Choosing a polymer for a drug that will not stay amorphous: resveratrol and seven polymers

Background and motivation

An amorphous solid dispersion (ASD) disperses a poorly soluble drug molecularly in a polymer, so that the drug dissolves faster and reaches a higher transient solubility than its crystal (Leuner and Dressman 2000, ref 1; Serajuddin 1999, ref 2; Chiou and Riegelman 1971, ref 3). The price is stability: the amorphous solid sits at a higher energy than the crystal and can recrystallize over pharmaceutically relevant times, which removes the solubility advantage (refs 4–9) C1. Keeping the drug amorphous during manufacture and storage is therefore the first condition for any benefit.

Polymers are widely accepted to slow this crystallization (refs 2, 10), but how they do it was debated. One view holds that the polymer raises the glass transition temperature (Tg) of the mixture and so lowers molecular mobility (Khougaz and Clas 2000, ref 11, and refs 12, 13). Yet inhibition has also been seen when Tg was unchanged or even lowered (refs 11, 14). Another view gives specific interactions, mainly hydrogen bonds between drug and polymer, the leading role (refs 15–20), and crystal growth measurements from the Taylor group had correlated growth inhibition with drug–polymer hydrogen bonding (Kestur and Taylor 2010, ref 21; Van Eerdenbrugh and Taylor 2010 and 2011, refs 22, 23) C2·C3. Note that the reference list prints three of these works twice (ref 11 = ref 18, ref 12 = ref 15, ref 14 = ref 16) X17.

Resveratrol (CAS 501-36-0, Figure 1a) is a phytochemical found as cis and trans isomers, the trans form being the more abundant and active; it has been suggested to have antiplatelet, antioxidant, antifungal, anticancer and cardioprotective properties (Baur 2010, ref 25) and to extend life span in yeast (ref 26). It is only moderately hydrophobic (log P 3.1) but poorly soluble in water, which the authors attribute in large part to its high melting point of 262 °C and strong lattice energy (refs 27, 28) M1·C4. Its tendency to crystallize is extremely high, which makes it a hard test case: if a polymer can keep resveratrol amorphous, the reasons should be visible. The authors state the hypothesis directly: disrupting resveratrol self-assembly through drug–polymer intermolecular interactions is critical both to form an amorphous dispersion and to keep it physically stable C5. They also ask a question that, in their words, the literature had neither confirmed nor refuted: does a polymer that makes an amorphous dispersion easy to form also keep it amorphous in storage C22?

To test this, they compared seven polymers chosen for their different hydrogen-bonding chemistry (Figure 1): two grades of poly(vinylpyrrolidone) (PVP K29/32 and PVP K-12), hydroxypropyl methylcellulose (HPMC), hydroxypropyl methylcellulose acetate succinate (HPMCAS, grade AS-MF), carboxymethyl cellulose acetate butyrate (CMCAB), poly(acrylic acid) (PAA) and Eudragit E100 (E100, a butyl methacrylate–dimethylaminoethyl methacrylate–methyl methacrylate copolymer, 1:2:1) M2–M8.

Background concepts

Crystallization tendency. Van Eerdenbrugh, Baird and Taylor (ref 29) sorted drugs into classes by whether they crystallize during or soon after fast solvent evaporation. A compound that crystallizes during or immediately after spin coating is a "rapid crystallizer"; such a compound cannot be studied as a neat glass, and its amorphous form exists only with an inhibitor C6.

X-ray amorphous. A powder is called X-ray amorphous when its X-ray powder diffraction (XRPD) pattern shows only a broad halo and no sharp peaks. Small crystalline fractions are invisible: in this paper the detection limit was of the order of 5–10% crystalline material, estimated from calibration samples spiked with crystalline resveratrol P4. "Amorphous" in what follows always means "no crystals detected at this limit".

Hydrogen-bond donors and acceptors. A hydrogen bond needs a donor (a hydrogen on an electronegative atom, here always an OH group) and an acceptor (a lone pair, for example on a carbonyl oxygen, an ether oxygen or an amine nitrogen). Resveratrol has three phenolic OH groups, which are strong donors, and only weak acceptors M1. In its crystal (Cambridge Structural Database code DALGON, ref 33) each hydroxyl oxygen takes part in two hydrogen bonds, building infinite chains along two crystal axes, and each hydroxyl hydrogen sits half the time in each of two positions, a disorder known as "flip-flop" hydrogen bonding (refs 34, 35) M1·C28. A polymer can only break this network if it offers resveratrol's OH groups an acceptor at least as attractive as the drug's own.

Etter's rules and the pKBHX scale. To rank the polymers, the authors follow the approach of Van Eerdenbrugh and Taylor (ref 23), built on Etter's hydrogen-bond rules (refs 36, 37): all good donors and acceptors are used in hydrogen bonds, six-membered intramolecular rings form first, and the best remaining donors and acceptors pair with each other C30. Acceptor strength is read from the pKBHX scale (ref 38), a measure of hydrogen-bond basicity derived from the 1:1 association constants of acceptors with 4-fluorophenol, tabulated for many drug-like groups. Each polymer group is assigned the value of a small model compound (1-methyl-2-pyrrolidone for the PVP amide, triethylamine for the E100 amine, ethyl acetate for esters, diethyl ether for ethers, ethanol for hydroxyls) and sorted into classes with fixed cut-offs (Table 4, footnote a) R17·E2:

In plain terms: a group with pKBHX above 2.25 is a very strong acceptor, one between 1.5 and 2.25 a strong one, and so on. Carboxylic acid groups have no tabulated acceptor value, so the authors gave their carbonyl the ester value of 1.07 (Table 4, footnote b) R17. For comparison, phenol, the closest model for resveratrol's own acceptor, has pKBHX −0.07 M1.

Reading hydrogen bonds in infrared spectra. When an OH group donates a hydrogen bond, its stretching band moves to lower wavenumber, and the stronger the bond the larger the shift. When a polymer carbonyl accepts one, a new carbonyl band appears a few tens of wavenumbers below the band of the free carbonyl, and the ratio of the two bands follows the fraction of bonded carbonyls (Tang et al. 2002, ref 20) C9. Shifts relative to the physical mixture are commonly used to infer that drug and polymer are mixed at the molecular level (Taylor and Zografi 1997, ref 12; Marsac et al. 2006, ref 31) C8. Because amorphous resveratrol could not be made, the authors had no amorphous reference spectrum; they followed how the bands change with composition instead C10.

Acid–base interaction. Resveratrol is a weak acid (phenol), and E100 carries basic dimethylamino groups. If a proton moves from phenol to amine, the two become an ion pair. Moustafine et al. (ref 32) assigned two E100 bands at 2820 and 2770 cm−1 to the non-ionized dimethylamino groups; their disappearance therefore signals that the amine has been engaged, which the authors read as an acid–base reaction R7·C15.

Glass transition and the two-thirds rule. The glass transition temperature (Tg) marks the change from a rigid glass to a mobile, rubbery state; a high Tg usually means low molecular mobility at storage temperature. Water lowers Tg (plasticization). Because amorphous resveratrol could not be isolated, its Tg was estimated with the empirical "2/3 rule" (Tg roughly two-thirds of the melting point on the kelvin scale; the paper names the rule without printing it), giving 83.7 °C M1·E1. Recomputed by the AI writer from the melting point of 262 °C, the rule gives about 83.6 °C, the printed value within rounding E1.

Full and partial crystallization. The authors distinguish dispersions that crystallize completely from those in which crystallinity rises to a plateau and then stays constant for months. The second pattern is called partial crystallization C26.

Methods in detail

The study has four strands, all on the same drug and polymer set: a formation screen, an infrared study of interactions, a storage study with Tg measurements, and a hydrogen-bond analysis of the polymers from the literature. A water sorption experiment on one dispersion was added in the discussion.

Contribution 1: attempts at neat amorphous resveratrol and a formation screen with polymers

The problem. Before comparing polymers, the authors needed a reference: amorphous resveratrol on its own. They tried spin coating solutions in ethanol, methanol, acetone and a 1:1 (by weight) mixture of dichloromethane and ethanol, varying concentration and spin speed; rotary evaporation from several solvents; and cryomilling the crystalline drug in a liquid nitrogen bath for up to 6 h. Melt quenching was not attempted, because resveratrol degrades on heating P2·R1.

What they did. All polymers were dried over phosphorus pentoxide for at least a week. Resveratrol and polymer were dissolved together in 1:1 dichloromethane–ethanol at dry-weight ratios from 5% to 75% drug in 5% steps; every solution was checked visually to be a single clear phase. The solvent was removed on a rotary evaporator with a 60 °C water bath, the product dried under vacuum for 24 h and then cryomilled in liquid nitrogen for a total of 4 min. XRPD (Shimadzu XRD-6000, Cu Kα, 5–35° 2θ at 8°/min, 0.04° steps) followed immediately; the diffractometer angle was checked each day against the silicon [111] reflection (28.423–28.463°), whose intensity was also used to normalize that day's data. A sample without sharp peaks was X-ray amorphous; the highest drug content that stayed X-ray amorphous is the polymer's formation limit P1·P4.

Why it is informative. Solvent evaporation is a severe test for a rapid crystallizer. The amount of polymer needed to suppress crystallization during this short step is a direct, if coarse, measure of how well each polymer interferes with nucleation and growth from a concentrating solution.

Contribution 2: infrared spectroscopy of drug–polymer interactions

What they did. Solutions with 5% to 60% drug in 1:1 dichloromethane–ethanol were dropped on thallium bromoiodide (KRS-5) windows and spun at 500 then 2500 rpm (18 and 30 s). Transmission spectra (Bio-Rad FTS 6000, DTGS detector) were collected from 4000 to 500 cm−1 at 4 cm−1 resolution with 128 scans, keeping the absorbance of the region of interest between 0.6 and 1.2 and purging with dry air P5. Two regions carry the argument: the resveratrol OH stretch (about 3550–2750 cm−1) and the polymer carbonyl stretch (about 1780–1610 cm−1). The compositions shown are 0, 10 and 30% drug for PVP K29/32, 0–60% for E100 and 0–25% for PAA (Figures 3–5); spectra of the HPMC, HPMCAS and CMCAB films are in the Supporting Information, which was not available here A3·R8.

A detail to watch. The captions of Figures 3 and 4 say panel (a) is the OH region and (b) the carbonyl region, and the text cites them that way, but the printed panels labelled (a) show the carbonyl region and those labelled (b) the OH region. The wavenumber axes settle which is which X1.

Contribution 3: storage stability and glass transitions

What they did. Dispersions were stored at five conditions: 0% RH over phosphorus pentoxide and 84% RH over potassium chloride, both at room temperature (22–25 °C, reported as 25 °C), and 0% RH over anhydrous calcium sulfate (Drierite) at 50, 90 and 120 °C. The two hottest conditions were added because the dispersions were stable at 50 °C, to force crystallization in the more stable ones. Storage lasted up to 400 days, with XRPD at intervals; the percent crystallized was estimated against calibration samples of each dispersion spiked with crystalline resveratrol, and physical mixtures served as fully crystalline references P3·P4.

Each dispersion contained 60% polymer, or more where 60% did not give an amorphous product: 70% for HPMCAS and CMCAB and 85% for PAA. The drug content of the stored samples therefore ranged from 40% down to 15% P3·X11.

Tg was measured by DSC (TA Q2000, nitrogen purge, 3–7 mg in pin-holed Tzero pans): a first heating to 120 °C to remove history, cooling at the maximum rate, and a second heating from 0 to 200 °C at 10 °C/min, from which the onset Tg was taken. The PVP dispersions needed modulated DSC (2 °C/min with ±1 °C every 60 s) to resolve Tg. Samples equilibrated at 84% RH were sealed hermetically, cooled to −50 °C and heated at 10 °C/min P6·P7·P8.

Contribution 4: a hydrogen-bond analysis of the polymers, and water sorption

For each polymer the authors listed every donor and acceptor group with its content in mmol per gram, its donor strength and its pKBHX acceptor class (Table 4) R17·P10. This turns the hypothesis into a prediction that can be set against the experiments. Separately, the water uptake of a 40% drug PVP K29/32 dispersion and of a physical mixture of the same composition, with crystalline drug, was measured gravimetrically (TA Q5000) at 25 °C from 0 to 95% RH in 5% steps, after drying at 50 °C; each step ended when the mass changed by less than 0.01% in 5 min or after 360 min P9.

Experimental analysis

Resveratrol alone is a rapid crystallizer

Every attempt failed. Spin-coated films crystallized during or immediately after coating whatever the solvent, concentration or speed; rotary evaporation gave crystals; even 6 h of cryomilling at liquid nitrogen temperature left the drug crystalline R1. The authors therefore class resveratrol as a rapid crystallizer and write that "it is clear" a crystallization inhibitor is necessary C6. A practical consequence runs through the rest of the paper: there is no measured Tg and no amorphous reference spectrum for the pure drug M1·C10.

Formation: how much polymer each one needs

Table A. Formation limits after rotary evaporation from 1:1 dichloromethane–ethanol (text and Figure 2), with the polymer properties from Table 4 (acceptor and donor classes; the strongest acceptor group was picked by the AI writer) and Table 5 (polymer Tg) R2·R17·M2–M8.

The spread is fourfold, from 15% drug with PAA to 60% with E100 R2. The authors single out E100 as the best inhibitor during solvent evaporation and note that PAA "required so much polymer that there was very little resveratrol" left C7. In their discussion the authors note that, in general, less polymer was needed for the polymers that formed stronger hydrogen bonds with resveratrol C42. In Table A the only strong and very strong acceptors (E100 and PVP) need the least polymer, and the polymer whose only groups are carboxylic acids (PAA) needs the most. The authors relate the PAA result to the earlier finding that acidic polymers are poor crystallization inhibitors for acidic compounds; resveratrol is a weak acid (Van Eerdenbrugh and Taylor 2011, ref 23) C44.

The authors add that faster evaporation by spin coating "consistently" needed less polymer, with the same order of polymers, but give no numbers R3. Note by the AI writer (not discussed in the paper): the IR section reports that the highest PVP K29/32 film that could be spin coated held 30% drug, because above that resveratrol crystallized at the tip of the pipette, which "seems to act as a nucleation site"; that is below the 45% reached by rotary evaporation, so the spin-coating statement cannot be checked from the data shown R4·C11·X5.

Interactions: what the infrared spectra show

Table B. Infrared band positions: OH from Table 2 (40% drug unless stated), carbonyl from Table 1, crystalline value and observations from the text R5·R6·R7·R8·R9·R10·R11.

PVP K29/32. With 30% drug the resveratrol OH maximum lay at about 3193 cm−1, below the 3238 cm−1 of the crystal and the lowest of any system, and the band was the broadest. The PVP carbonyl at 1682 cm−1 gained a shoulder at 1656 cm−1, 26 cm−1 lower, which was larger at 30% than at 10% drug (Figure 3) R5·R10. The authors read the OH shift as "consistent with" stronger drug–polymer than drug–drug hydrogen bonds and as "suggesting" that PVP forms the strongest hydrogen bonds of the set; the new shoulder is the bonded carbonyl C12·C13.

E100. The OH maximum moved steadily from 3425 cm−1 at 20% drug to 3395, 3391, 3377 and 3370 cm−1 at 30, 40, 50 and 60% (peak labels in Figure 4), and the band broadened slightly, less than with PVP R6. A new carbonyl band appeared at 1701 cm−1 beside the ester band at 1730 cm−1 (labelled 1705 cm−1 on the 60% trace) and grew with drug content R6·R10. Most striking, the two E100 bands at 2820 and 2770 cm−1, which belong to non-ionized dimethylamino groups, faded quickly as drug was added and were gone at 50% drug R7. The authors take this as evidence of an acid–base reaction between the phenol and the amine, in addition to weak hydrogen bonds to the ester carbonyl; the abstract calls it "somewhat surprising" C14·C15.

Cellulose derivatives. With 60% polymer the OH maximum was 3358 cm−1 with HPMC, 3369 cm−1 with HPMCAS and 3399 cm−1 with CMCAB. The band broadened with HPMC and HPMCAS but not with CMCAB, and the carbonyl bands of HPMCAS and CMCAB moved from 1744 to 1738 and from 1750 to 1730 cm−1 with 30% drug, the shifted band growing relative to the free one as polymer content rose. By OH position, resveratrol forms the weakest hydrogen bonds with CMCAB and similar ones with HPMC and HPMCAS R8·R10·R11·C17.

PAA. PAA's own strong OH absorbance hid the resveratrol band. In the carbonyl region the main band at 1711 cm−1 changed shape slightly but did not shift significantly (1713 cm−1 with 25% drug), and a small new band appeared at 1760 cm−1 (Figure 5). "These data suggested" limited resveratrol–PAA hydrogen bonding R9·R10·C18.

The two rankings. From the OH positions the authors rank hydrogen-bond strength PVP K29/32 > HPMC > HPMCAS > E100 > CMCAB, and they note that the carbonyl data agree for all polymers that have a carbonyl C19. Adding the ionic interaction, which they regard as "typically stronger" than hydrogen bonding, they suggest an order of crystallization-inhibiting performance E100 > PVP K29/32 > HPMC > HPMCAS > CMCAB > PAA C20·C21. Two cautions apply. The PVP value comes from a 30% drug film and the others from 40%; since the E100 band moves by 55 cm−1 between 20 and 60% drug, loading matters, although the 200 cm−1 gap between PVP and E100 is far larger X12. And E100 ranks only fourth by OH shift, below HPMC and HPMCAS; its first place in the performance ranking rests on the acid–base evidence, not on hydrogen-bond strength X13.

Storage: who stays amorphous

Table C. Days before crystals were seen by XRPD, from Table 3. ">X": monitoring stopped at X days with no crystals; "1": crystals at the first sampling (day 1), the dispersion going on to crystallize fully or only partly R12.

The storage results split the polymers cleanly at 84% RH. The two PVP grades and E100 kept their dispersions amorphous for the whole monitoring period, 100 to 400 days. Every cellulose derivative and PAA held crystals at the first sampling and went on to crystallize completely R12. The two PVP grades behaved alike, although PVP K-12 has a much lower molecular weight and Tg (91 against 159 °C); the authors call the PVP dispersions "remarkable" in their stability R12·M2·M3·C24. Dry storage at 25 °C caused no crystallization in any dispersion within 100–400 days, and at 50 °C only the HPMC dispersion crystallized, partly R12.

Two qualifications come from the table itself. PVP K-12 was followed for only 100 days at every condition, and PVP K29/32 for 200 and 100 days at 90 and 120 °C, so the text's "for up to 400 days" applies to K29/32 at the three milder conditions only X6. And the dispersions differ in drug content (40% with 60% polymer, 30% with 70%, 15% with 85%) and in how far they sit below their own formation limit: HPMC and PAA were stored exactly at the highest loading that had formed an amorphous product, E100 20 points below it X11.

Heat and humidity crystallize in different ways

Table D. Extent and time course of crystallization, from the text and Figures 6 and 7; percentages are XRPD estimates against calibration samples (approximate) R13·R14·R15·R16.

Two patterns emerge. At high temperature in dry air, crystallization often started within a day and then stopped, leaving a constant crystalline fraction for months; the higher the temperature, the larger that fraction. Under humidity, crystallization ran to completion R13·R14. The authors sort the outcomes into three behaviours, none (PVP), complete (always for samples that crystallized at 84% RH) and partial (only at 0% RH and elevated temperature), and conclude that high humidity promoted crystallization "to a much greater extent" than heat C26·C39. This holds at 50 and 90 °C. At 120 °C, however, the HPMC, HPMCAS, CMCAB and PAA dispersions crystallized completely as well X8.

Some figure details do not match the text. The text cites Figure 6 for E100 crystals after 1 day at 90 °C, but the earliest stored trace in Figure 6a is at 3 days X4. Figure 7b is captioned "after 100 days", whereas the text gives the HPMC values after 275 days, and its lowest trace is labelled "0 °C", a temperature not used in the study X3. The smallest crystallinity values, about 5% for HPMC at 50 °C and about 10% for E100 at 90 °C, lie at the stated XRPD detection limit of 5–10% X15.

Glass transition temperatures do not explain the order

Table E. Onset Tg by DSC, from Table 5. * Crystallization seen by DSC above the Tg. The estimated Tg of amorphous resveratrol is 83.7 °C M2–M8·R18·M1.

If the polymer with the highest Tg were the best inhibitor, stability would follow PVP K29/32 > HPMC > CMCAB > PAA > HPMCAS > PVP K-12 > E100. The results "clearly do not follow this order": E100, the polymer with by far the lowest Tg, kept its dispersion amorphous at 84% RH, and PVP K-12 did as well as PVP K29/32. The authors conclude that polymer Tg "does not appear to be the only factor" C33·R12. The dry dispersions themselves lie between 63 and 85 °C, a narrow window, and after equilibration at 84% RH the three dispersions that stayed amorphous still had Tg values of 48–58 °C, well above the 25 °C storage temperature R18.

The authors also write that E100 dispersions have Tg values "similar to or lower than" those of the cellulose polymers. Note by the AI writer: in Table 5 this is true for the neat polymers (41 °C against 111–145 °C) but not for the dispersions, where E100 (77 °C) sits above HPMC (75), HPMCAS (68) and CMCAB (63 °C) C34·X9. Either way, Tg cannot explain why E100 and PVP do best.

Water uptake of the PVP dispersion

Table F. Water sorption at 25 °C of 40% resveratrol with 60% PVP K29/32, as a physical mixture with crystalline drug and as an amorphous dispersion; read from Figure 9 (approximate) R19.

At every humidity the dispersion took up less water than the physical mixture, about a third as much at 95% RH R19. The authors present this as evidence that resveratrol–polymer hydrogen bonds suppressed moisture absorption C37. Note by the AI writer (not discussed in the paper): the physical mixture holds the drug as crystals, so its uptake reflects mainly the free polymer; the comparison shows that mixing at the molecular level lowers uptake, but it does not by itself identify hydrogen bonding as the cause, and no amorphous drug-free or other-polymer reference was measured X18.

Discussion

The authors' explanation: a better hydrogen-bond partner

The discussion starts from the crystal. Resveratrol's three hydroxyls form an extensive three-dimensional hydrogen-bond network in a very planar crystal (torsion angles of −3° and 8° for the rings and the bridging CH=CH group); the double bond restricts rotation, the molecule has few low-energy conformers, and the melting point is high. The high crystallization tendency, the authors write, "can be attributed" to these features, which "will promote" crystallization (Figure 8) C27·M1. A polymer should then inhibit crystallization if it gives resveratrol a better hydrogen-bonding partner than resveratrol itself ("it is reasonable to expect") C29.

Applying Etter's rules, the authors note that the only donors in the system are OH groups; resveratrol's phenolic OH is a stronger donor than the aliphatic OH of the cellulose polymers, and the carboxylic acids of PAA, HPMCAS and CMCAB are the strongest polymer donors C30. Since resveratrol is a strong donor and a weak acceptor, the best partner is the polymer with the strongest acceptor: PVP (very strong), then E100 (strong). The others offer only medium acceptors, and the polymers with strong donors, "in particular PAA and CMCAB", "may interact poorly" because resveratrol has no acceptor that can compete with theirs C31. The chain they propose is:

The authors find these predictions "in excellent agreement" with the storage results and "in line with" the IR inference that PVP forms the strongest hydrogen bonds C32. In the abstract and conclusions they put it more cautiously: inhibitor performance "could be correlated" with polymer structure, stability was "suggested" to be linked to the type and strength of interactions, and effectiveness "is rationalized" on that basis; they conclude that "it thus appears" necessary, for a compound like resveratrol, to select the polymer for its ability to disrupt molecular self-assembly C45·C47.

Why E100 does so well

By hydrogen-bond strength alone E100 ranks fourth, yet it formed the most concentrated dispersion and resisted humidity. The authors attribute this to the acid–base reaction seen in the infrared: it "could be responsible" for the easier formation, and the moisture stability is "likely due to" it C16·C38. That E100 needed slightly less polymer than PVP is put down "perhaps" to solvent effects or to the ionic interaction C43. None of these explanations was tested separately.

Humidity against temperature, and the felodipine contrast

The authors find the moisture stability of the PVP dispersions "very interesting", because PVP is very hygroscopic (Taylor et al. 2001, ref 39) and because earlier work on felodipine had shown the opposite: felodipine–PVP dispersions crystallized much faster under humidity, whereas felodipine–HPMCAS dispersions resisted (Rumondor et al. 2009, ref 8) C35·R20. Felodipine is the more hydrophobic drug (log P 4.5 against 3.1). They offer two "possible explanations" for the PVP behaviour: strong drug–polymer hydrogen bonds that suppress water sorption (Rumondor et al. 2010, ref 40), or the absence of moisture-induced amorphous–amorphous phase separation because resveratrol is more hydrophilic (Rumondor et al. 2011, ref 41) C36. Figure 9 is offered for the first C37; the second was not examined.

More generally, the dispersions other than PVP and E100 were "in general" much less stable to humidity than to heat: moisture caused rapid, complete crystallization, while heat caused partial crystallization followed by long plateaus. The authors say they "do not fully understand" the size of the humidity effect, and that the observations "highlight the difficulty of selecting appropriate conditions for accelerated stability testing" C39·C40·C41.

Formation and stability go together, mostly

The paper's second question was whether a polymer that makes a dispersion easy to form also keeps it stable. The authors answer that "in general" it does: relatively high polymer contents were needed in every case, and the polymers that maintained solid-state stability also facilitated formation C42. At the ends of the ranking this is clear, with E100 and PVP best at both and the cellulose polymers and PAA worse at both, especially under humidity R2·R12.

Note by the AI writer (not discussed in the paper): within the middle group the two orders diverge. HPMC formed dispersions with more drug than HPMCAS and CMCAB (40 against 35%) and ranks above them by OH shift, yet its dispersion was the only one to crystallize at 50 °C, where HPMCAS and CMCAB stayed amorphous for more than 100 days. PAA, ranked last and called "an extremely poor stabilizer", stayed amorphous for more than 175 days at 90 °C, where the three cellulose dispersions crystallized partly within a day. Part of this may come from the different drug contents of the stored samples (15% for PAA, 30% for HPMCAS and CMCAB, 40% for HPMC) X10·X11·R12.

What it suggests for formulation

The authors' own recommendation is general: for a compound with a high crystallization tendency, choose the polymer by its ability to disrupt drug self-assembly through drug–polymer interactions C47; acidic polymers are a poor choice for acidic drugs C44; and humidity must be part of any stability assessment C41.

Note by the AI writer (not stated in the paper): for a drug that is a strong hydrogen-bond donor and a weak acceptor, a table like Table 4 is a cheap first screen, because it can be filled from the polymer structure and the pKBHX database before any experiment. In this study a single day at 84% RH separated the polymers as clearly as months of dry storage, while hot storage could make a dispersion look stable because crystallinity stopped at a plateau. Polymer Tg was a poor guide: the polymer with the lowest Tg, E100, was among the best.

Limitations

Limitations stated by the authors. The authors acknowledge few limits explicitly. They could not obtain amorphous resveratrol, so they had no amorphous reference spectrum and no measured Tg for the drug C10·M1. For PAA the OH region could not be used because of the polymer's own absorbance R11. They state that they "do not fully understand" why humidity has so much more effect than temperature C40, and they present the explanations for the moisture stability of the PVP dispersions as "possible" C36.

The AI writer's own judgement (1): what the data support. The strongest results are direct and large: resveratrol alone always crystallized; the formation limits span 15 to 60% drug in a 5% step screen; and at 84% RH the PVP and E100 dispersions stayed amorphous for 100–400 days while every other dispersion held crystals after one day R1·R2·R12. The argument against Tg as the controlling factor is also solid, since the lowest-Tg polymer (E100) and the low-Tg PVP grade (K-12) were among the most stable R12·M3·M8. The link to interaction strength is weaker. Hydrogen-bond strength was ranked from band positions and literature acceptor values, not measured as an energy, and it was never related to crystallization rates; it fits the top two polymers and PAA in formation, but not the order of the middle group in storage X10·X13.

The AI writer's own judgement (2): statements that go beyond the data. Several summary sentences claim more than Table 3 shows. PVP K-12 was monitored for 100 days, not 400 X6. The conclusions list E100 among the dispersions resistant at 90 and 120 °C, but it partly crystallized within a day at both temperatures X7. The contrast "moisture complete, heat partial" fails at 120 °C X8. The E100 dispersion Tg is higher, not lower, than those of the cellulose dispersions X9. These do not change the main conclusions, but a reader should take the table rather than the summary sentences.

The AI writer's own judgement (3): design and reporting. The stored dispersions carried 15–40% drug and sat at different distances below their own formation limits, so the stability ranking mixes the effect of the polymer with that of loading X11. Infrared positions were compared at 30% drug for PVP and 40% for the others X12. Replicate numbers and sampling schedules are not given, and several crystallinity values lie at the XRPD detection limit X15·X16. Figure panels and labels disagree with the text in four places (panel letters in Figures 3 and 4, 124 against 126 days, the "0 °C" label and "100 days" caption of Figure 7b, the missing day-1 trace in Figure 6a) X1·X2·X3·X4. Finally, the motivation for an amorphous form, faster dissolution and higher transient solubility, was not measured.

Conclusion

Resveratrol is a rapid crystallizer that cannot be made amorphous without a polymer R1·C6. Among seven polymers, PVP and Eudragit E100, the only ones with strong or very strong hydrogen-bond acceptors and no competing donors, formed amorphous dispersions with the least polymer (45% and 60% drug) and were the only ones whose dispersions survived 84% RH; HPMC, HPMCAS, CMCAB and PAA needed more polymer and crystallized under humidity, and PAA was the poorest former R2·R12·R17. Infrared spectra point to the strongest hydrogen bonds with PVP and to an acid–base reaction with E100 C12·C15. Polymer Tg did not predict the order C33. Humidity was far more damaging than heat at 50–90 °C, where heat produced only partial crystallization that stopped at a plateau C39·X8. The authors rationalize polymer performance by the type and strength of drug–polymer interactions, an explanation that fits the extremes well but rests on a qualitative ranking C45·X10.