 at about 2 h. With HPMCAS or HPMC there was no substantial nucleation or growth, and the starting supersaturation lasted beyond the 8 h of the test; the same held for HPMCAS combined with either PVP grade (Figure 2) R1·R2. After nucleation the concentration fell more slowly with PVP-K29/32 than with PVP-K12, and more slowly with either than without polymer C12. The authors read the PVP + HPMCAS results as showing that PVP does not impair HPMCAS ("indicating") C13.

Table A. Induction times from the text; concentrations read from Figure 2 (approximate), CEX starting at 22 µg/mL in pH 6.8 buffer at 37 °C R1·R2.

Result 2: neat amorphous CEX crystallises on its own particles and gives almost no supersaturation. Under polarized light, amorphous particles in plain buffer showed bright spots after 2 min; after 10 min the particles were widely crystalline and needle-like crystals had grown in the surrounding liquid (Figure 9) R8. With 5 µg/mL HPMC (Figure 10) or HPMCAS (Figure 11) in the buffer, no crystals formed in the liquid, but small crystalline regions still appeared on the particles and grew by 60 min, which the authors describe as matrix crystallization at "a much slower rate" R9·R10·C26. In the slurry, crystallization of pure amorphous CEX was complete within an hour (spectra not shown) R4.f. The dissolution curves agree. Amorphous CEX released faster at first, about 0.49 µg/mL at 20 min against about 0.13 for the crystal, but the two then ran together and ended at about 1.70 and 1.85 µg/mL after 16 h (Figure 8) R7.a·R7.b. With 5 µg/mL HPMCAS or HPMC in the buffer, amorphous CEX reached about 3.05 and 3.50 µg/mL, which the authors call "only slightly higher" R7.c·R7.d·C25.

The authors conclude that pure amorphous CEX crystallises rapidly and "primarily via the matrix route", which they attribute to water lowering the Tg and raising molecular mobility (not measured here) C1·C32. The low concentrations "indicate either" that matrix crystallization is rapid and complete or that the seed crystals formed grow fast enough to consume whatever the residual amorphous material dissolves C34. Dissolved polymers are "clearly unable" to stop crystals forming in the amorphous matrix, and the observations "confirm the need" to prevent matrix crystallization if supersaturation is to be generated at all C35·C36. Whether 22 µg/mL dissolved polymer slowed crystallization in the slurry is stated two ways: "not substantially" retarded in the Results and extended time to complete crystallization in the Discussion; neither set of data is shown R4.g·C16·C33·X7.

Result 3: in the ASDs the polymer sets both release speed and how long supersaturation lasts. All eight ASDs produced supersaturated solutions; those without HPMC reached 21–22 µg/mL, the full dose R5·C60. The two PVP ASDs released fastest and at almost the same initial rate, about 16 and 17 µg/mL at 2 h and about 20 µg/mL at 4 h R6.a·R6.b·C20. CEX:PVP-K12 peaked at about 21.0 µg/mL near 6 h and began to fall about 2 h after release was complete; by 16 h it was down to about 6.5 µg/mL, with wide error bars during the decline. CEX:PVP-K29/32 reached about 22.0 µg/mL near 10–11 h, held its peak about twice as long as PVP-K12, and was about 19.7 µg/mL at 16 h (Figure 5) R6.a·R6.b·C20. CEX:HPMCAS released more slowly, about 9.0 µg/mL at 2 h, but kept rising to about 21.1 µg/mL without falling; CEX:HPMC was slowest and reached only about 16.4 µg/mL at 16 h, still rising R6.c·R6.d·C21. In the slurry the PVP-K12 ASD crystallised within 4 h, while the PVP-K29/32 and HPMCAS ASDs stayed amorphous for 7 h; the authors read this as better resistance with the higher PVP grade R4.a·R4.c·R4.d·C15·C37.

Table B. CEX in solution (µg/mL) from 17.6 mg ASD in 400 mL buffer, 37 °C; all values read from Figures 5, 6 and 7 (approximate), means of three runs; full release is 22 µg/mL R6.

Result 4: one fifth of HPMCAS or HPMC in place of PVP stops the fall but slows release. None of the four ternary ASDs desupersaturated within 16 h; they ended at about 16.7–21.9 µg/mL, against about 6.5 µg/mL for CEX:PVP-K12 (Figure 6, Table B) R6.a·R6.e–h. In the slurry, both PVP/HPMCAS ternaries stayed amorphous for 7 h, whereas the PVP-K12 binary had crystallised by 4 h; the HPMC-containing ASDs were not reported R4.a·R4.b·R4.e·X11. The cost is release rate. Replacing 20% of the PVP-K12 with HPMCAS made release "much slower", close to the HPMCAS binary: about 10.1 against 9.0 µg/mL at 2 h and 14.6 against 14.5 at 4 h; the PVP-K29/32/HPMCAS ternary behaved similarly R6.c·R6.e·R6.f·C22. With HPMC the PVP grade mattered: the PVP-K12/HPMC ternary was "much slower" (about 6.7 µg/mL at 2 h and 16.7 at 16 h) than the PVP-K29/32/HPMC ternary (about 12.4 and 20.8 µg/mL), a difference not seen with HPMCAS R6.g·R6.h·C23. The PVP-K29/32/HPMC ternary did release faster than the CEX:HPMC binary, which the authors call enhanced release C55.

Result 5: the inhibitor also works from the medium. When the buffer already held 5 µg/mL HPMCAS or HPMC, the CEX:PVP-K12 ASD released slightly more slowly than in plain buffer, about 13.6 and 13.5 µg/mL at 2 h against 16.1, but did not desupersaturate, ending at about 22.5 and 21.0 µg/mL after 16 h (Figure 7) R6.a·R6.i·R6.j·C24. At 2 h the ASD in polymer-containing buffer was ahead of the PVP-K12/HPMCAS ternary (about 10.1 µg/mL) R6.e·R6.i. Note by the AI writer (not discussed in the paper): if a 5:4:1 ternary dissolves fully at this dose, it supplies about 4.4 µg/mL of cellulosic polymer, close to the 5 µg/mL used in the medium, so the two routes deliver a similar inhibitor level to the solution X8.

Table C. Crystalline CEX solubility after 48 h at 37 °C in pH 6.8 buffer, mean ± SD of three (Table II); the supersaturation ratio S = 22/solubility is calculated by the AI writer, not by the paper R11·E5·X9.

Side result: polymers barely change the crystal solubility. At 22 µg/mL, the polymers raised the crystal solubility from 1.5 to 1.89–2.14 µg/mL, which the authors describe as no substantial change (Table II) R11·C27. They use this to argue that the supersaturation is the same in all polymer-containing solutions C43. Note by the AI writer: Table C shows the ratio is indeed similar among the polymers (about 10–12), but about 1.3–1.4 times higher without polymer, and the solubilities were measured at 22 µg/mL rather than the 5 µg/mL of the induction tests X9·X8.

Discussion

How the authors put the results together. The argument runs from the neat drug to the dispersions. Neat amorphous CEX crystallises on contact with buffer, so a polymer in the medium cannot rescue it; the polymer has to be in the solid C1·C35·C36. In the ASDs, even at 50% loading, CEX is much more stable on exposure to buffer C37. Because the ASDs released completely, the authors take matrix crystallization as "completely inhibited" ("it is apparent") and attribute the remaining differences to crystallization from solution C39. Growth needs nuclei, so the polymer's effect on nucleation is "of paramount importance", and the induction time is its measure C40·C41. The induction times are "quite predictive" of the dissolution results: PVP-K12 is the weakest nucleation inhibitor and its ASD is the first to desupersaturate C48.

The chain as the authors give it, drawn by the AI writer from claims C39, C42, C48, C49 and C57 C39·C42·C48·C49·C57.

Why the cellulosic polymers work better. By classical nucleation theory, a slower nucleation rate at constant temperature needs a lower supersaturation or a higher interfacial tension C42. The authors rule out the first: with the polymer barely changing the crystal solubility, they consider the supersaturation the same in all polymer-containing solutions; their text cites "Table I" here, but the solubilities are in Table II C43·X1. They rule out the second as well: a polymer adsorbing on the nucleus would lower the interfacial tension and speed nucleation, the opposite of what is seen (they cite "Eq. 2"; the interfacial tension appears in Eq. 3) C43·X1. They conclude that CNT "does not readily allow for obvious mechanistic insights" C44. Their proposal draws on the two-step picture: the cellulose derivatives "may" interact more strongly with dense liquid clusters of CEX, whereas the more hydrophilic PVP also interacts with the solvent and competes less well for the clusters C45·C46. This is in line with earlier studies in which amphiphilic polymers, including many cellulose derivatives, extended the induction times of lipophilic drugs more than hydrophilic polymers did (Ilevbare 2013, ref 10; Warren 2010, ref 26) C47. No cluster or polymer–cluster interaction was measured in this paper C46.

The ternary strategy. The authors' central formulation idea is to combine a polymer that releases the drug fast but inhibits solution crystallization poorly, here PVP, with a small amount of a good inhibitor C51. In their words, the ternaries show that the higher tendency of PVP-K12 ASDs to crystallise from solution "can be mitigated" by a small amount of HPMCAS or HPMC, and they infer that the cellulose polymers "dissolve sufficiently rapidly" from the ternary to act in solution C49·C50. They regard the induction-time results as confirming that PVP does not impair the cellulosic inhibitor, and they note that only very low polymer concentrations, 5 µg/mL, are needed C52·C53. They also flag the cost: relative to CEX:PVP-K12, release is "somewhat surprisingly" delayed when a small amount of HPMCAS or HPMC is added to the matrix, which "clearly may not be desirable" C56. The polymer-in-medium experiment suggests a way around this: the inhibitor "need not be present in the dispersion itself" and "could be potentially" added to the formulation as a separate component C57·C58. The authors end with a recommendation that using very hydrophilic polymers for rapid release together with a small amount of an effective inhibitor "should be investigated further" C59.

Where the approach applies. All results concern one drug, CEX, at one loading (50%), one dose (17.6 mg ASD in 400 mL, giving 22 µg/mL at full release) and one medium (pH 6.8, 100 mM phosphate buffer, 37 °C, 300 rpm), followed for 16 h P5. The polymer levels differ between tests: 5 µg/mL in the induction, microscopy and polymer-in-medium tests, 22 µg/mL in the solubility test and in the unshown slurry test with dissolved polymer X8. The comparison between solution and ASD results is therefore cleanest for the 5:4:1 ternaries, whose full release would supply about 4.4 µg/mL cellulosic polymer (calculation by the AI writer) X8.

What it suggests for formulation. Note by the AI writer (not discussed in the paper): the results separate two jobs a polymer does in an ASD. The polymer inside the dispersion must stop the wetted solid from crystallising, because a dissolved polymer could not do that for neat amorphous CEX R8·R9·R10. A polymer in the solution must slow nucleation after release; at 5 µg/mL the cellulosic polymers did this for 8–16 h, whereas PVP gave only 1–2 h R1·R6. At this loading, PVP alone did the first job but not the second, and adding 10% of a cellulosic polymer to the ASD did both but slowed release; dosing the inhibitor outside the ASD protected the solution with a smaller loss of release rate, although only as pre-dissolved polymer R6·X15.

Limitations

Limitations stated by the authors. The authors acknowledge that Raman spectroscopy alone does not identify the route by which the PVP-K12 ASD crystallised in the slurry C38. They describe the slower release of the ternaries relative to CEX:PVP-K12 as "somewhat surprising" and as something that "clearly may not be desirable" C56. They state that CNT does not readily explain how polymers extend induction times, and they present the cluster mechanism with "may therefore relate" C44·C46. The in vivo usefulness of the strategy is offered as "might be useful", and the separate-component idea as "could be potentially" C51·C58.

The AI writer's own judgement (1): statements that go further than the data. The inference that matrix crystallization was "completely inhibited" in all ASDs rests on complete release to 21–22 µg/mL, but the HPMC binary and the PVP-K12/HPMC ternary reached only about 16.4 and 16.7 µg/mL in 16 h, the PVP-K29/32/HPMC ternary about 20.8, and none of the HPMC-containing ASDs was examined in the slurry C39·R6.d·R6.g·R6.h·X3·X11. The claim that the induction experiments confirm that neither HPMCAS nor HPMC is impaired by PVP holds for HPMCAS only; no PVP + HPMC solution was tested C52·X4. The abstract says the PVP/cellulosic combination "enabled improved drug release", while the discussion reports delayed release relative to PVP alone; the ternaries were ahead of PVP-K12 only at late times, after its desupersaturation, and faster than a cellulosic binary only for PVP-K29/32/HPMC against CEX:HPMC C4·C55·C56·X5. The abstract also links crystallization inhibition to CEX being "molecularly mixed" with the polymer, but miscibility of the 50% ASDs was not characterised, and the PVP-K12 ASD did crystallise in the slurry C2·X6. Storage stability, named in the conclusions as a property a second polymer could provide, was not studied C6·X10. The separate-component strategy was tested only with polymer already dissolved in the medium C58·X15.

The AI writer's own judgement (2): smaller points in the data. The solubility rise with polymer, from 1.5 to 1.89–2.14 µg/mL (26–43%), is larger than "not substantially" suggests, so the supersaturation was similar among polymer solutions but higher without polymer C27·X9. Late in the UV tests, crystalline CEX reached about 1.85 µg/mL and was still rising, and the no-polymer induction solution levelled at about 3.1 µg/mL, both above the HPLC solubility of 1.5 µg/mL; the two methods are not directly comparable R7.a·R2.a·X12. Two statements about the slurry with dissolved polymer disagree and rest on data not shown X7. Some microscopy panels show a different field from the others in the same figure (Figure 10c, Figure 11a), so the same particles are not followed through time X13. The CEX:PVP-K12 dissolution curve in Figures 5, 6 and 7 is one data set shown three times, not three experiments X14. All values in this article other than Table II were read from plots and are approximate.

Conclusion

The paper makes three points. First, as nucleation inhibitors in solution, HPMCAS and HPMC far outperform PVP: at 5 µg/mL they held 22 µg/mL CEX for more than 8 h, against about 1 h for PVP-K12 and 2 h for PVP-K29/32 R1·C3. Second, neat amorphous CEX crystallises on its own particles within minutes and generates almost no supersaturation, and dissolved polymer cannot prevent this; in a dispersion, the polymer keeps the solid amorphous long enough for release R7·R8·C35·C37. Third, at 50% loading, the binary ASDs force a choice between fast release that collapses (PVP-K12) and stable but slow release (HPMCAS, HPMC), while 5:4:1 ternaries, or 5 µg/mL of cellulosic polymer in the medium, kept the solution supersaturated for 16 h; the ternaries released more slowly than PVP alone R6·C22·C24·C56. In the authors' words, this "shows the feasibility" of ASDs in which one polymer provides release and a minor second polymer prevents crystallization C6.

The results come from one drug, one loading and one dose in buffer, and several links in the argument were inferred rather than measured: how fast the minor polymer leaves the ternary ASD, how the cellulosic polymers block nucleation, and whether a solid inhibitor dosed beside the ASD would act as fast as a pre-dissolved one C50·C46·X15. Testing these, and the trade-off between release rate and inhibitor level, would show how far the two-polymer approach can be taken.