Resveratrol solid dispersions in water: crystallization, drug–polymer complexes and how much of the drug is really free

Background and motivation

An amorphous solid dispersion (ASD) holds a poorly water-soluble drug in the amorphous state inside a polymer. On dissolution it can produce a supersaturated solution, a drug concentration above the solubility of the crystal, and this is how ASDs raise bioavailability (Leuner 2000; Serajuddin 1999). The approach works only if the matrix stays amorphous during dissolution and the high concentration lasts long enough for absorption; recrystallization to the stable crystal removes the solubility advantage (Hancock 1997). Polymers can change nucleation rates, growth rates and crystal morphology in supersaturated solutions (Alonzo 2012 and refs 5–6), and the polymer chosen strongly affects physical stability in storage (refs 7–10). The authors therefore require a useful polymer to prevent crystallization in three settings: the solid matrix, the hydrated matrix and the supersaturated solution.

The introduction raises a second, less studied issue: solubilization by the polymer. If dissolved polymer raises the thermodynamic solubility of the crystalline drug, the same measured concentration means less supersaturation and a smaller driving force for membrane transport. Such solubilization by ASD polymers "may sometimes occur" (Loftsson 1996) but had not been widely reported. A higher concentration from a dissolving ASD can therefore come from supersaturation, from solubilization, or from both. The distinction matters because a supersaturated solute has a higher chemical potential than its crystal, which drives both crystallization and faster membrane transport (Mullin 2001; Raina 2014). When drug and polymer form a complex, the total concentration may rise but the free drug does not, so transport is not faster (Dahan 2010; Loftsson 1994).

The model compound is resveratrol, a plant polyphenol whose trans isomer (Figure 1) is thought to be the more abundant and biologically active form. It is moderately hydrophobic (logP 3.1) and poorly soluble in water (38.8 μg/ml at 25 °C and pH 6; Sun 2008), which the authors attribute largely to its high melting point of 262 °C and strong crystal lattice. An amorphous form should therefore give large gains in dissolution rate and apparent solubility. The molecule carries three phenol groups, which later matters for its behaviour with one polymer. In the classification of Van Eerdenbrugh 2010 it is a rapid crystallizer, so any amorphous form of it is at risk.

The study builds on the group's storage study of the same dispersions (Wegiel 2013, ref 7), where stability depended on the specific drug–polymer interactions. Poly(vinylpyrrolidone) (PVP), which forms strong hydrogen bonds with resveratrol, and Eudragit E100 (E100), which takes part in an ionic interaction with it, prevented crystallization at elevated relative humidity (RH). Three cellulose derivatives form much weaker hydrogen bonds with resveratrol, and their dispersions crystallized rapidly at high RH: hydroxypropyl methyl cellulose (HPMC), hydroxypropyl methyl cellulose acetate succinate (HPMCAS) and carboxymethyl cellulose acetate butyrate (CMCAB) (refs 7, 16, 17). The authors reasoned that these less stable dispersions "might also be expected" to crystallize rapidly from the matrix in dissolution media.

The paper tests one hypothesis: dispersions whose interactions gave physical stability at elevated RH will also resist crystallization during dissolution and so give higher resveratrol concentrations than dispersions that crystallize rapidly. The broader aim is to understand the link between crystallization, drug–polymer complexation and solution concentration. Four kinds of measurement were combined: crystal solubility in polymer solutions, in situ Raman monitoring of crystallization in stirred slurries, drug and polymer concentrations at the end of each slurry, and membrane flux in a diffusion cell. The data did not support the hypothesis as stated: the two dispersions that stayed amorphous gave the lowest concentrations, and the highest came from a dispersion that partly crystallized.

Background concepts

Crystal solubility, supersaturation and solubilization. The reference point throughout is the equilibrium solubility of crystalline resveratrol, measured as 42 ± 6 μg/ml in 50 mM phosphate buffer at pH 6.8 (PB 6.8) and 37 °C. A solution above this value is supersaturated only to the extent that the extra drug is free. A dissolved polymer can bind drug and so raise the total amount that dissolves in equilibrium with the crystal; this is solubilization, and the authors state that for polymeric systems "solubilization results from the formation of drug-polymer complexation". To separate the two effects they measure crystal solubility in buffer containing the same amount of dissolved polymer as the dispersion solution, and define apparent supersaturation as the excess above that value (note to Table III, card equation E2):

Here $C_{\mathrm{o}\mathrm{b}\mathrm{s}}$ is the resveratrol concentration measured in the dispersion solution (μg/ml), $c_{\mathrm{p}\mathrm{o}\mathrm{l}}$ is the dissolved polymer concentration in the same solution (mg/ml), and $S_{\mathrm{c}\mathrm{r}\mathrm{y}\mathrm{s}\mathrm{t}}(c_{\mathrm{p}\mathrm{o}\mathrm{l}})$ is the solubility of crystalline resveratrol in buffer containing that much polymer, taken from Figure 2. In plain terms: the excess counts only the part of the concentration that the crystal could not reach in the same medium. The paper prints this relation in words. The excess is still a total-drug quantity and does not say how much of the extra drug is free.

Free drug and drug–polymer complexes. A drug molecule bound to a dissolved polymer chain counts in the total concentration measured by ultraviolet (UV) spectroscopy but is not free. The paper distinguishes two kinds of complex, drawn in Figure 11. A soluble complex stays in solution and raises the total concentration without raising the free concentration; an insoluble complex precipitates and removes drug from solution. For PVP and phenols, the authors cite older work (Kabadi 1966; Molyneux 1961; Bandyopa 1972) that found a critical concentration for precipitation. At low phenol levels little phenol is bound and the complex stays dissolved; as the phenol-to-PVP ratio rises, phenol displaces hydrogen-bonded water from the polymer, and the desolvated complex precipitates. The drug-to-polymer ratio in solution, not the polymer alone, decides which regime applies.

Membrane flux as a read-out of free drug. In a side-by-side diffusion cell, a donor and a receiver solution are separated by a membrane, and the rate at which drug appears in the receiver is measured. Following Raina 2014, the authors take this rate to be set by free drug: faster flux is "consistent with higher concentrations of free drug", and the degree of supersaturation is "directly proportional to the flux" (card equation E3):

Here $J$ is the flux across the membrane (μg/min) and $\sigma$ is the degree of supersaturation of the donor, the free drug concentration relative to the crystal solubility. In plain terms: if a saturated crystal suspension gives a certain flux, a donor with twice that flux holds about twice as much free drug, however high its total concentration is. The paper's reference is a saturated suspension of crystalline resveratrol in buffer without polymer, which has a supersaturation of one by definition.

The Raman peak ratio as a crystallinity indicator. Two aromatic ring-stretching bands of resveratrol, at 1630 and 1605 cm⁻¹, change their relative height when the drug crystallizes (Figure 4, crystalline resveratrol against a 40:60 resveratrol/PVP dispersion). The intensity ratio is 0.76 for the crystal and about 1.1 for the amorphous dispersions, so a falling ratio means crystallization. The authors use the ratio "qualitatively", because the amorphous value "differed slightly for all the solid dispersions and no pure amorphous reference was available". A plateau therefore shows more or less crystallization but cannot be converted into percent crystallinity.

How the five polymers behave in the medium. PVP (grade K29/32) is described as extremely soluble in pH 6.8 buffer, and HPMC (grade 606) as highly soluble. HPMCAS (grade AS-MF) dissolves at pH 6.8 but is essentially insoluble in hydrochloric acid solution at pH 1.2 (HAS 1.2), which the authors use to block drug release in one experiment. E100 and CMCAB have limited solubility at pH 6.8; for E100 the authors write that the polymer "does not dissolve at pH 6.8". A polymer that does not dissolve can neither carry drug into solution nor solubilize it.

Methods in detail

The work falls into three contributions. The first measures what dissolved polymers do to crystalline resveratrol, with no dispersion involved: how much they raise its solubility and, for PVP, how drug and polymer behave together in solution. The second slurries the dispersions and follows crystallization in situ while varying polymer type, polymer content and pH, then compares the 24 h concentration with the crystal solubility at the measured level of dissolved polymer. The third varies the amount of one HPMCAS dispersion in the medium and uses membrane flux to separate free from bound resveratrol. The flow below shows how the three connect.

Contribution 1: What dissolved polymers do to crystalline resveratrol

① Problem. To judge whether a dispersion solution is supersaturated, one needs the right reference: if a dissolved polymer raises the crystal's solubility, comparison with 42 μg/ml in plain buffer overstates supersaturation. For PVP a further difficulty appeared early. A first series with rising PVP concentration showed "no clear trends" in resveratrol solubility and a gel-like precipitate, "suggesting the formation of an insoluble complex", so no simple solubility curve could be measured for PVP.

② What they did. For the solubility study, 20 mg of crystalline resveratrol was dispersed in 20 ml of PB 6.8 containing pre-dissolved HPMC or HPMCAS at 5, 10 or 20 mg/ml; E100 and CMCAB were used as saturated polymer solutions because both dissolve poorly at pH 6.8. After 24 h at 37 °C the samples were centrifuged (40,000 g, 15 min), and the supernatant was measured by UV at 305 nm with a 1 cm dip probe, calibrated from 1.5 to 6.5 μg/ml; HPMC and HPMCAS did not affect the spectrum. For PVP the design was turned around: PVP was held at 20 mg/ml in 18 ml of buffer, and 11.84 to 198.67 mg of crystalline resveratrol was added, 658 to 11,037 μg/ml nominal (Table II), again for 24 h at 37 °C. PVP shifted the UV spectrum with both PVP and resveratrol concentration, so a dedicated calibration at the peak trough (258 nm) was built by varying both. The precipitate was examined by X-ray powder diffraction (PXRD; the paper also writes XRPD) and, after isolation and drying, by infrared spectroscopy (IR). Replicate numbers are not printed.

③ Why it works. Measuring crystal solubility in the presence of polymer gives the correct baseline for each polymer level, so the dispersion data can later be corrected for solubilization. The PVP titration holds polymer constant and varies drug, the variable that, according to refs 26–27, decides whether a phenol–PVP complex stays dissolved or precipitates, so one series covers both regimes. PXRD and IR of the precipitate test whether it is recrystallized drug or a drug–polymer solid.

④ How it differs from earlier work. Because solubilization by ASD polymers had "not been widely reported", dispersion concentrations were usually compared with the crystal solubility in plain buffer. Here the polymer-dependent solubility is measured, and the dissolved polymer in each dispersion solution is used to apply the correction. The PVP titration takes the classic phenol–PVP complexation experiment from the polymer literature (refs 26–28) and applies it to a compound of formulation interest.

Contribution 2: Following crystallization in slurries and correcting the end concentration

① Problem. The storage study ranked the polymers under humidity. Whether that ranking holds in water, where the matrix hydrates, the polymer may dissolve and the drug may crystallize in the matrix or from solution, was not known. A single end-point concentration cannot show when or where crystallization happened, or whether a high concentration reflects supersaturation at all.

② What they did. Resveratrol and polymer (dried over P₂O₅ for at least a week) were dissolved in 1:1 (by weight) dichloromethane:ethanol, rotary-evaporated with a 60 °C water bath, vacuum-dried for 24 h, cryo-milled for 4 min under liquid nitrogen and sieved to 53–250 μm; all dispersions were X-ray amorphous before use. For each slurry, about 300–500 mg of solid was added to 5 ml of PB 6.8 or HAS 1.2 at 37 °C, with a 785 nm Raman probe above the slurry. Spectra were taken every 5 min for the first hour, then at widening intervals up to hourly, to 24 or 48 h (n = 3). Three designs were used: polymer type, all five polymers at 35 wt% resveratrol (65% polymer, Figure 5); polymer content, cellulosics only, with HPMCAS at 85, 75 and 65%, CMCAB at 95, 90, 85 and 65% and HPMC at 80, 70 and 60% (Figure 6); and pH, a 25 wt% resveratrol HPMCAS dispersion in PB 6.8 and HAS 1.2, with polarized light microscopy (PLM) after 24 h (Figures 8 and 9).

At the end of each run the dried residue was checked by XRPD. The supernatant (40,000 g, 15 min) was analysed for resveratrol by UV and, for HPMC and HPMCAS dispersions, for dissolved polymer by high performance liquid chromatography (HPLC) with an aerosol-type detector (Table I). The HPMC calibration ran from 2 to 10 mg/ml with a limit of quantification (LOQ) of 2.6 mg/ml, the HPMCAS calibration from 4 to 10 mg/ml with an LOQ of 3.4 mg/ml. The measured polymer level gave the crystal solubility from Figure 2 and hence the apparent supersaturation (E2). One condition is unclear: the Methods give 300–500 mg of solid in 5 ml, about 60–100 mg/ml, while the Results describe the polymer comparison "at a solid dispersion loading of about 20 mg/ml" (extractor note X6). This matters because Contribution 3 shows that loading changes the concentration strongly.

③ Why it works. The probe reads crystallinity in the stirred slurry without sampling, so changes in the first hour are visible, and XRPD of the residue gives an independent check. The polymer assay makes the solubilization correction possible. The pH design is a control on where crystals form: at pH 1.2 HPMCAS does not dissolve, so drug cannot be released and any crystallization must happen in or on the matrix. If the kinetics at pH 6.8 look the same, dissolution is probably not needed for crystallization. CMCAB, also poorly soluble at pH 6.8, gives a second case of the same kind.

④ How it differs from earlier work. The storage study (ref 7) exposed powders to humidity and temperature. This study places the same dispersions in water and measures the solid and the solution together, coupling a time-resolved solid-state signal with a polymer-corrected solution read-out instead of relying on a dissolution concentration alone.

Contribution 3: Loading series and membrane flux

① Problem. Even after the solubilization correction, the excess counts total drug. It cannot show whether the high concentrations from HPMCAS dispersions are free resveratrol, which would drive absorption, or resveratrol bound to dissolved HPMCAS, which would not.

② What they did. Weighed amounts of a 15:85 resveratrol:HPMCAS dispersion, about 6 to 35 mg/ml (read from Figure 10), were stirred in 32 ml of PB 6.8 for 24 h at 37 °C. Resveratrol and HPMCAS in solution were measured (Figure 10a), and the same mixtures served as donors in a side-by-side diffusion cell (PermeGear) with a regenerated cellulose membrane of 6–8 kDa molecular weight cut-off (MWCO) and a 30 mm orifice (7.065 cm²). Each 34 ml compartment held 32 ml at 37 °C; the receiver held PB 6.8 under sink conditions, below one-third of the crystal solubility. UV dip probes in both compartments followed the concentrations continuously, and flux was calculated from the receiver curve (card equation E1):

Here $V$ is the receiver volume (32 ml), $C_r$ is the receiver concentration (μg/ml) and $t$ is time (min); the slope comes from linear regression over the linear region. In plain terms: the flux is how many micrograms of resveratrol cross the membrane per minute, the rate of rise of the receiver concentration times the receiver volume. The reference donor was a saturated suspension of crystalline resveratrol in PB 6.8 without polymer, stirred for 24 h at 37 °C.

③ Why it works. By E3, flux reports free drug, and the crystal suspension fixes the flux at a supersaturation of one. Varying the loading changes total resveratrol and dissolved polymer together; if total resveratrol rises while flux falls, the extra resveratrol cannot be free. Note by the AI writer (not discussed in the paper): the printed molecular weight of HPMCAS (Mw 17,000) is above the 6–8 kDa cut-off, so polymer-bound resveratrol would be expected to stay in the donor; the paper does not make this argument.

④ How it differs from earlier work. The polyphenol–cellulose dispersion studies the authors cite (Li 2013, refs 32–36) reported high solution concentrations. This paper adds a free-drug measurement to the same kind of system, using the diffusion-cell approach of Raina 2014, and compares the result with felodipine ASDs, whose flux reached ten times that of the crystal (Alonzo 2011).

Experimental analysis

Setup in brief. One drug, trans-resveratrol, was tested with five polymers: PVP K29/32 (Mw 58,000), E100 (Mw 47,000), HPMC 606 type 2910 (Mw 35,600), HPMCAS AS-MF (Mw 17,000) and CMCAB 641-0.5 (Mw 35,000). Nearly all experiments used PB 6.8 at 37 °C. Solution data are end points after about 24 h; crystallization was followed over 24–48 h. Only Tables II, III and IV of the paper print numbers; in the tables below, ≈ marks values read from a figure.

Main result 1: HPMCAS raises the crystal's solubility steeply, HPMC only slightly.

Table A. Solubility of crystalline resveratrol in PB 6.8 at 37 °C after 24 h with dissolved HPMC or HPMCAS (read from Figure 2, approximate, except values stated in the text). Saturated E100 and CMCAB solutions gave no increase; no values are printed.

HPMCAS raised the crystal solubility about tenfold at 10 mg/ml and roughly thirtyfold at 20 mg/ml, a rise that steepens with polymer concentration. HPMC changed little up to 10 mg/ml and reached about twice the plain-buffer value at 20 mg/ml. The authors describe an increase "above a threshold polymer concentration", with a "very large impact" for HPMCAS and "more modest" increases for HPMC. One point cannot be taken at face value: the method adds 20 mg of crystal to 20 ml (1000 μg/ml nominal), yet Figure 2 plots about 1230 μg/ml at 20 mg/ml HPMCAS, so either the amount added or the plotted value is wrong (X5). By the paper's own definition the HPMCAS rise is solubilization, the first sign that complexation will matter.

Main result 2: in PVP solution resveratrol first dissolves well, then precipitates with the polymer.

Table B. Crystalline resveratrol added to 20 mg/ml PVP in 50 mM PB 6.8, 24 h at 37 °C (Table II, transcribed in full; precipitate column from Figure 3 and the text).

Every measured value lies far above the crystal's 42 μg/ml; at the peak, 1610 μg/ml is close to forty times that value. Yet adding more crystal lowered the dissolved resveratrol, down to 99 μg/ml. The precipitate was non-crystalline by PXRD, and IR of the dried precipitate showed both resveratrol and PVP ("data not shown"). The authors read the rise as soluble complex at certain resveratrol:PVP ratios and the fall as an insoluble complex removing resveratrol from solution: "it is clear that both soluble and insoluble complexes are formed", and "it can be inferred" that the increase is due to soluble complex. Free resveratrol was not measured. At the lowest addition, the measured value exceeds the nominal one where complete dissolution is described, which implies at least about 8% error in UV values in PVP solutions (X12). The concentrations printed on the vial photographs in Figure 3 are about 0.91 times the Table II values, while the plotted points match the table (X4).

Main result 3: staying amorphous did not raise the concentration.

Table C. 35 wt% resveratrol dispersions slurried in PB 6.8 at 37 °C. Raman 1630/1605 cm⁻¹ ratios read from Figure 5 (approximate; cellulosic last points at about 1260 min; n = 3); solution columns from Table III.

The three cellulosic curves dropped within the first hour and then levelled off; "complete crystallization was never reached". CMCAB crystallized most, then HPMCAS, then HPMC, and XRPD of the residues agreed. The PVP and E100 dispersions stayed amorphous, as their powders had at high RH. The concentration ranking did not follow the stability ranking. The two amorphous dispersions gave 42 and 46 μg/ml, no more than the crystal, and the PVP dispersion "did not readily dissolve". The partly crystallized HPMCAS dispersion gave 458 μg/ml, about eleven times the crystal value, and HPMC gave 62 μg/ml. In the abstract the authors attribute the lack of supersaturation from PVP and E100 dispersions "most likely" to resveratrol–polymer complexation.

For the cellulosics, the dissolved polymer decides how much of the gain is solubilization. HPMC at 3.7 mg/ml was below the level that raises crystal solubility, so its 20 μg/ml excess counts as supersaturation. HPMCAS at 4.7 mg/ml was "sufficiently high to solubilize resveratrol", and after correction "it is evident that supersaturation exists even at the 24 h time point". The correction itself does not reconcile. The printed excess (155 μg/ml) equals 458 − 302 within rounding, so the 302 entry behaves as the crystal solubility in the polymer medium, not as an increase over 42 μg/ml; read as an increase, the excess would be 114 μg/ml (X3). Figure 2 gives only about 141 μg/ml at 5 mg/ml HPMCAS, and the same 4.7 mg/ml is later assigned to the 85% HPMCAS dispersion with a crystal solubility of 83 μg/ml (X2). Some excess remains in every version of the arithmetic, but its size is unresolved. Two further points: the Raman series for PVP and E100 end before 24 h, so their 24 h status rests on Table III (X14), and the E100 result is explained by complexation in the abstract but by the polymer not dissolving at pH 6.8 in the discussion ("can be rationalized"), with no complexation data for E100 (X8).

Main result 4: more cellulosic polymer, less crystallization and a larger gain.

Table D. Polymer content series in PB 6.8 at 37 °C. Raman ratios from Figure 6 (approximate; last points at about 1440 min for HPMCAS and HPMC and about 2700 for CMCAB, whose time axis has no printed unit and is read as minutes); concentrations and crystal solubilities from the bars of Figure 7 (approximate). Figure 7b labels the HPMC bars 85, 75 and 65%; they are listed as the 80, 70 and 60% dispersions of Figure 6, whose Raman values they match.

The highest level of each polymer stayed flat and every lower level crystallized. Table IV lists the minimum polymer needed to inhibit crystallization as 95% CMCAB, 85% HPMCAS and 80% HPMC, and 65% for both E100 and PVP. For the cellulosics these are the lowest tested levels without crystallization; E100 and PVP were tested only at 65%, so their entries are upper bounds (X7). The concentrations follow crystallinity: "unsurprisingly the larger the extent of crystallization, the less enhanced the solution concentration relative to the crystal solubility". Without crystallization, the gain over the polymer-corrected crystal solubility was about 1166 μg/ml for 85% HPMCAS against 50 μg/ml for HPMC. Part of this is the polymer released: the non-crystallizing HPMC dispersion released 2.6 mg/ml, which leaves crystal solubility unchanged, while the HPMCAS dispersion released 4.7 mg/ml, which raises it from 42 to 83 μg/ml. The authors conclude that "it appears that at least some part of the concentration enhancement can be attributed to the formation of a soluble resveratrol-polymer complex". Reporting conflicts sit inside this result: the HPMC dispersion with the 50 μg/ml increase is called "90% polymer" on one page and 80% HPMC on another, the Figure 7b bar labelled 65% HPMC (about 56 μg/ml) differs from Table III (62 μg/ml), and the 2.6 mg/ml HPMC value equals the assay's LOQ (X1, X11).

The proposed mechanisms are hedged and were not measured in the hydrated state. Crystallization from the cellulosics "is thought to be due to the insufficiently strong intermolecular interactions" with resveratrol. Its dependence on polymer amount "most probably reflects the number of functional groups available on the polymer for interaction", since the cellulosics carry a limited number of hydroxyl and other groups that can hydrogen-bond with resveratrol. At low drug loading resveratrol is also more finely dispersed, which "will also hinder crystallization".

Main result 5: crystals form in the matrix, not from the solution. The 25 wt% resveratrol HPMCAS dispersion crystallized to a similar extent at both pH values (Figure 8): the Raman ratio fell from about 1.05 to about 0.96 at pH 6.8 and from about 1.05–1.08 to about 0.93 at pH 1.2 over about 1440 min, with error bars of about ±0.02–0.04. Under polarized light after 24 h (Figure 9), particles at pH 6.8 had dispersed and dissolved and left small crystalline particles, while at pH 1.2 the particles, about 200–300 μm across, stayed intact with birefringent regions inside the matrix. The authors state "It is clear that there was no significant difference" in rate and extent, without a statistical test, and infer that the pH 6.8 crystals "most likely resulted from crystallization from the matrix, rather than first dissolving and then crystallizing from a supersaturated solution". Note by the AI writer: the 25 wt% dispersion is the 75% HPMCAS dispersion, whose pH 6.8 curve in Figure 6a also ends at about 0.96, so the two experiments agree.

Main result 6: more HPMCAS dispersion raised the concentration but lowered the flux. Tables E and F give the 15:85 loading series. The percentage columns and the last two columns of Table F are the AI writer's arithmetic from the card values, not numbers in the paper.

Table E. 15:85 resveratrol:HPMCAS dispersion stirred 24 h in 32 ml PB 6.8 at 37 °C (read from Figure 10a, approximate). Percentages compare each value with complete dissolution (15% or 85% of the loading). HPMCAS values above 10 mg/ml lie outside the HPLC calibration range.

Table F. Membrane flux of the same donor solutions (read from Figure 10b, approximate; crystal value 2.0 μg/min from the text). The last two columns assume, as E3 does, that flux is proportional to free resveratrol with the same constant for every donor.

Up to about 18.8 mg/ml loading, total resveratrol rose from about 530 to about 1610 μg/ml while the flux fell from about 4.4 to about 2.6 μg/min. Beyond that point the concentration fell steeply and a gelatinous precipitate formed, while the flux stayed at about 2.8–3.1 μg/min. On the rising branch the falling flux is "indicating that the amount of free resveratrol was reduced"; beyond the maximum, free resveratrol "appears to become relatively constant". Every dispersion solution gave a higher flux than the crystal suspension, "indicating that some extent of supersaturation was generated", but the degree of supersaturation did not rise with the amount released and showed "the opposite trend". The discussion turns this into the paper's main conclusion: "These results indicate that the resveratrol concentration enhancement is mainly due to the formation of a soluble resveratrol-HPMCAS complex", with free resveratrol at most about twice a saturated solution. The stated inverse correlation of flux with concentration holds only on the rising branch, up to about 18.8 mg/ml (X9).

Note by the AI writer (not discussed in the paper): if flux scales with free drug as E3 assumes, free resveratrol stayed between about 55 and 92 μg/ml across the series while the total ranged from about 180 to about 1610 μg/ml; at the concentration maximum only about 3–4% of the dissolved resveratrol would be free. Table E also shows that on the rising branch about half of the added resveratrol was in solution, as the authors state, while a larger share of the HPMCAS dissolved (about 92% at the lowest loading). The solution was therefore poorer in resveratrol than the 15:85 dispersion, and the undissolved residue was not characterized. These figures rest on plot readings and on the ±6 μg/ml uncertainty of the crystal solubility, so they indicate size only.

Comparisons and controls. The five polymers fall into three groups. PVP and E100 kept resveratrol amorphous but released no extra drug, for different stated reasons: an insoluble complex for PVP and an undissolved polymer for E100. CMCAB neither prevented crystallization nor dissolved, and gave 39 μg/ml. HPMC and HPMCAS crystallized at 65% polymer but still raised the concentration, by 20 μg/ml and by several hundred μg/ml; the difference lies mostly in how strongly the dissolved polymer holds resveratrol (Table A). The saturated crystal suspension rules out the possibility that the dispersion solutions hold less free drug than a saturated solution: the lowest dispersion flux still exceeded 2.0 μg/min. The pH 1.2 arm and the CMCAB dispersions show that resveratrol crystallizes quickly even when it cannot be released. Several controls are missing: no flux for the PVP, HPMC or 65% HPMCAS slurries; no measurement of dissolved PVP, for which the Table III footnote cites polymer insolubility although PVP is called extremely soluble (X10); no complexation test for E100; and solution concentrations only at the end of each slurry.

Discussion

How the authors connect the findings. The discussion calls the behaviour of an ASD in water "extremely complex": the matrix hydrates and may change phase, drug and polymer dissolve and interact, and the drug may crystallize. Because resveratrol has a very high melting point and is a rapid crystallizer, effective polymeric inhibitors are "essential". The chain below follows their argument and the scenarios of Figure 11.

For PVP, "It is proposed that the low resveratrol solution concentrations observed from this system result from the formation of an insoluble resveratrol-PVP complex". The support is that the dispersion "did not readily dissolve" and that crystalline resveratrol in PVP solution gave an amorphous precipitate containing both components, so "it appears that the insolubility of the dispersion arises as a result of resveratrol-polymer complexation". The same complex explains the absence of crystals: an insoluble amorphous complex "would be expected to hinder crystallization". The general lesson the authors draw is that strong drug–polymer interactions "may be favorable" for preventing crystallization in storage "but may be detrimental in terms of compound release upon dissolution".

Where the approach applies. The method is general: measure crystal solubility with the dissolved polymer present, measure how much polymer actually dissolves from the dispersion, and measure flux against a crystal reference. Any formulation where the polymer may bind the drug can be tested this way. The findings are narrower: one drug, these five polymer grades, mostly pH 6.8 and 37 °C, solution data after about 24 h, and flux only for the 15:85 HPMCAS dispersion. Within those limits the paper shows that a polymer ranking based on storage stability does not predict the concentration delivered in buffer, and that a high total concentration from a solubilizing polymer can carry little extra free drug.

What remains unresolved. Neither complex was detected directly: the HPMCAS complex is inferred from flux falling while concentration rose, the PVP complex from the precipitate and an IR spectrum that is not shown. Free resveratrol and stoichiometry in PVP solutions were not measured, nor was the undissolved residue of the PVP dispersion. The gelatinous precipitate above the maximum of the HPMCAS loading series resembles the PVP case but was not analysed. The matrix route at pH 6.8 is inferred from similar kinetics, not observed. The authors raise one open question themselves: soluble complex formation "might explain" the very high concentrations reported for ellagic acid, quercetin, curcumin, naringenin and resveratrol in cellulose derivative dispersions (refs 32–36).

Note by the AI writer (not discussed in the paper): the complex formed when the dispersion dissolves appears to hold far more resveratrol than the crystal-solubility experiment predicts. At 10 mg/ml HPMCAS, 414 μg/ml dissolves in equilibrium with the crystal, so about 372 μg/ml is bound while free drug is at saturation (42 μg/ml). At the peak of the loading series, about 11.2 mg/ml HPMCAS holds about 1610 μg/ml while the flux implies only about 55 μg/ml free: nearly four times as much bound resveratrol per milligram of polymer, with free drug only about 1.3 times higher. Either binding rises very steeply with free drug, or the dispersion produces drug–polymer associations that the crystal does not form at equilibrium, or part of the "dissolved" resveratrol is in a form, such as polymer-associated aggregates, that survives centrifugation at 40,000 g. The paper does not test these possibilities. The same gap explains why the polymer-corrected excess of Figure 7 (up to about 1166 μg/ml) is so much larger than the free-drug excess implied by flux.

What it suggests for formulation and further work. The authors recommend that "it is important to first determine if the formulation is stable to crystallization when exposed to aqueous media", and then whether any concentration gain is supersaturation or solubilization; strong drug–polymer interactions "may in certain cases lead to polymer:drug complex formation, which must be assessed". The AI writer's reading, not stated in the paper: a storage-stability screen alone can pick the wrong polymer, since PVP kept resveratrol amorphous and still gave no gain. Where a polymer solubilizes the drug, concentration-based dissolution tests overstate the benefit, and flux or another free-drug measure is needed to rank formulations. For phenolic drugs with PVP, a precipitate that appears as the drug-to-polymer ratio rises warns of insoluble complexation. The polymer levels that stopped crystallization in water (80% HPMC, 85% HPMCAS, 95% CMCAB) allow only 5–20% drug, which limits the dose per unit. The authors cite NMR studies that distinguished supersaturated from solubilized solutions (Ueda 2012, 2013; refs 29, 30); a direct measurement of that kind on these solutions would test the proposed complexes, and absorption data would show whether the small free-drug gain matters in vivo.

Limitations

Limits the authors state. The paper has no limitations section, but the text carries several caveats. The Raman ratio is used "qualitatively" because "no pure amorphous reference was available", and in the cellulosic dispersions "complete crystallization was never reached". The IR evidence for the PVP precipitate is "data not shown". In the loading series, concentrations "were only approximately half" of complete dissolution, and the polymer data after the maximum "were more scattered". The inverse flux–concentration correlation is stated "for this particular system". The mechanistic claims carry the authors' own hedges: the insoluble PVP complex "is proposed", the HPMCAS complex is what "These results indicate", the matrix route "appeared", the E100 result "can be rationalized", and the abstract's complexation explanation is "most likely".

The AI writer's own judgement (1): the complexes are inferred, not shown. The card rates the evidence as moderate overall. The crystallization ranking of the polymers is solid: triplicate Raman runs agree with XRPD and polarized light microscopy, and the flux series has a crystal control. The mechanistic claims are weaker. The HPMCAS complex rests on flux falling while concentration rose, and free drug in the 65% HPMCAS slurry, the one that gave 458 μg/ml, was not measured. The insoluble PVP complex rests on a titration with crystalline drug and an unseen IR spectrum, while the residue and dissolved PVP of the actual dispersion slurry are unknown (X10). E100 complexation is stated in the abstract without data (X8). The flux reading also assumes that dissolved HPMCAS does not change the membrane or the diffusion of free resveratrol, which the paper does not test.

The AI writer's own judgement (2): numbers that do not reconcile, and scope. The HPMC polymer levels conflict between figures, table and text (X1). The 4.7 mg/ml HPMCAS level is given for both the 65% and 85% dispersions with crystal solubilities of 302 and 83 μg/ml (X2, X3), so the corrected supersaturation values of 155 and about 1166 μg/ml should not be used quantitatively. The 20 mg/ml HPMCAS solubility point exceeds the drug added (X5), the slurry solids loading is unresolved (X6), and UV values in PVP solutions carry at least about 8% error (X12). On scope: the E100 and PVP minimum polymer levels are upper bounds (X7), "no significant difference" between pH values has no statistical test, the Raman ratio cannot give percent crystallinity (X13), and the study covers one drug, mostly one pH, end-point solution data only and no absorption data. None of these reverses a qualitative conclusion, but the ranking and the direction of the effects are better supported than any single number. Whether the complexation lesson holds for non-phenolic drugs, this paper cannot say.

Conclusion

The paper tested whether resveratrol dispersions that resisted crystallization in humid storage would also resist it in water and so give higher concentrations. The crystallization part held: PVP and E100 kept resveratrol amorphous in pH 6.8 buffer, while the cellulosic dispersions partly crystallized within an hour unless they held at least 80% HPMC, 85% HPMCAS or 95% CMCAB. The concentration part did not. The amorphous PVP and E100 dispersions gave 42 and 46 μg/ml, no more than the crystal, while partly crystallized HPMCAS dispersions gave 458 μg/ml at 35% drug and up to about 1600 μg/ml in the 15:85 loading series. In the authors' account, PVP binds resveratrol into an insoluble complex that blocks release, E100 does not dissolve, and HPMCAS holds resveratrol in a soluble complex.

The flux data carry the main message. The 15:85 HPMCAS solutions reached at most about twice the crystal's flux, and less as their total concentration rose, so the authors conclude that the gain is "mainly due to" a soluble complex and that resveratrol supersaturation was "very limited". The abstract generalizes: "the extent of supersaturation was limited not only by crystallization, but also by soluble and insoluble complex formation". For formulation work the lesson is to test an ASD in water and not only in storage, and to measure free drug, by flux or otherwise, whenever the polymer may bind the drug.