Keeping high-loading celecoxib amorphous under heat and humidity: which polymer, and does a second polymer hurt?

Background and motivation

Amorphous drugs dissolve faster and reach higher apparent solubility than their crystals, but they tend to crystallize during storage, and a polymer is mixed in at the molecular level to prevent it (Introduction, refs 1–2). Two mechanisms are usually invoked. A polymer with a high glass transition temperature (Tg) raises the Tg of the mixture and so lowers the mobility of the drug, an effect called antiplasticization; but several studies found that crystallization inhibition cannot always be explained by Tg alone. Polymers also form specific interactions with the drug, such as hydrogen bonds or ionic interactions, and the strength of these interactions "appears to have some correlation" with how well a polymer inhibits crystallization (Wegiel and co-workers, ref 11). C6

Water complicates both mechanisms. Amorphous solids take water into their bulk, not only onto their surface, and water, with a Tg of −137 °C, lowers the Tg of the dispersion and raises molecular mobility (refs 12–15). M12 Water can also break drug–polymer hydrogen bonds by bonding to a hydrophilic polymer itself (Marsac and co-workers, ref 16). Either effect may split the dispersion into a drug-rich and a polymer-rich amorphous phase, amorphous–amorphous phase separation (AAPS; Rumondor and Taylor, ref 17; Purohit and Taylor, ref 18), after which crystals form preferentially in the drug-rich domains; or the higher mobility may let the drug crystallize directly. C7

A further complication is that the drug must stay amorphous on the shelf and also during dissolution, when the dispersion produces a supersaturated solution, and the best inhibitor for one phase is often poor in the other. Polyacrylic acid, for example, inhibits the growth of acetaminophen crystals from the supercooled liquid very well but does not inhibit nucleation or growth in aqueous solution (Trasi and Taylor, refs 26, 27). The authors therefore see combinations of polymers, one chosen for the solid and one for solution, as a possible advance in formulation, and note that the miscibility, the interactions and the physical stability of such ternary systems had not been studied in depth. C8·C9

The paper continues the group's own earlier work. Xie and Taylor (ref 32) had found that high-loading celecoxib (CEX) dispersions made with polyvinylpyrrolidone (PVP) released the drug better when 20% of the PVP was replaced by hydroxypropyl methylcellulose (HPMC) or HPMC acetate succinate (HPMCAS), two polymers that inhibit celecoxib crystallization from supersaturated solution very effectively. R27 The open question was whether the solid state pays for this. The paper sets out to evaluate the physical stability of celecoxib dispersions made with different polymers under high temperature and high relative humidity (RH), and the miscibility and stability of the ternary dispersions relative to the matching binary ones. A1·A3·A4

Background concepts

Drug loading and ratios. All compositions are celecoxib:polymer by weight. A 6:4 dispersion holds 60% drug and 40% polymer; a ternary 6:3.5:0.5 dispersion holds 60% drug, 35% PVP and 5% HPMCAS or HPMC, so 12.5% of the polymer is the cellulosic; in 8:1.6:0.4 and 5:4:1 the cellulosic is 20% of the polymer. A2·A3

Reading hydrogen bonds in an infrared spectrum. Celecoxib carries a sulfonamide NH2 group, a hydrogen-bond donor; PVP and the vinylpyrrolidone units of PVP/VA carry an amide carbonyl, an acceptor; HPMCAS carries ester and acid carbonyls; HPMC has only hydroxyl and ether oxygens (Figure 1). M1·M2·M4·M5·M6 When the NH2 group is hydrogen-bonded, its stretching band moves to a lower wavenumber, and a lower N–H frequency goes with a shorter, stronger hydrogen bond (Nakamoto and co-workers, ref 34). C10 On the acceptor side, a carbonyl that takes part in a hydrogen bond absorbs at a lower wavenumber than a free one, so a new low-wavenumber shoulder on the polymer C=O band marks the bonded fraction. These are the two signatures the paper reads in Figures 2–4.

Glass transition, antiplasticization and plasticization. A single Tg between those of the components is taken as evidence that drug and polymer are mixed in one amorphous phase. C26 A polymer with a higher Tg raises the Tg of the mixture (antiplasticization); sorbed water, with a very low Tg, lowers it (plasticization). Here the onset Tg is reported, measured by differential scanning calorimetry (DSC) on the second heating at 20 °C/min for dry samples. P6

AAPS seen by DSC. If water splits the dispersion into two amorphous phases of different composition, each phase shows its own glass transition. The authors therefore exposed 6:4 dispersions to humidity, sealed them hermetically so the water stayed in, and counted the Tg events in the first heating: one event was read as one phase, two as partial demixing, and the lower Tg was "surmised" to belong to a drug-rich domain. P7·C32·C33 Amorphous celecoxib itself has a Tg of about 56 °C, below that of every polymer used (about 97–170 °C, read from Figure 6). M1·M2·M3·M4·M5·M6

Hygroscopicity. The water a material takes up at a given RH, measured here by dynamic vapor sorption (DVS): the sample is dried and then held at increasing RH steps while its mass is recorded. P8

Crystal growth rate from the supercooled melt. A physical mixture of crystalline drug and polymer is melted and quench cooled into a glass, small nuclei are allowed to form, and the growth of their diameter is followed under a polarized light microscope at a set temperature (method of Kestur and Taylor, ref 23). It measures how strongly a polymer slows growth at a fixed composition, independent of nucleation. P9

Powder X-ray diffraction (PXRD). An amorphous sample gives a broad halo; sharp peaks on the halo mark crystalline drug. Stability is reported as the time until such peaks are first seen. P5

Methods in detail

The study has one drug and five polymers: two grades of PVP (K29/32, Mw 58,000; K12, MW 2000–3000), PVP-vinyl acetate (PVP/VA, Kollidon VA64), HPMCAS (grade AS-MF) and HPMC (Type 606), plus ternary dispersions in which PVP is the major polymer and HPMCAS or HPMC the minor one. M1·M2·M3·M4·M5·M6·A3 Four kinds of measurement were made to explain the storage results, which come last.

Preparing the dispersions

Celecoxib and the polymer or polymers were dissolved in a mixture of ethanol and dichloromethane, the solvent was removed by rotary evaporation, and the solid was dried in a vacuum oven overnight, ground with mortar and pestle, sieved to 106–250 µm and kept over phosphorus pentoxide at room temperature (RT) until use. P1 The paper does not give the solvent ratio, the solids concentration or the evaporation settings. P1

Interactions by infrared spectroscopy

Mixtures were dissolved in the same solvent, dropped onto thallium bromoiodide (KRS-5) crystals and spun into thin films, which were measured in transmission (64 scans, 4 cm−1 resolution, 4000–500 cm−1) under a dry-air purge. P2 Crystalline celecoxib and melt-quenched amorphous celecoxib were measured by attenuated total reflectance on the same instrument. P3 Drug:polymer ratios ran from 9:1 to 1:9 (Figure 2); PVP:HPMCAS blends without drug ran from 10 to 90% HPMCAS (Figure 3); ternary films carried 10% or less, 40% or 60% HPMCAS (Figures S4–S6). P2·R7·R8 To see what water does to the bonds, films were exposed to 94% RH at RT and measured after 1 and 18 days (Figure 4); the paper does not describe the humidity chamber. P4·X16

Thermal analysis

Dry Tg was measured by DSC on 3–6 mg in Tzero pans with a pinhole lid: heating at 20 °C/min to about 30 °C above Tg, cooling, and reheating at 20 °C/min to 180 °C, with the cycle repeated twice and the onset Tg of the second heating reported. P6 For the effect of moisture, 6:4 binary dispersions were held in open pans at RT for 3 days at 75% or 94% RH (Table 1 also lists 56% RH), then sealed hermetically, equilibrated at −20 °C, heated to 100 °C at 20 °C/min, cooled to −20 °C and reheated to 100 °C twice. P7 The method says the onset Tg of both the first and second heating was reported; Table 1 gives one value per sample without saying which heating. X5

Water sorption

Samples were dried at 45 °C under dry nitrogen (to less than 0.01% w/w change in 2 min, at most 60 min) and then taken from 5% to 95% RH in 10% steps at 25 °C, with at most 180 min per step. P8 The samples were amorphous celecoxib, the five polymers, 5:5 binary dispersions and 5:4:1 ternary dispersions with PVP-K29/32. P8·R12·R13

Crystal growth rate

Crystalline celecoxib and polymer were mixed in a cryogenic mill, melted between two cover slips on a hot plate and quench cooled on a metal surface. The slips were held at 80 °C until small nuclei became visible, and the increase in nucleus diameter with time was then followed with a polarizing microscope on a hot stage at 80, 90 or 100 °C. P9 The compositions were neat celecoxib, 90:10 binaries with HPMCAS, PVP/VA, PVP-K12 and PVP-K29/32, and 90:8:2 ternaries with PVP-K12 or PVP-K29/32 plus HPMCAS; HPMC was not included. P9·X11

Storage stability

Dispersion powders in open 20 mL vials were stored under seven conditions: 80, 60 and 40 °C over Drierite (0% RH); RT (about 20 °C) over saturated sodium bromide (56% RH), sodium chloride (75% RH) and potassium nitrate (94% RH); and 40 °C over sodium chloride (75% RH). PXRD patterns were taken periodically from 5° to 35° 2θ at 4°/min. P5 The main text reports three of the seven conditions: 80 °C/0% RH with 80% drug, RT/94% RH with 60% drug and 40 °C/75% RH with 80% drug; a summary of all conditions is said to be in Table S1. P5·X2 Figure 9 shows one pattern per condition (after 6, 28 and 33 days); the sampling schedule is not given. X8

Experimental analysis

Which polymers bind celecoxib

Neat crystalline celecoxib shows its NH2 stretches at 3322 and 3225 cm−1; in the amorphous drug they move up to 3351 and 3268 cm−1, so the authors conclude that hydrogen bonding in the amorphous drug "appears weaker" than in the crystal. R1·C10 Table A collects the band changes on mixing with each polymer.

Table A. Infrared band positions as stated in the text (Figures 2 and 3; PVP/VA and HPMC films in Figures S2 and S3, not supplied). R1·R2·R3·R4·R5·R6·R7·C11·C12·C14·C15·C16

With PVP the band at 3196 cm−1 grows at the expense of the 3268 cm−1 band as polymer is added, and the shoulder at 1661 cm−1, where the drug itself does not absorb, grows as the drug:polymer ratio rises; together these "clearly" indicate a hydrogen bond between the celecoxib NH2 and the PVP carbonyl. R3·C11 PVP/VA gives the same signature on its pyrrolidone carbonyl, so the bond strength is comparable; but at a given drug loading "it appears" that more drug molecules are bonded with PVP-K29/32, which the authors "presumably" attribute to the vinyl acetate units diluting the pyrrolidone groups (Van Eerdenbrugh and Taylor, ref 35). R4·C12·C13 With HPMCAS the shifts are smaller, read as a much weaker interaction, and with HPMC, which has no carbonyl, there is no clear sign of interaction at all. R5·R6·C14·C15

For the ternary systems, the authors first showed that PVP and HPMCAS bond with each other: in drug-free blends the PVP carbonyl gains a shoulder at 1643 cm−1, assigned to PVP bonded to the HPMCAS carboxylic acid, at a lower wavenumber than the drug–PVP shoulder, so the polymer–polymer bond is the stronger one. R7·C16 The 1682 cm−1 band stays dominant even at high HPMCAS content, "consistent with" the low fraction of acid groups in HPMCAS. R7·C17 Yet in ternary films with 10% HPMCAS or less, the spectra look like celecoxib:PVP and the 1643 cm−1 shoulder is absent; the celecoxib–PVP bond persists up to 40% HPMCAS, and the PVP–HPMCAS bond appears only at 60% HPMCAS, less pronounced than without drug. R8 The authors conclude that the celecoxib–PVP bond is favoured over other pairings and "is not substantially disrupted" by HPMCAS, while noting that it is "unclear" whether PVP–HPMCAS bonds exist at all at the low HPMCAS levels. C19·C18

Dry state: Tg, crystal growth and storage at 80 °C

Every binary dispersion showed one Tg between those of its components, which "suggests" that celecoxib is miscible with each polymer (Figure 6). R14·C26 The vinylpyrrolidone polymers raised Tg most. PVP-K12 (about 97 °C) and PVP/VA (about 102 °C) have a lower Tg of their own than HPMCAS (about 117 °C) and HPMC (about 140 °C), yet they "apparently" raised the Tg of the dispersions more effectively. M3·M4·M5·M6·C27 PVP-K29/32, with a Tg of about 170 °C, gave the highest values throughout. M2·R14.a The ternary dispersions with PVP and HPMCAS also showed a single Tg, close to but slightly below that of the PVP binary, as the authors expected from the IR results (Table S1). R15·C31 PVP:HPMCAS blends without drug showed one Tg at every ratio. R16·C29

Crystal growth from the quench-cooled melts followed the same order (Figure 8). Neat celecoxib grew fastest, at about 0.010, 0.14 and 0.97 µm/s at 80, 90 and 100 °C. R21.a With 10% HPMCAS, growth at 80 °C fell to about 0.00031 µm/s; with 10% PVP/VA, PVP-K12 or PVP-K29/32 it fell further, to about 0.00007, 0.000037 and 0.00003 µm/s. R21.b·R21.c·R21.e·R21.g In the authors' words, growth was "much faster" with HPMCAS than with PVP, PVP/VA gave "slightly faster" growth than PVP, possibly because PVP carries more pyrrolidone carbonyls per gram ("may be because"), and the higher PVP molecular weight reduced growth only "minimally". C36·C37 Replacing a fifth of the PVP by HPMCAS (90:8:2) raised the growth rate slightly, to about 0.000053 µm/s with PVP-K29/32 and 0.00007 µm/s with PVP-K12 at 80 °C, still far below the HPMCAS binary. R21.d·R21.f·C36

Table B. Dry-state measures side by side. Tg from Table 1 (unexposed); growth rates read from Figure 8 (approximate); storage times as stated in the text. R17·R21·R22·M1

Storage at 80 °C/0% RH matched this ranking (Figure 9a). At 80% drug, the binaries with PVP-K29/32 and PVP/VA and the ternaries with PVP plus a little HPMCAS or HPMC showed signs of crystallization only after about 120 days, whereas the HPMCAS and HPMC binaries crystallized fastest, with crystallinity visible after 6 days. R22·C38 The 6-day patterns show only a halo for the four vinylpyrrolidone-based dispersions and sharp peaks for HPMCAS (strongest) and HPMC; the 120-day point itself is not shown. R22·X8 The authors take two lessons: the stability of a ternary "could be largely predicted" from the binary of similar composition, and growth rates, which are quick to measure, "appear to be predictive" of the relative crystallization of the dispersions and "might potentially" shorten long-term stability studies. C39·C40

Note by the AI writer (not discussed in the paper): the growth rates were measured at 90:10 on melts and the storage at 80:20 on solvent-evaporated powders, and HPMC was not in the growth set, so the agreement is a rank agreement for the polymers that were measured. X11

High humidity: the ranking changes

At RT/94% RH and 60% drug the order changed (Figure 9b). Celecoxib:PVP/VA was the most stable binary, amorphous for up to 150 days; celecoxib:HPMCAS followed, stable for up to 115 days; the binaries with HPMC or PVP and the ternaries with PVP as the major polymer were "much less stable", with diffraction peaks within a month. R23·C41 The 28-day patterns show strong sharp peaks for the PVP-K29/32:HPMC ternary, smaller peaks for celecoxib:PVP-K29/32, weak features for the HPMCAS ternary, at most faint features for celecoxib:HPMC, and only a halo for celecoxib:PVP/VA, celecoxib:HPMCAS and celecoxib:PVP-K12. R23

Note by the AI writer (not discussed in the paper): the text groups "the binary ASDs with HPMC or PVP" among those with peaks within a month, but at 28 days the PVP-K12 dispersion in Figure 9b shows no peaks and the HPMC dispersion at most faint ones; the 150- and 115-day points are stated, not shown. X9·X8

Water uptake explains much of this (Figure 5). The neat polymers take up very different amounts of water: at 95% RH about 68 wt% for PVP-K29/32, 67 for PVP-K12, 48 for PVP/VA, 31 for HPMC and 12 for HPMCAS, against only 2 wt% for amorphous celecoxib. R12·C23 In the 5:5 dispersions the order follows the polymers, with one exception: celecoxib:PVP/VA takes up only about 7.0 wt% at 95% RH, close to celecoxib:HPMC (about 8.0 wt%), so its uptake was "largely suppressed". R13·C24 The PVP dispersions rise steeply above 75% RH, to about 21.8 (PVP-K12) and 17.2 wt% (PVP-K29/32) at 95% RH, while celecoxib:HPMCAS reaches only about 3.6 wt%. R13·C24

Table C. Humidity measures side by side. Water uptake read from Figure 5 (approximate, 25 °C); Tg from Table 1 (3 days at RT; the heating scan is not stated); Tg drop computed from the Table 1 means (Figure 10 plots the same); stability as stated in the text. R13·R17·R20·R23·X5

Table 1 shows what the water does to Tg. After 3 days at 94% RH the Tg of the PVP-K29/32 dispersion fell from 103 to 37.4 °C, and that of PVP-K12 from 84.4 to 22 °C; the other three fell to 38–42.1 °C. R17 Plotted as a reduction (Figure 10), the PVP dispersions lose the most Tg at every RH, about 39–42 °C already at 56 and 75% RH for PVP-K29/32 and about 62–66 °C at 94% RH for both grades, against about 25–39 °C for the others. R20·C47 But Tg after exposure does not rank stability: four dispersions sit between 37.4 and 42.1 °C at 94% RH, yet range from peaks within a month to 150 days, and the authors conclude that "it is clear" the Tg values alone cannot explain the differences. R17·R23·C46

Evidence for amorphous–amorphous phase separation

After 3 days at 94% RH, the first DSC heating of the two PVP dispersions shows two glass transitions, marked by arrows in Figure 7a: at about 28 and 53 °C for PVP-K12 and about 55 and 84 °C for PVP-K29/32 (read from the figure). PVP/VA, HPMCAS and HPMC show a single Tg, "implying" that water uptake did not cause AAPS. R18·C32 On reheating, all samples show a single Tg; "a possible explanation" is that the separated phases remixed during heating (Marsac and co-workers, ref 16). R18·C34 After 75% RH the two-Tg signature "appeared less pronounced" in PVP-K12 and was absent in PVP-K29/32 and the others. R19·C35

The infrared spectra of celecoxib:PVP films at 94% RH point the same way. The bonded N–H band at 3196 cm−1 loses intensity while the 3268 cm−1 band gains, and the bonded carbonyl shoulder at 1661 cm−1 decreases slightly (Figure 4); the film crystallized within 18 days. R9 The authors read this as water irreversibly disrupting "some" of the celecoxib–PVP hydrogen bonds and "possibly" inducing some AAPS before crystallization. C20 Celecoxib:HPMCAS and celecoxib:PVP/VA films "appeared resistant" to both, with no notable change over 18 days. R10·C21 Ternary films changed like celecoxib:PVP films, which the authors put down to PVP being the larger polymer fraction. R11·C22

Note by the AI writer (not discussed in the paper): two points weaken this evidence. In Figure 7a the arrow at about 55 °C for PVP-K29/32 sits on a sharp endothermic peak, and peaks at about 53–55 °C also appear in the PVP/VA, HPMC and HPMCAS traces, which the authors read as single-Tg samples; the second event at about 84 °C is the clearer one. X6 And the abstract's statement that ternary dispersions also "appeared to" undergo AAPS rests on film IR only (Figure S9), because no DSC was done on ternary powders after humidity. X7·C5 The text also says crystallization occurred in the 7:3 film, while Figure 4 shows a 3:7 film with the same note in its caption. X3

Combined heat and humidity: 40 °C/75% RH

At 40 °C/75% RH and 80% drug (Figure 9c), celecoxib:PVP/VA 8:2 was again the most stable, amorphous for up to 4 months; celecoxib:PVP-K29/32 and its HPMCAS ternary stayed amorphous for up to 70 days; celecoxib:HPMCAS was stable for up to 33 days and crystallized thereafter. R24·C42 The authors call it "interesting" that PVP now outperforms HPMCAS. C42 At 75% RH the water uptake of the PVP and HPMCAS dispersions differs much less than at 94–95% RH (about 5.3 against 2.2 wt% for 5:5), and the authors write that the results "suggest" that drug–polymer interactions became the dominant factor at this condition. R13·C53

Note by the AI writer (not discussed in the paper): the 33-day pattern of celecoxib:HPMCAS in Figure 9c already carries small sharp features, so 33 days is better read as about the first detection; and the text's statement that crystallization occurred "at an earlier time point in all systems" does not name the comparison (PVP/VA lasted about 4 months here, against about 120 days at 80 °C/0% RH). X10·X13

Ternary against binary

Table D. The ternary dispersions next to the PVP binary. Water uptake and growth read from Figures 5 and 8 (approximate); storage as stated in the text. R13·R15·R21·R22·R23·R24

Across every test the ternaries followed the PVP dispersion (Table D). The cellulosic lowered water uptake slightly, but in the authors' words this reduction "was not large enough" to change physical stability; the ternaries were stable dry but crystallized after high-RH exposure. R13·C51 The authors connect this to the IR result: a small amount of HPMCAS "was not able to disrupt" the drug–PVP interactions "and hence did not substantially impact" crystallization kinetics. C45 Note by the AI writer (not discussed in the paper): at 94% RH the PVP-K29/32:HPMC ternary shows more and stronger diffraction peaks at 28 days than the PVP binary; since crystallinity was not quantified, the equivalence holds for the time to first peaks, not necessarily for the extent of crystallization. X14

Discussion

The authors' explanation

The authors explain the results with two factors whose weight depends on the stress. In the dry state, the vinylpyrrolidone polymers raise Tg more and form stronger hydrogen bonds with celecoxib, which "may contribute" to their better stability; the stronger bonds "may restrict the diffusive motion and hence self-association" of the drug molecules. C2·C44 They cite Miyazaki and co-workers (ref 7), who found that polyacrylic acid stabilized amorphous acetaminophen better than PVP through a stronger hydrogen bond and longer enthalpy relaxation times despite a similar Tg. R29 Under high humidity, hygroscopicity takes over: the rapid crystallization of the PVP dispersions "can therefore most likely be accounted for" by their large water uptake, which disrupts drug–polymer interactions and leads to crystallization. C48 The chain, with its two outcomes, is drawn below.

PVP/VA dispersions "absorb much less water and hence are more resistant to both plasticization and AAPS", in line with Rumondor and Taylor (ref 17), who found for felodipine and quinidine that PVP/VA bonds the drug as strongly as PVP but shows less moisture-induced phase separation, linked to lower water uptake. C49·R28 The cellulosics form a third case. HPMCAS takes up the least water, yet celecoxib:PVP/VA still outlasts celecoxib:HPMCAS at 94% RH, "most likely due to the stronger drug–polymer interaction strength"; by the same argument PVP/VA outlasts HPMC, which takes up similar water. C50 At 40 °C/75% RH, where the difference in water uptake between PVP and HPMCAS dispersions is small, the results "suggest" that the interaction became the dominant factor, so PVP outlasted HPMCAS there. C53 In the conclusions this becomes a general statement for celecoxib: in the absence of moisture, polymers that form stronger hydrogen bonds and raise Tg inhibit crystallization better; under high RH, stronger interaction and lower hygroscopicity protect against AAPS and crystallization, and Tg "was not predictive". C2·C4·C46

What this means for polymer combinations

The central practical result is that a minor cellulosic does no harm in the solid state. The authors call this "an important observation" that "lends support" to their view that ASDs with two polymers "may be useful to optimize performance in different scenarios (e.g., storage and dissolution)", and in the conclusions they speak of "proof of concept" for miscible polymer pairs that maximize crystallization inhibition in both the solid and the solution phase. C3·C43 They add that the cellulosics are very effective inhibitors in solution even at very low concentration, but that this "clearly" does not translate into the solid matrix exposed to high RH when the hygroscopic PVP is present. C52·R27

The authors also consider a single polymer. PVP/VA stabilized amorphous celecoxib well at elevated temperature and humidity, and in solution it extended the nucleation induction time to about 13 h at 5 µg/mL, against 5 min without polymer and under 2 h with PVP (Figure S11). R25 So "the same goal might be reached" with PVP/VA alone. C54 Its dispersion, however, dissolved much more slowly than the PVP:HPMCAS ternaries and gave a lower supersaturation (Figure S12), which "may be undesirable" for a drug used against pain. R26·C55 Six and co-workers (ref 29) reported a similar trade-off for itraconazole: stable but slow-releasing with PVP/VA (45% released after 3 h), fast-releasing but poorly stabilized with Eudragit E100, and both fast and stable as a ternary. R30 Note by the AI writer (not discussed in the paper): the reference given for the induction-time and supersaturation comparators (ref 34) is a 1955 paper on hydrogen-bond stretching frequencies, so these comparators cannot be traced in the reference list. X1

Note by the AI writer (not discussed in the paper): for a formulator the paper shows that dry-heat and humid stress tests can rank the same polymers in opposite ways. PVP ranks with PVP/VA in dry heat but behind HPMCAS at 94% RH. A small cellulosic fraction does not change this, so in a PVP-major ternary the moisture protection has to come from the choice of the major polymer or from the packaging. And because Tg after moisture exposure was similar for dispersions with very different stability, water uptake and the persistence of drug–polymer bonds under humidity are worth measuring alongside Tg.

Limitations

Limits the authors state. The authors note that DSC has "practical limitations" as a probe of drug–polymer mixing, although it remains the most frequently used method (refs 40–42). C25 They write that the exact compositions of the two amorphous phases after 94% RH "cannot be quantified", so the drug-rich assignment of the lower Tg is "surmised". C33 And it is "unclear" whether PVP and HPMCAS hydrogen-bond with each other in ternaries with 10% HPMCAS or less, and whether such bonds would form at the expense of the drug–PVP bond. C18

The AI writer's own judgement (1): what the storage data can carry. The dry-state ranking is consistent across four independent measures (IR band shifts, dry Tg, growth rates and storage at 80 °C), which makes it the strongest part of the paper. R3·R14·R21·R22 The storage evidence itself is thin in what it shows. Each condition is illustrated by one PXRD pattern; the time points that carry the ranking (about 120, 150, 115 days, about 4 months, 70 days) are stated without patterns or a sampling schedule; and crystallinity was not quantified. X8 Two details of Figure 9b do not follow the text (no peaks for PVP-K12 at 28 days; stronger peaks for the HPMC ternary than for the PVP binary), and at 40 °C/75% RH the HPMCAS pattern already shows features at the point described as stable. X9·X14·X10 The results for four of the seven storage conditions are not reported in the main text, and Table S1, said to summarize them, is cited elsewhere for ternary Tg values. X2

The AI writer's own judgement (2): the humidity explanation. The explanation combines measurements made at different compositions: sorption on 5:5 dispersions at 25 °C, Tg after humidity on 6:4 dispersions at RT, growth on 90:10 melts, and storage on 8:2 and 6:4 powders, with the ternaries at 5:4:1, 6:3.5:0.5 and 8:1.6:0.4. X12·X11 Rankings can be compared across these tests, absolute values cannot. The AAPS evidence rests on two-Tg DSC for the binaries, where one of the two PVP-K29/32 events coincides with an endotherm that the single-Tg samples share, and on film IR for the ternaries. X6·X7 Table 1 does not say from which heating its values come, and they lie close to the single Tg of the second heating, so they probably describe the remixed rather than the separated state. X5 Hydrogen bonding was measured on dry films, not on hydrated powders, so the step "stronger interaction persists under humidity" for PVP/VA against HPMCAS is inferred rather than measured. C50·X16

Conclusion

The paper makes three contributions. First, it ranks five polymers for high-loading celecoxib dispersions by interaction strength (IR), Tg, crystal growth and water uptake, and shows that the dry-state ranking (PVP and PVP/VA ahead of HPMCAS and HPMC) follows interaction strength and Tg. C1·C2·R22 Second, it shows that high humidity reorders the polymers: PVP, the most hygroscopic, takes up the most water, loses the most Tg, "appeared to" phase-separate and crystallized within a month, while PVP/VA and HPMCAS lasted 150 and 115 days, and Tg after humidity did not predict the outcome. C5·C46·R23 Third, it shows that replacing 12.5–20% of PVP by HPMCAS or HPMC leaves the stability of the dispersion essentially that of the PVP binary under all three conditions, which the authors present as proof of concept for miscible polymer pairs that combine solid-state and solution inhibition. C3·C43·A3 What remains open is whether the ternary powders themselves phase-separate under humidity, and whether the dissolution advantage of the pair survives storage. X7·C43