Fast release without losing the drug to crystals: pairing polyacrylic acid with a cellulose inhibitor in high-loading celecoxib dispersions

Background and motivation

Many drug candidates are so poorly soluble in water that they need an enabling formulation. One route adds solubilizers, such as surfactant micelles or cyclodextrins, which raise the total amount dissolved but leave the free drug concentration limited by the crystal. The other route is the amorphous solid dispersion (ASD): the drug is dispersed in a polymer in the amorphous state and, on dissolution, produces a supersaturated solution whose free concentration exceeds that of the crystal. Because passive absorption across the gut wall "is expected" to scale with the free concentration, the degree and duration of supersaturation matter (Introduction; Brouwers 2009, ref 3; Raina 2014, ref 4; Dahan 2012, ref 5). C2

Supersaturation is metastable, and an amorphous formulation can lose drug to crystals in two ways: crystals can form directly in the wetted solid (matrix crystallization) or from the supersaturated solution after the drug has dissolved (solution crystallization). Polymers can slow both, and the authors argue that a formulator therefore needs to know how a polymer acts on each route and on nucleation and growth separately (Alonzo 2010, ref 11). C3

A second constraint is drug loading. Higuchi and co-workers showed for sulfathiazole–PVP dispersions that release slows sharply as the drug fraction rises, even without crystallization. Corrigan explained this with a model in which, at low loading, drug and polymer dissolve together at the polymer's rate, whereas at high loading the polymer may dissolve faster and leave a drug-rich layer behind, so release becomes controlled by the drug (refs 8, 16, 17). Yet low drug:polymer ratios mean large or multiple dosage units, so there is a practical need to raise loading while keeping the drug amorphous, keeping the solution supersaturated and still releasing fast, especially for indications such as pain where rapid onset matters. C4·C5

Celecoxib, a cyclooxygenase-2 inhibitor used mainly for arthritis pain, is a convenient test case. It crystallizes rapidly from water when no polymer, or a poor inhibitor, is present, and its release from ASDs slows as loading rises (Gupta 2004; Abu-Diak 2011). Albers et al. achieved fast release at 50% loading from a polymethacrylate extrudate, but the supersaturated solution crystallized in less than ten minutes. In a clinical study, Morgen et al. found that celecoxib nanoparticles made with ethylcellulose and casein were absorbed much faster than a 50:50 celecoxib–HPMCAS spray-dried dispersion. C6

The authors' idea is to split the two jobs between two polymers. PAA, a highly water-soluble, weakly acidic polymer, should give fast release but is known to be a poor inhibitor of solution crystallization; HPMC and HPMCAS are widely used ASD polymers and good crystallization inhibitors for many drugs (refs 22–28). The paper sets out to combine them in one dispersion so that a high-loading celecoxib ASD releases quickly and stays supersaturated under non-sink conditions, and to measure the polymers' effects on celecoxib nucleation and growth so that the release results can be explained. C7

Background concepts

Supersaturation. The driving force for crystallization from solution is the supersaturation ratio, defined in the paper as

In plain terms: $S$ compares the concentration $c$ of dissolved drug with the concentration $c^{*}$ of a solution saturated with respect to the crystal; the larger $S$, the faster crystals nucleate. The paper does not print a value, but with its own numbers, 22 µg/mL against a crystalline solubility of 1.5 µg/mL, $S$ is about 15 (calculation by the AI writer). E3·R1

Amorphous solubility. An amorphous drug can raise the free concentration only up to its "amorphous solubility", the highest free concentration achievable; for celecoxib the paper cites a value of about 22 µg/mL from earlier experimental and calculated work (refs 19, 20, 29, 31). The authors therefore treat 22 µg/mL as the maximum possible supersaturation, and they use the same value as the dose in their dissolution tests. M1·C10

Induction time. In an unseeded solution, crystals are first detected after an induction time, which the paper splits into the time needed to form critical nuclei and the time needed for them to grow to a detectable size:

In plain terms: the measured induction time $t_{\mathrm{i}\mathrm{n}\mathrm{d}}$ is the true nucleation time $t_n$ plus the growth time $t_g$; a polymer can lengthen it by slowing either step. The onset is read from a sudden rise in light scattering at 350 nm, where celecoxib does not absorb, together with a fall in absorbance at 249 nm, where it does. E4·C9

Seeded growth. When crystals are added to a supersaturated solution, nucleation is no longer needed and the fall in concentration measures growth. Comparing the slope of this fall with and without polymer gives a relative growth rate (Eq. 5, below). Combining it with induction times gives a rough estimate of how strongly the polymer slows nucleation (Eq. 6).

Polymer-controlled and drug-controlled release. Corrigan's model, quoted in the introduction, links the dissolution rates of two components A and B of a dispersion:

In plain terms: component B dissolves at the rate $G_A$ of component A scaled by the ratio of their proportions $N$, that is, B is carried out by A. This holds when

In plain terms: A must be in sufficient excess relative to the product of diffusion coefficient $D$ and solubility $C_S$ of each component. With the polymer as A, this describes polymer-controlled release at low drug loading; when the condition fails at high loading, release becomes controlled by the drug. The paper uses the model as background only and does not fit it. E1·E2·C4

Non-sink test at the amorphous solubility. Release tests were run with a dose that gives 22 µg/mL if fully dissolved, the reported amorphous solubility. This is far above the crystalline solubility, so the solution becomes supersaturated, but it is "at sink" with respect to the amorphous drug. The design has a useful consequence the authors point out: an ASD that stays amorphous should release everything, whereas any drug that crystallizes in the particle cannot dissolve afterwards, because the solution is already supersaturated with respect to the crystal. Incomplete release thus signals crystallization. C8·P5

Polymer miscibility. Two polymers mix at the molecular level only if mixing is favoured by specific interactions, because the entropy gained by mixing two large molecules is small (refs 46, 47). In infrared (IR) spectra, a blend of immiscible polymers shows the sum of the two spectra, whereas specific interactions, such as new hydrogen bonds, shift or create bands. Pure PAA has a carbonyl band at 1710 cm^-1, which the paper assigns to hydrogen-bonded carboxylic acid dimers (refs 22, 45); the authors read a new band near 1735 cm^-1 in blends as a sign that these dimers are disrupted. C58·C63·R20·R21

Methods in detail

The study has three parts: solution experiments that separate the polymers' effects on solubility, nucleation and growth; release experiments designed so that crystallization shows up as missing drug; and miscibility measurements that try to explain why the two cellulose polymers behave differently in combination with PAA. The method flow is:

Materials and preparation. Celecoxib came from Attix Pharmaceuticals, PAA from Sigma-Aldrich (molecular weight not given), HPMCAS (type AS-MF) and HPMC (606 grade) from Shin-Etsu (Figure 1). Weighed drug and polymer(s) were dissolved in ethanol:dichloromethane 50:50 v/v, the solvent was removed by rotary evaporation, the solid dried overnight under vacuum, ground and sieved to 106–250 µm. The 10% HPMCAS dispersion (CEX:HPMCAS 1:9) was cryo-milled instead. All dispersions were amorphous by powder X-ray diffraction before use and were stored over phosphorus pentoxide. All aqueous work used pH 6.8, 100 mM sodium phosphate buffer (SPB) at 37 °C. M1–M5·P1·P2·R6

Contribution 1: separating the polymers' effects on solubility, nucleation and growth

① The problem. A polymer can keep a solution supersaturated for three different reasons: it may solubilize the drug, lowering $S$; it may slow nucleation; or it may slow growth of crystals that already exist. These matter differently in a dissolving ASD, where crystals may come from the matrix.

② What they did. First, crystalline solubility was measured with and without 22 µg/mL of each polymer (excess crystals, 20 mL buffer, 48 h, ultracentrifugation, HPLC, triplicate). Second, unseeded induction times: 275 µL of a 4 or 2 mg/mL methanol stock was added to 50 mL buffer (0.55% methanol), giving 22 or 11 µg/mL celecoxib, with or without pre-dissolved polymer; the figure legend gives 5 µg/mL of each polymer. Concentration and scattering were recorded every minute for 8 h without polymer and 16 h with polymer, in triplicate. Third, seeded desupersaturation: 7 mg of celecoxib crystals were suspended in 50 mL buffer with 0, 5, 11 or 22 µg/mL HPMCAS, and the drug stock was then added to give 22 or 11 µg/mL. P3·P4·P7

The seeded curves give a relative growth rate,

In plain terms: $R_0$ is the mass growth rate of the seeds without polymer and $R_p$ the rate with polymer, both taken from the slopes of the concentration–time curves; the ratio says how many times slower the seeds grow with the polymer. The authors took $R_0$ from the linear part of the first 5 min without HPMCAS and, "for simplicity", $R_p$ from the slope over the whole 16 h with HPMCAS. E5·C38

The growth ratio is then combined with induction times to estimate the effect on nucleation (refs 39, 42):

In plain terms: $J_0$ and $J_p$ are the nucleation rates without and with polymer, and $t_{u,0}$ and $t_{u,p}$ the corresponding induction times. A polymer that lengthens the induction time a lot but slows growth only a little must be slowing nucleation strongly; the powers 3 and 4 come from the cited model. For 22 µg/mL celecoxib with 5 µg/mL HPMCAS no induction was seen within the experiment, so the authors set $t_{u,p}$ = 960 min, which underestimates it. E6·C39

③ Why it works. Measuring solubility first rules out solubilization as the cause of longer induction times. Seeded experiments isolate growth, which is what happens if crystals made in a dissolving particle reach the medium. ④ Difference from earlier work. The approach follows the group's earlier felodipine and ritonavir work (refs 36, 39, 40); here it is applied to celecoxib so that the release experiments can be interpreted.

Contribution 2: release tests that expose matrix crystallization

① The problem. When release from an ASD stops short, the drug may have crystallized in the particle or after dissolving. The two need different remedies.

② What they did. Each ASD was added to 400 mL buffer at 37 °C, 300 rpm, in an amount containing 8.8 mg celecoxib, so that full release gives 22 µg/mL (Table 1 lists 18 dispersions and their masses, 14.7 to 88 mg). Concentration was recorded at 249 nm every minute for 16 h, in triplicate. The binary dispersions were CEX:PAA 5:5, 3:7 (and 1:9), CEX:HPMCAS 6:4, 5:5, 3:7, 1:9 and CEX:HPMC 5:5, 3:7; the ternaries were CEX:PAA:HPMCAS 5:2:3, 5:3:2, 5:4:1, 4:5:1, 3:6:1, 1:8:1 and CEX:PAA:HPMC 2:1:1, 5:4:1, 3:6:1 (weight ratios of drug to polymer(s); drug loading 10–60%, with 2:1:1 being 50% drug). To separate the two crystallization routes, CEX:PAA 5:5 and 1:9 were also dissolved in buffer that already contained 5 µg/mL HPMCAS: if HPMCAS in solution stops solution crystallization, any drug still missing must have crystallized in the particle. Solids retrieved soon after the peak were examined under polarized light. Separately, particles on a glass slide were wetted with a few drops of buffer and watched under polarized light, where crystals appear as birefringent needles. P5·P6·P8·C8

③ Why it works. Because the dose equals the amorphous solubility, an ASD that stays amorphous should reach 22 µg/mL; a plateau below it with no fall means drug lost in the particle, while a rise followed by a fall means crystallization from solution. ④ Difference from earlier work. Most ASD dissolution tests use sink conditions, which hide crystallization; the authors argue that non-sink conditions are needed to evaluate amorphous formulations mechanistically (Section Dissolution Studies of ASDs).

Contribution 3: ternary dispersions and polymer–polymer miscibility

① The problem. Combining two polymers only helps if the inhibitor is where the drug is. If the polymers do not mix, the drug may end up partly in an inhibitor-poor phase.

② What they did. Glass transitions (Tg, onset of the second heating, about 20 °C/min) were measured by DSC for the pure components and the 1:1 binary ASDs. Because the polymer Tgs are close (117, 128 and 140), DSC could not test polymer–polymer mixing, so IR spectra (transmission, 64 scans, 4 cm^-1) were recorded of films spin-coated on KRS-5 from PAA:HPMCAS and PAA:HPMC blends at 7:3, 5:5 and 3:7 and of 2:1:1 films containing celecoxib, focusing on the carbonyl region 1600–1800 cm^-1. P9·P10·C56·M2–M4

③ Why it works. Spectral additivity in a blend means no specific interaction, which for polymers usually means no mixing; a new band means the groups now interact. ④ Difference from earlier work. Polymer–polymer miscibility of ASD carriers is hard to predict a priori (Marks 2014, ref 49); the paper checks it directly for the two pairs it uses.

Experimental analysis

The polymers barely change crystalline solubility

The equilibrium solubility of crystalline celecoxib in the buffer was 1.5 µg/mL, in good agreement with Abu-Diak et al. At 22 µg/mL, each polymer raised it only slightly (Table 2). The authors conclude that the polymers "do not solubilize the drug" and therefore do not change the supersaturation, the driving force for nucleation, to any substantial extent. R1·C11·C12

Table A. Crystalline solubility at 37 °C after 48 h, mean ± SD of three (from Table 2; card R1).

HPMCAS and HPMC block nucleation; PAA does nothing

Without polymer, celecoxib at 22 µg/mL began to crystallize within 5 min: the concentration fell from about 20.7 µg/mL to about 8 µg/mL at 20 min and about 4 µg/mL at 1 h (Figure 2). At 11 µg/mL the induction time was about 20 min, as classical nucleation theory predicts for a lower supersaturation. With 5 µg/mL PAA the curve was unchanged: about 9 µg/mL at 15 min, about 3.6 µg/mL at 1 h and about 2.7 µg/mL from 8 h onward (Figure 3). With 5 µg/mL HPMCAS or HPMC, the concentration drifted only slowly, to about 19.0 and 20.0 µg/mL after 16 h, and PAA plus HPMCAS behaved like HPMCAS alone. R2·R3·C13·C14·C15

The authors draw three conclusions. PAA is "an ineffective crystallization inhibitor". The large effect of the cellulose polymers cannot come from solubilization and "must be due to subtle molecular level interactions that are difficult to probe experimentally". And because PAA does not weaken HPMCAS, the result "lends support" to combining two polymers, "at least for the system studied herein". They find PAA's failure "perhaps not particularly surprising": earlier work suggests an inhibitor needs some hydrophobicity, and PAA is very hydrophilic and "is expected" to interact more with water than with celecoxib (Warren 2010; Ilevbare 2013; Schram 2015). C14·C16·C17·C18

Card check: in Figures 2, 3, 4 and 9 the crystallized solutions settle at about 2.6–2.9 µg/mL (about 3.3–3.5 µg/mL at the end of Figure 2), above the measured crystalline solubility of 1.5–1.72 µg/mL; the paper does not comment on this residual level (story note X4).

Binary dispersions: fast but crystallizing, or stable but slow

CEX:PAA 5:5 dissolved extremely fast and reached a maximum of 11 µg/mL within 3 min, but the drug then crystallized and the concentration fell to about 3.3 µg/mL at 1 h and about 2.7 µg/mL thereafter (Figure 4). The 30% PAA dispersion (3:7) peaked at about 9.3 µg/mL and fell the same way. The authors attribute the fast start, "presumably", to the full ionization of PAA at pH 6.8 (pKa 4.28) and its hydrophilicity; celecoxib (pKa 11.1) stays un-ionized, and the medium pH was unchanged. They leave open whether release stopped because the matrix crystallized first or because the dissolved drug crystallized so fast ("suggests either"). The particles were amorphous before wetting, so the loss is not from initial crystals. R4.a-b·R5·R6·C23·C24·C26·C27

Card check: the text says the 30% and 10% PAA profiles were "data not shown", but the 30% (3:7) profile is plotted in Figure 4 (story note X2).

The HPMCAS dispersions never lost supersaturation, but their release rate depended strongly on loading: at 10% drug release reached the amorphous solubility within 4 h, at 50% only after 16 h, and the 60% dispersion, shown but not discussed, reached about 15.8 µg/mL by 16 h. The authors take the sustained levels as confirmation that HPMCAS inhibits solution crystallization, and complete release as a sign that it also inhibits matrix crystallization. HPMC dispersions were slower still and did not finish within 16 h (Figure 13). R4.c-f·R17.a-b·C20·C22·C48

Table B. Celecoxib in solution (µg/mL; full release = 22) from binary dispersions, read from Figs. 4 and 13 (approximate; card R4, R17).

Card check: the 10% HPMCAS dispersion was cryo-milled, not sieved to 106–250 µm like the others, and its particle size is not given, so its fast release is not size-matched (story note X10).

Under the microscope, CEX:PAA 5:5 particles dissolved within 20 s and needle-shaped crystals were visible by 5 min (Figure 5); CEX:HPMCAS 3:7 particles dissolved slowly and showed no crystals in 30 min (Figure 6). The PAA process was so fast that the authors could not tell matrix crystallization from solution crystallization "based on these images alone". R7·R8·C28·C29

An inhibitor in the medium is not enough

The binary results suggested to the authors that a PAA dispersion that also contained a good inhibitor "could lead to improved overall dissolution performance". As a first test, CEX:PAA 5:5 was dissolved in buffer that already contained 5 µg/mL HPMCAS (Figure 7). Release was again very fast, but stopped at about 9.7 µg/mL; this time the concentration did not fall, decreasing only to about 8.8 µg/mL over 16 h. The 10% dispersion (1:9) behaved the same way, at about 11.6 to 10.8 µg/mL. The error bars are wide, about ±4 and ±3.4 µg/mL. Solids taken out soon after the peak were birefringent needles, "consistent with" the stable crystal form of celecoxib (Figure 8). R9·R10·C30·C32

The authors' reading is the key step of the paper: HPMCAS in solution now prevents the dissolved drug from crystallizing, so the missing half must have crystallized in the particles; "It therefore appears that the incomplete release is due to the crystallization of CEX in the matrix." Release and matrix crystallization compete, HPMCAS in the medium cannot influence the matrix, and once the matrix has crystallized no further drug can dissolve. Because a lower loading did not help, PAA is "a very poor inhibitor of matrix crystallization, irrespective of the drug loading". C31·C33

Card check: the paper does not locate the crystals or measure their amount, and identifies the retrieved solids by microscopy only. Crystals forming in the concentrated layer around fast-dissolving particles, where the local concentration exceeds the bulk, would also fit these data (story note X6).

Growth on seeds is slowed, nucleation far more

Why did the needle crystals from the particles not drain the supersaturated medium? Seeds would normally grow and pull the concentration down to the crystalline solubility. The seeded experiments (Figure 9) answer this: without polymer, seeds consumed the supersaturation at once, with no lag, from about 22.5 to about 4 µg/mL in 30 min. With HPMCAS the fall was slow. "It is clear", the authors write, that HPMCAS is an effective growth inhibitor at both supersaturations; at 11 µg/mL with 11 µg/mL HPMCAS very little desupersaturation occurred, and "This explains why" crystals from the PAA particles did not deplete the medium. HPMCAS is "a less effective inhibitor of growth at a higher supersaturation", in line with theory (Kubota and Mullin) and earlier work, although growth at 22 µg/mL was still "fairly slow". R11·C34·C35·C36·C37

Table C. Seeded desupersaturation (7 mg seeds in 50 mL, 37 °C), read from Fig. 9 (approximate), with the relative growth rates printed on p.22 (card R11, R12).

For 22 µg/mL celecoxib with 5 µg/mL HPMCAS, Eq. 6 with $R_0/R_p$ = 70 and $t_{u,p}$ = 960 min gives $J_0/J_p$ of about 4000. The authors conclude ("it can be surmised") that 5 µg/mL HPMCAS inhibits both nucleation and growth substantially, but that its effect on nucleation is "far more pronounced", even though the induction time with polymer was underestimated. R13·C40

Card check: the slopes and the induction time without polymer used in Eqs. 5 and 6 are not printed. Taking $t_{u,0}$ = 5 min from the text reproduces the value: $(1/70{)}^3\times (960/5{)}^4\approx 3960$. Figure 9 shows only 0–240 min, while $R_p$ was taken over 16 h; slopes read from the figure over 4 h give about 67, 150 and 115, close to the printed 70, 158 and 130 (card E5, E6; story note X8).

Card check: the explanation for Figure 7 rests on the seeded test at 11 µg/mL celecoxib with 11 µg/mL HPMCAS, but Figure 7 used 5 µg/mL HPMCAS; at 5 µg/mL seeds were tested only at 22 µg/mL celecoxib (story note X7).

Ternary dispersions with HPMCAS: complete at 30%, a trade-off at 50%

Putting HPMCAS into the PAA dispersion (Figure 10) gave mixed results. With HPMCAS as the minor polymer at 40–50% loading (4:5:1 and 5:4:1), a burst was followed by a sustained level; the authors call this "a slight improvement" over HPMCAS in the medium, but "it is apparent that some matrix crystallization was still occurring" because release was incomplete. Replacing more PAA by HPMCAS (5:3:2) "appeared to be reasonably successful" against matrix crystallization, reaching about 19 µg/mL by 16 h and still rising, but the burst fell to about 6 µg/mL within 5 min and the rest came slowly; 5:2:3 had an even smaller burst. The authors say this apparently biphasic release "immediately raises the suspicion of (partial) phase separation between PAA and HPMCAS". At 30% loading (3:6:1) release was full and rapid with the supersaturation kept, and 1:8:1 (10% drug) behaved the same. R14·C41·C42·C43·C44

Table D. Celecoxib in solution (µg/mL; full release = 22) from CEX:PAA:HPMCAS dispersions, read from Fig. 10 (approximate; card R14).

Card check: the text gives 12–14 µg/mL for the sustained level of 4:5:1 and 5:4:1, but Figure 10 shows plateaus of about 16.8 and 15.9 µg/mL from about 2–4 h; 12–14 µg/mL matches only the first minutes (story note X1).

Under the microscope the 3:6:1 dispersion dissolved rapidly and the matrix collapsed into fine particles, with no birefringence (Figure 11). A 2:1:1 dispersion, which was not dissolution-tested, dissolved very fast at first and then shed many particles of various sizes that dissolved slowly, again without crystals (Figure 12). The authors conclude that the HPMCAS ternaries improved both rate and crystallization inhibition over the binaries, but that rapid and complete release at 50% loading was not possible, and that the biphasic profiles "suggest that a formulation strategy combining PAA and HPMCAS would lead to very complex systems". R15·R16·C45·C46·C47

Ternary dispersions with HPMC: better at 50%

HPMC was tried next because it is also a strong nucleation inhibitor (Figure 3). The HPMC ternaries released much faster than the HPMC binaries, "benefiting from the presence of the fast releasing PAA" (Figure 13). At 30% loading (3:6:1, i.e. 60% PAA and 10% HPMC) release was essentially complete by 2 h, about 20.4 µg/mL, and the supersaturation lasted another 14 h. At 50% loading the 5:4:1 dispersion released fast, about 15.8 µg/mL at 1 h and about 19.0 µg/mL at 4 h, "about 85%" overall, with the supersaturation "substantially maintained"; this was more than the HPMCAS 5:4:1 ternary at only a slightly slower rate. The 2:1:1 dispersion rose more slowly, to about 17.2 µg/mL at 4 h and about 20.8 µg/mL at 16 h, and did not show the biphasic shape of the HPMCAS ternaries; the 5:2:3 and 5:3:2 HPMC ratios were not tested because their profiles "were expected to be similar" to 2:1:1. R17.c-e·C49·C50·C51

Table E. The 50% loading systems side by side (µg/mL; full release = 22), read from Figs. 4, 10 and 13 (approximate; card R4, R14, R17).

The microscope showed a different way of dissolving: 3:6:1 HPMC particles stayed whole and shrank, eroding from the edges, without breaking up into small particles as the HPMCAS ternaries did (Figure 14). The authors argue that this slower erosion "appears to be beneficial" for creating and keeping supersaturation: to prevent matrix crystallization, "it is important that the inhibitory polymer remains in close contact with the drug during hydration", and since the HPMC dispersions "appear to undergo erosion", drug and HPMC "should remain in contact" and be released together. R18·C52·C53·C54

Why HPMC and HPMCAS differ: polymer miscibility

Each 1:1 binary ASD had a Tg value between those of its components (Table 3): CEX:HPMCAS 67, CEX:HPMC 71 and CEX:PAA 83, against 56 for celecoxib and 117, 140 and 128 for the polymers. The authors read this as "some miscibility" of the drug with each polymer. R19·C55

Table F. Carbonyl bands (cm^-1) of spin-coated films. 1710, 1735 and 1743 are given in the text; the other positions are read from Figs. 15 and 16 (approximate; card R20, R21).

In PAA:HPMCAS blends both carbonyl bands appeared with little change in position: the spectra are additive, "suggesting a lack of specific intermolecular interactions", and the authors conclude that the IR data "suggest that PAA and HPMCAS are immiscible". They explain this by the low entropy of mixing of polymers and by PAA's acid dimers, which would break only for a more favourable interaction that HPMCAS "appears" unable to offer; adding celecoxib did not change the picture. In PAA:HPMC blends, by contrast, a shoulder at 1735 cm^-1 grew at the expense of the 1710 cm^-1 dimer band as the HPMC fraction rose, although HPMC itself has no band there. The authors read this as HPMC "may disrupt" the PAA dimers, through hydrogen bonds between HPMC OH and PAA C=O or between PAA OH and HPMC acceptors ("might occur through either"), and as "some degree of miscibility", which celecoxib does not undo. R20·R21·C57·C58·C60·C63·C64

Card check: the 2:1:1 CEX:PAA:HPMC film has a PAA:HPMC ratio of 1:1, but its trace resembles the 7:3 blend (maximum near 1713 cm^-1) more than the 5:5 blend (maximum near 1735 cm^-1), while the text calls it "that same spectroscopic pattern"; the traces also appear to be drawn to a common maximum without a stated normalization (story note X14).

Discussion

The authors' chain of reasoning. Put together, the paper argues as follows. PAA dissolves very fast and releases a burst of celecoxib, but it does nothing against crystallization. A cellulose polymer in the medium protects the dissolved drug, yet cannot reach the drug inside the wetting particle, where it crystallizes. The inhibitor therefore has to be inside the particle, close to the drug, and leave with it. Whether that happens depends on whether the inhibitor mixes with PAA:

The authors' overall conclusion is cautious: the advantages of combining a fast-release polymer with a good matrix and solution crystallization inhibitor "were realized at a drug loading of 30% and partially realized at higher drug loadings", and combining two polymers "thus appears to be a viable formulation strategy, although attention needs to be paid to the miscibility of the ternary systems" (Conclusions). The abstract is more confident, saying that certain HPMC:PAA ratios achieved rapid release and crystallization inhibition "even at high drug loadings". C68·C69·C71

What is well supported. The solution experiments are unambiguous: at 5 µg/mL, PAA has no effect on celecoxib nucleation, while HPMCAS and HPMC hold a 22 µg/mL solution for 16 h. The polymers barely change crystalline solubility, so the effect is kinetic. The release tests, all at the same dose, medium and temperature in triplicate, show directly that at 30% loading both ternaries release about 20–21 µg/mL within about 2 h and keep it, far ahead of every binary. R1·R3·R14.e·R17.e·P5

What rests on inference. Two steps of the chain are inferred rather than observed. Matrix crystallization in PAA dispersions is deduced from the plateau with HPMCAS in the medium and from needles retrieved after the peak; the crystals were not located or quantified. The two-domain structure of the PAA:HPMCAS dispersions is deduced from IR spectra of thin polymer films; the dissolution-tested 50% loading ternaries were never examined for domains, and the 2:1:1 HPMCAS film used for IR was not dissolution-tested. The statements about HPMC as a growth and matrix inhibitor are extrapolated from HPMCAS data and from release extents. C31·C61·X5·X6·X13

Note by the AI writer (not discussed in the paper). Several observations fit the authors' picture but would also fit other readings. HPMCAS and HPMC were equally strong nucleation inhibitors in solution, yet only HPMC worked well in the 50% ternaries; this is the strongest argument in the data that something other than inhibitory strength, such as mixing with PAA or the way the particle dissolves (erosion versus break-up), decides the outcome. The test dose equalled the amorphous solubility, so the experiment asks whether a dispersion can reach that level, not whether it can stay above it; at higher doses, which are common in vivo, drug-rich phases could form and the balance between PAA and the inhibitor might shift. PAA's speed comes from its ionization at pH 6.8, and the strategy was not tested at low pH, where PAA is un-ionized.

What it suggests for formulation. The authors' own advice is to combine a fast-release polymer with an inhibitor and check the miscibility of the ternary, and to keep the inhibitor in contact with the drug during hydration, because an inhibitor in the medium does not stop matrix crystallization; growth from seeds is harder to stop than nucleation, so preventing crystals in the dissolving matrix is, in the abstract's word, "critical". C69·C54·C70 Note by the AI writer: in these data a 10% inhibitor fraction protected a 30% loading but not a 50% loading with HPMCAS, so the inhibitor-to-drug ratio and the polymer pairing both deserve screening; a quick IR check of the two carriers' mixing could be a cheap filter before release testing.

Limitations

Limits the authors state. The induction-time effect of the cellulose polymers "must be due to subtle molecular level interactions that are difficult to probe experimentally". Microscopy could not "unambiguously differentiate matrix crystallization from solution crystallization" in the PAA dispersions. DSC could not be used for polymer–polymer miscibility because the polymer Tgs are similar. The induction time with 5 µg/mL HPMCAS was capped at 960 min and is therefore "underestimated". The hydrogen-bond pattern between PAA and HPMC "might occur through either" of two routes. The 5:2:3 and 5:3:2 HPMC ternaries were not tested. C17·C28·C56·C39·C63·C51

The AI writer's own judgement. The core results, the solution tests and the 30% loading release data, are strong. The mechanism is moderately supported at best. Specific points from the card's extractor notes:

• Text and figures disagree in places: the HPMCAS 5:4:1 and 4:5:1 plateaus are 12–14 µg/mL in the text but about 16–17 µg/mL in Figure 10; the 30% PAA profile is "data not shown" in the text but plotted in Figure 4; two cross-references point to the wrong figure (Figure 3 for the dissolution of Figure 4, Figure 2 for the PAA induction result of Figure 3). X1·X2·X3
• Crystallized solutions settle at about 2.7 µg/mL, not at the measured crystalline solubility of 1.5 µg/mL, although the text says the PAA dispersions "approached the crystalline solubility". X4
• The polymer concentrations differ between experiments: 22 µg/mL for solubility, 5 µg/mL for induction times, 5–22 µg/mL for seeded growth and 5 µg/mL in the Figure 7 medium, so the link between solubility and induction results, and between the seeded test and Figure 7, is qualitative. X7·X9
• The growth and nucleation ratios cannot be reproduced exactly from printed data; Figure 9 shows 4 h while $R_p$ used 16 h. X8
• Table 3 prints no unit or replicate number, and its uncertainties reach ±5–6; drug–polymer miscibility is inferred only from intermediate Tg values. X15
• The abstract's "even at high drug loadings" is stronger than the conclusions' "partially realized". X12
• The cover page title spells the drug "Celexocib", and the reference list repeats three works (refs 16/17, 23/29, 26/28). X16

Conclusion

The paper makes three contributions. It separates the effects of PAA, HPMCAS and HPMC on celecoxib solubility, nucleation and growth, showing that the cellulose polymers block nucleation strongly and growth less, while PAA does neither. It uses release tests at a dose equal to the amorphous solubility, with and without an inhibitor in the medium, to show that PAA dispersions lose drug that dissolved inhibitor cannot recover, which the authors attribute to crystallization in the matrix. And it shows that PAA with 10% HPMCAS or HPMC releases celecoxib completely and quickly at 30% loading, while at 50% the HPMC pairing works better, a difference the authors link to the partial miscibility of PAA with HPMC and the immiscibility of PAA with HPMCAS. Within one drug, one medium and one dose, the two-polymer approach is shown to work at moderate loading; its extension to higher loadings, other drugs and the proposed domain mechanism remain to be tested.