Choosing a crystallization inhibitor for polyphenol dispersions with mid-infrared spectroscopy

Background and motivation

Naringenin and quercetin (Figure 1) are flavonoids, polyphenolic plant compounds used in foods, beverages and dietary supplements and studied for antioxidant, antitumour and antibacterial activity. Both are sparingly soluble in water: the paper cites 45 µg/ml for naringenin (37 °C, pH 3) and 0.4 µg/ml for quercetin (room temperature, pH 3), and links this poor solubility to their limited absorption after oral administration. Particle size reduction, solid dispersions and cyclodextrin inclusion complexes had all been tried. The authors single out the amorphous solid dispersion (ASD) as especially attractive here. Both compounds melt high (250 °C and 316 °C), which reflects strong crystal lattice energies, so disrupting the crystal "can be expected" to raise solution concentration and biological exposure substantially. The weakness of the approach is that the polyphenol may crystallize over pharmaceutically relevant times and lose the solubility advantage.

In an ASD the polymer is there to inhibit crystallization. The tendency of a drug to crystallize from a dispersion depends both on the drug itself and on how well the polymer inhibits it (Van Eerdenbrugh and Taylor 2010). The difficulty the authors point to is practical. The usual tools, powder X-ray diffraction (PXRD), differential scanning calorimetry (DSC), Raman spectroscopy and polarized light microscopy, only confirm that a sample has crystallized; they do not predict whether it will. Stability studies across several polymers are therefore costly, and a formulation that crystallizes late means reformulation. The rule commonly used in practice, choosing a polymer that raises the glass transition temperature ($T_g$) of the dispersion above that of the amorphous drug, is imperfect: adequate crystallization inhibition has been achieved even when the blend $T_g$ was unchanged or lower (Khougaz and Clas 2000; Matsumoto and Zografi 1999).

The Taylor group had been building an alternative based on specific drug–polymer interactions. Crystal growth studies with felodipine and with two other compounds (Kestur and Taylor 2010; Kestur et al. 2011) and an ab initio polymer selection scheme based on hydrogen-bond rules (Van Eerdenbrugh and Taylor 2011) suggested that polymers forming stronger intermolecular interactions with the drug inhibit crystallization better. For two other polyphenols, resveratrol and curcumin (Wegiel et al. 2013, the second then submitted), infrared (IR) spectroscopy had proved useful for monitoring these interactions and anticipating ASD stability. This paper extends the IR approach to naringenin and quercetin. They are a deliberate test case: unlike resveratrol, both have intramolecular as well as intermolecular hydrogen bonds in the crystal, which "may affect" how well they bond to a polymer, and quercetin normally crystallizes as a dihydrate, which makes humidity a relevant storage variable. Six polymers with different chemistry were compared (Figure 2): poly(vinylpyrrolidone) (PVP), Eudragit E100 (E100), carboxymethylcellulose acetate butyrate (CMCAB), hydroxypropylmethylcellulose (HPMC), HPMC acetate succinate (HPMCAS) and poly(acrylic acid) (PAA).

Background concepts

Reading hydrogen bonds from an IR spectrum. An O–H group absorbs infrared light in the OH stretching region, roughly 3600–3100 cm⁻¹. When the OH group donates a hydrogen bond, the O–H bond weakens and its band moves to lower wavenumber and broadens; the stronger or more extensive the hydrogen bonding, the larger the shift (Taylor and Zografi 1997). The partner side can be read the same way. A carbonyl group (C=O) that accepts a hydrogen bond absorbs at a lower wavenumber than a free carbonyl, so a polymer carbonyl band that moves down, or grows a shoulder at lower wavenumber, when the drug is added indicates drug-to-polymer hydrogen bonds. The paper uses both readings: the position of the polyphenol OH band, and the shift of the polymer carbonyl band, both at the same drug content. One complication is that each method works only where the other component does not absorb strongly: PAA has its own large OH band, and HPMC has no carbonyl.

Acid–base (ionic) interactions. Naringenin and quercetin are weak acids because of their phenolic OH groups. E100 is a methacrylate copolymer carrying a basic dimethylaminoethyl group, the only basic group among the six polymers. A phenol and a tertiary amine can react: the phenol gives up its proton to the amine, and the two ions attract. The authors note that such ionic interactions are "typically stronger" than hydrogen bonds. In IR, the non-ionized dimethylamino group of E100 has two characteristic bands near 2820 and 2770 cm⁻¹ (Moustafine et al. 2006); their loss marks amine groups that have been protonated.

Hydrogen-bond donors, acceptors and the pKBHX scale. Whether two molecules hydrogen-bond depends on the strength of the best donor on one and the best acceptor on the other. Etter's empirical rules for organic crystals (Etter 1990, 1991) state that all good donors and acceptors are used, that six-membered intramolecular hydrogen-bond rings form more readily than intermolecular bonds, and that the best donors and acceptors left after intramolecular bonding pair with each other. To grade acceptors, the paper uses the pKBHX scale, a database of hydrogen-bond basicity values designed for medicinal chemists (Laurence et al. 2009). Following Van Eerdenbrugh and Taylor (2011), acceptors are classed as weak (below 0.75), medium (0.75–1.5), strong (1.5–2.25) or very strong (above 2.25). For example, 1-methyl-2-pyrrolidone, a model of the PVP repeat unit, has a pKBHX of 2.38 (very strong), triethylamine, a model of the E100 amine, 1.98 (strong), and ethyl acetate, diethyl ether and ethanol 1.01–1.07 (medium).

Intramolecular hydrogen bonds in the two flavonoids. Figure 1 draws the intramolecular hydrogen bonds as dashed lines. In naringenin, the 5-OH group bonds to the neighbouring ketone oxygen; in quercetin both the 5-OH and the 3-OH bond to the ketone. Naringenin has three phenolic OH groups, of which two are free to form intermolecular bonds; quercetin has five, of which three are free. The ketone, the strongest acceptor on either molecule, is already used intramolecularly. In the crystal, the OH groups build layers of hydrogen-bonded sheets that stack (Cambridge Structural Database codes DOLRIF for naringenin, AKIJEK and FEFBEX for quercetin monohydrate and dihydrate). The authors note that an anhydrous form of quercetin also exists but was never seen in this study.

Crystallization tendency and what "X-ray amorphous" means. Van Eerdenbrugh et al. (2010) classified compounds by how easily they crystallize when a solvent is removed quickly. As the paper applies the scheme, a compound that cannot be obtained amorphous even as a fast spin-coated film is a rapid crystallizer, and one that gives an amorphous spin-coated film but not an amorphous product by slower rotary evaporation is an intermediate crystallizer. A sample is called X-ray amorphous when its PXRD pattern shows no sharp peaks; a small crystalline fraction below the detection limit may still be present. In the stability tables, "partial crystallization" means that crystalline peaks appeared but did not grow further during the study, leaving an amorphous fraction.

Methods in detail

The study has three parts. First, IR spectra of polyphenol–polymer films rank the polymers by interaction strength, and a structural donor–acceptor analysis gives an independent expectation. Second, the highest polyphenol content that each polymer can hold in an X-ray amorphous dispersion is measured. Third, one dispersion per drug–polymer pair is stored at five temperature and humidity conditions and checked for crystals. The two crystallization tests are then compared with the IR-based prediction.

Contribution 1: ranking polymer interactions by IR

Problem. To rank polymers before any stability test, one needs a measure of how strongly each polymer binds the polyphenol in the amorphous state, measured the same way for every pair.

What they did. Single compounds and polyphenol–polymer blends were dissolved in a 1:1 (by weight) mixture of dichloromethane and ethanol, the same solvent used to make the dispersions, and each solution was checked by eye to be a single phase. Two or three drops were placed on a thallium bromoiodide (KRS-5) crystal and spun at 500 rpm for 18 s and then 2500 rpm for 30 s. The thin films were measured in absorbance on a Bio-Rad FTS 6000 spectrometer (globar source, KBr beamsplitter, DTGS detector) from 4000 to 500 cm⁻¹ at 4 cm⁻¹ resolution with 128 co-added scans, with the absorbance of the region of interest kept between 0.6 and 1.2 and both sample and spectrometer flushed with dry air. All polymers had first been dried over phosphorus pentoxide for at least a week. The comparison was made at 30% (w/w) polyphenol, the highest content that gave a thin amorphous film with every polymer. Concentration series are shown for three pairs: naringenin–HPMC in the OH region (Figure 3), naringenin–PVP in the carbonyl region (Figure 4) and quercetin–E100 in the amine region (Figure 6).

How the spectra were read. Two assumptions carry the OH-region comparison. The OH band of the amorphous polyphenol alone is assumed to lie at the same position for naringenin and quercetin, because their structures are similar and amorphous quercetin could not be made as a reference. Then, at constant polyphenol content, a lower OH band position is read as stronger hydrogen bonding with that polymer. In the carbonyl region, a larger downward shift of the polymer carbonyl band relative to the pure polymer is "assumed to correspond to stronger hydrogen bonding". The two regions are plotted against each other in Figure 5 (OH band position against carbonyl shift, with a straight trend line). HPMC (OH value only) and PAA (carbonyl value only) are placed on that line from their single value, so that all six polymers can be ordered on one scale.

Why it works, and how it differs. The approach measures the interaction directly in the amorphous blend rather than inferring it from $T_g$, and both partners of the bond are observed: the donor (polyphenol OH) and the acceptor (polymer C=O). It repeats the logic used earlier for resveratrol and curcumin by the same group. For E100, whose basic amine can react with the phenol without any carbonyl hydrogen bond, the authors add a separate marker: the loss of the two non-ionized amine bands at 2820 and 2770 cm⁻¹.

Contribution 2: a structural expectation from donor and acceptor strengths

What they did. In the Discussion, the authors list every hydrogen-bonding group of each polymer with its content in mmol/g, whether it can donate or accept, its donor strength and the pKBHX acceptor strength of a model compound (Table 4), following the scheme of Van Eerdenbrugh and Taylor (2011). The polyphenols' only donors are phenolic OH groups, two free ones in naringenin and three in quercetin, and their strongest acceptor, the ketone, is tied up in an intramolecular bond. Under Etter's rules, intermolecular bonds should therefore form "most likely" between the free phenolic OH donors and acceptor groups on the polymer. The polymer with the strongest acceptor should form the best hydrogen bonds: PVP (amide, very strong, 2.38), then E100 (tertiary amine, strong, 1.98); all other acceptors are medium, with pKBHX near 1. Polymers that are mainly strong donors, in particular PAA and CMCAB with their carboxylic acids, "are not expected to be effective" inhibitors, because the polyphenols offer no free strong acceptor for them. No pKBHX value exists for a carboxylic-acid acceptor, so it was taken as comparable with an ester.

Note by the AI writer (not discussed in the paper). Table 4 marks the HPMCAS carboxylic acid as donor only, whereas the same group in PAA and CMCAB is marked donor and acceptor; an acceptor strength is printed for all three, so the missing mark is probably an omission (card note X13). How the mmol/g values were calculated is not stated.

Contribution 3: two crystallization tests

Test 1, the highest loading that stays amorphous. Neither polyphenol could be made amorphous on its own by rotary evaporation. Each polyphenol was therefore co-dissolved with each polymer at several dry weight ratios (the ratios are not listed) in 1:1 dichloromethane:ethanol, the solvent was removed on a rotary evaporator with a 60 °C water bath, the residue was dried under vacuum for 24 h and cryo-milled in liquid nitrogen for 4 min in total, and PXRD was run immediately (Shimadzu XRD-6000, Cu Kα, 5–35° 2θ at 8°/min, 0.04° steps, silicon standard checked daily). Samples with sharp peaks were called crystalline; those without, X-ray amorphous. The highest polyphenol content that came out X-ray amorphous with each polymer (Figure 7) was taken as "a strong indicator" of that polymer's crystallization-inhibiting performance: a polymer that needs less of itself to keep the drug amorphous is "likely a better crystallization inhibitor".

Test 2, storage. X-ray amorphous dispersions were stored for long-term stability. The methods state that dispersions with 40% polymer were selected and that the polymer level was raised for systems that could not be made at 40%; Table 3 lists the polymer content actually used for each pair, from 30% (quercetin–CMCAB) to 80% (quercetin–PAA). Five conditions were used: room temperature (22–25 °C) over phosphorus pentoxide (0% RH); room temperature over potassium chloride (84% RH); 40 °C over sodium chloride (75% RH); and 50 °C and 90 °C over anhydrous calcium sulfate (Drierite, 0% RH). Samples were checked periodically by PXRD, and the result is the number of days before crystalline material was first seen. The methods give storage "for up to 350 days", while Table 3 reports monitoring stopped at >200, >300 or >450 days, so the end of storage is stated two ways (card note X1). Onset $T_g$ of the dry polymers and dispersions was measured in triplicate by DSC (TA Q2000; second heating at 10 °C/min; PVP dispersions by modulated DSC at 2 °C/min with ±1 °C every 60 s).

Experimental analysis

Main result 1: PVP stands apart in hydrogen bonding. The concentration series in Figure 3 shows what the OH-region readout looks like. As naringenin in HPMC rises from 10% to 50%, the OH stretching band broadens and its maximum moves from about 3470 cm⁻¹ (the HPMC band, almost unchanged at 10% naringenin) to about 3335 cm⁻¹ at 50% (read from Figure 3, approximate). The ranking itself uses the 30% films (Table A). With PVP the polyphenol OH band sits at 3138 cm⁻¹ (naringenin) and 3135 cm⁻¹ (quercetin), roughly 190–260 cm⁻¹ below its position with any other polymer. HPMCAS follows (3329 and 3361 cm⁻¹), and E100, HPMC and CMCAB lie close together at 3377–3399 cm⁻¹. From these values the authors rank the OH-region hydrogen bonding PVP > HPMCAS > E100 ≥ HPMC ≥ CMCAB. PAA is left out because its own large OH band swamps the polyphenol band.

Table A. Mid-IR band positions in spin-coated films at 30% (w/w) polyphenol (from Tables 1 and 2; card R2, R4). Shift = pure polymer peak minus dispersion peak, in cm⁻¹.

The carbonyl region gives the same picture from the acceptor side. In naringenin–PVP films (Figure 4) a shoulder grows at 1661 cm⁻¹ as naringenin increases, 21 cm⁻¹ below the free PVP carbonyl at 1682 cm⁻¹; with quercetin the shifted band lies at 1658 cm⁻¹, a shift of 24 cm⁻¹. The authors take this "clear shift" as evidence that the drug hydrogen-bonds to the polymer. The other carbonyls barely move: 2–3 cm⁻¹ for HPMCAS, 1–2 cm⁻¹ for CMCAB, none for E100, and for PAA the band moves up by 2 cm⁻¹, which the authors read as "weak to no" hydrogen bonding, or bonding no different from that within PAA itself. The carbonyl ranking is therefore PVP > HPMCAS > CMCAB > E100 > PAA, in "reasonably good agreement" with the OH ranking. Plotting one region against the other (Figure 5) gives a straight trend, and with HPMC and PAA placed on the line from their single values the authors propose the combined hydrogen-bond order PVP > HPMCAS > HPMC ≥ CMCAB ≥ E100 > PAA.

Note by the AI writer (not discussed in the paper). Two details limit how finely this ranking can be read. First, E100, HPMC and CMCAB differ by only 3–15 cm⁻¹ in the OH region at a stated resolution of 4 cm⁻¹, no replicate spectra are reported, and the order between them differs between the drugs: for quercetin E100 (3377) < HPMC (3384) < CMCAB (3392), for naringenin E100 (3394) < CMCAB (3396) < HPMC (3399) (card note X8). Second, Figure 4 shows that amorphous naringenin itself absorbs strongly between about 1665 and 1580 cm⁻¹, with its main band near 1641 cm⁻¹ (read from Figure 4, approximate). The 1661 cm⁻¹ shoulder assigned to hydrogen-bonded PVP carbonyl therefore sits on the edge of a drug band; the authors mention band overlap only as a general limitation of the method (card note X9). Neither point affects the conclusion that PVP stands far apart from the rest.

Main result 2: E100 reacts with the polyphenols instead of hydrogen-bonding. E100 shows no carbonyl shift and only a modest OH shift, so by hydrogen-bond criteria alone it would rank near the bottom. Figure 6 shows the quercetin–E100 series between 3600 and 2400 cm⁻¹. The two sharp bands of the non-ionized dimethylamino group near 2820 and 2770 cm⁻¹ are clear in pure E100 and at 10% quercetin, smaller at 30%, and gone at 50% (w/w) quercetin, while a broad OH band near 3400 cm⁻¹ grows. The authors read the loss of these bands as the loss of un-ionized amine groups, which "provides evidence" of an acid–base reaction between the flavonoid and the polymer; for naringenin–E100, "similar spectral observation indicated" the same, but those spectra are not shown. Because ionic interactions are "typically stronger" than hydrogen bonds, the authors combine both readings into a predicted order of crystallization-inhibiting performance: E100 (ionic interactions) > PVP > HPMCAS > HPMC ≥ CMCAB > PAA. The structural analysis of Table 4 agrees with the IR in placing PVP first among hydrogen-bond acceptors, but expected E100 to be second through hydrogen bonds to its amine; the IR shows that this amine is consumed by the acid–base reaction instead, which the authors call the "main outlier".

Main result 3: the highest loading that stays amorphous. Figure 7 gives the first test of the ranking. E100 kept the most polyphenol amorphous and PAA the least (Table B): only 15% E100 was needed for naringenin and 30% for quercetin, whereas PAA had to make up 70% and 80% of the dispersion. Between these ends, the cellulosic polymers and PVP "performed similarly", allowing roughly 65–75% naringenin and 30–55% quercetin. The authors conclude that the result is "in line with" the IR ranking: E100 stood out as the best inhibitor, and PAA needed so much polymer that very little polyphenol remained.

Table B. Highest polyphenol content giving an X-ray amorphous dispersion by rotary evaporation from 1:1 dichloromethane:ethanol (read from Figure 7, approximate; end values confirmed by the text; card R8).

Main result 4: storage stability. Table C reproduces the storage results (Table 3). The E100 and PVP dispersions of both polyphenols showed no crystals at any condition until monitoring stopped: more than 450 days (naringenin–E100) or more than 300 days at the three dry conditions and at 84% RH, and more than 200 days at 40 °C/75% RH. This includes 90 °C and the two humid conditions, the most demanding ones in the study. For the other four polymers, humidity was decisive. At 25 °C/84% RH and at 40 °C/75% RH, every naringenin dispersion with HPMCAS, HPMC, CMCAB or PAA crystallized within 1 day, and every quercetin dispersion within 3 days (1 day with PAA). The authors summarize: "It is clear" that the polymers have very different inhibitory abilities, with E100 and PVP "the most resistant to crystallization".

Table C. Days before crystalline material was observed by PXRD (from Table 3; card R9). ">X" means monitoring stopped at X days with no crystals; "partial" means the system only partly crystallized. T_g is the onset value of the dry dispersion; dry polymer T_g: E100 41 ± 1, PVP 159 ± 2, HPMCAS 111 ± 1, HPMC 145 ± 3, CMCAB 137 ± 3, PAA 128 ± 0.5 °C.

Dry storage and the two drugs. Without moisture the picture is more mixed. At 25 °C/0% RH most dispersions held for the whole study; the exceptions were naringenin–HPMCAS (56 days, partial) and quercetin–PAA (14 days, partial). At 50 °C/0% RH naringenin–HPMCAS (3 days, partial), naringenin–PAA (7 days) and naringenin–CMCAB (75 days, partial) crystallized. At 90 °C/0% RH all four naringenin dispersions with cellulosic polymers or PAA crystallized within 1 day, whereas quercetin–HPMCAS and quercetin–HPMC stayed amorphous for more than 300 and 200 days. The authors draw the general point that naringenin "had a higher tendency to crystallize from the dispersions": it crystallized earlier from most systems, and at conditions under which the quercetin systems were stable. This is notable because quercetin was the faster crystallizer on its own (only naringenin could be spin-coated amorphous).

Partial crystallization. Seven entries in Table 3 are marked as only partly crystallized, five with cellulosic polymers and two with quercetin–PAA. The authors describe these systems as crystallizing partly and then showing no further crystal growth over the study, and point to Figure 8, PXRD patterns of a naringenin–HPMC dispersion at 60% (w/w) naringenin. At 50 °C/0% RH (Figure 8a) the pattern stays a broad halo through 250 days, with a weak sharp feature near 22° 2θ from 75 days onward; at 25 °C/84% RH (Figure 8b) sharp crystal peaks are present from day 1 on top of a remaining halo and change little up to 200 days (read from Figure 8, approximate). Note by the AI writer (not discussed in the paper): Table 3 lists naringenin–HPMC (40% polymer, the same 60% naringenin composition) as showing no crystals for more than 450 days at 50 °C/0% RH and as crystallizing in 1 day, without the partial mark, at 25 °C/84% RH, so the figure and the table do not agree on these two conditions (card note X7).

The glass transition temperatures. Table 3 also reports the onset $T_g$ of each dry dispersion, from 69 °C (naringenin–CMCAB) to 128 °C (naringenin–PVP). The E100 dispersions (85 and 91 °C) lie 44–50 °C above E100 itself (41 °C). The paper does not discuss these values. Note by the AI writer (not discussed in the paper): they do not follow the stability order. For example, naringenin–PAA (87 °C) and naringenin–E100 (85 °C) have almost the same $T_g$, yet at 50 °C/0% RH the first crystallized in 7 days and the second not in 450 days; naringenin–HPMC (77 °C) also held for more than 450 days there (card note X14). This fits the introduction's argument that $T_g$ alone is a poor guide to crystallization inhibition.

Discussion

The authors' causal chain. The Discussion links the results through the structure of the flavonoids. Both compounds have a high inherent crystallization tendency, seen in the failure to make neat amorphous samples, which the authors say "can be attributed" to their hydroxyl groups building an extensive hydrogen-bonded network in the crystal. A polymer inhibits crystallization if it offers interactions more favourable than the flavonoid–flavonoid ones ("it is reasonable to expect"). Resveratrol, which has no intramolecular hydrogen bonds, could be stabilized only by strong drug–polymer hydrogen bonds; curcumin, whose donors are all held intramolecularly, was hard to stabilize by hydrogen bonding at all. Naringenin and quercetin sit in between, with some donors free, so "some stabilization" through polyphenol–polymer hydrogen bonds is expected.

Why E100 works. E100 is the case where the IR and the structural expectation part ways. The pKBHX scale makes its amine the second-best acceptor, but the IR shows little hydrogen bonding. The authors explain this by the ionization of the dimethylamino group by the polyphenol: once protonated, the amine is "no longer available" for hydrogen bonding, and the resulting ionic interaction is "much stronger" than a hydrogen bond. The same interaction had raised the stability of resveratrol and curcumin dispersions with E100, and ionic drug–polymer interactions have been linked to stability for dexamethasone phosphate and indometacin; Yoo et al. reported ionic systems stable for up to 50 days at 60% RH and 25 °C while others recrystallized. The authors therefore write that ionic interactions "appear to be the reason" for the extremely good stability of the flavonoid–E100 dispersions, and in the Conclusions that this "is suggested" to be due to the "proposed" ionic interaction.

Why PVP survives humidity. PVP is very hygroscopic, and at high humidity crystallization should speed up because water raises molecular mobility; felodipine–PVP dispersions, for example, crystallize much faster in the presence of moisture (Rumondor et al. 2009). Naringenin (log P 2.60) and quercetin (log P 1.82) are less hydrophobic than felodipine (log P 4.5), so their PVP dispersions "would be expected" to hold more water and crystallize more readily. They did not. The authors take this as further evidence of strong polyphenol–PVP hydrogen bonds and offer two "Possible explanations": that the hydrogen bonds suppress moisture sorption, as reported for resveratrol, or that the more hydrophilic flavonoids do not undergo amorphous–amorphous phase separation in the presence of water. Neither was measured here.

Partial crystallization. The authors state that they "do not fully understand" the partial crystallization. For quercetin–PAA at 0% RH, they suggest that quercetin, which crystallizes as a mono- or dihydrate, needs water to crystallize; PAA, being extremely hygroscopic, "likely" kept a limited amount of water from production and drying, enough for some quercetin to crystallize but not all. This cannot explain the partial crystallization with the cellulosic polymers, which occurred with the anhydrous-crystallizing naringenin and at humid conditions.

Acidic polymers. PAA was "by far" the least effective polymer during solvent evaporation. The authors connect this to an earlier finding that acidic polymers are poor crystallization inhibitors for acidic compounds (Van Eerdenbrugh and Taylor 2011), which fits the weak-acid character of both flavonoids.

Scope. The comparison covers two flavonoids of similar structure, both weak acids with intramolecularly bonded OH groups; six polymers; dispersions made by solvent evaporation and cryo-milling; and storage of one composition per pair at five conditions. The IR measurements were made on spin-coated thin films at 30% polyphenol, the dispersions in the loading test and the stability study were rotary-evaporated powders, and the stored dispersions held 20–70% polyphenol. No dissolution, solubility or supersaturation data were collected.

What it suggests for formulation. The authors conclude that mid-IR is "a useful tool" to assess drug–polymer interactions that "could be used as a polymer selection tool" for ASDs with optimal physical stability, and in the Conclusions that "it can be concluded that this analytical approach is an important tool". Note by the AI writer (not discussed in the paper): in practice, the data support using IR to find the strongest candidates (here a very strong amide acceptor and a base that reacts with the phenol), and they show that an acid–base interaction can be invisible in the usual hydrogen-bond shifts, so a separate check for it is needed when the drug is acidic or basic.

Limitations

Limitations stated by the authors. The authors list three limits of IR for this purpose. The drug and polymer must have spectral features well resolved from each other; overlapping peaks need more advanced analysis, such as multivariate methods. Absorbed moisture must be removed, because water also hydrogen-bonds to both drug and polymer and makes drug–polymer bonds hard to define. For molecules with little propensity for hydrogen bonding, IR can hardly distinguish a molecularly mixed system from a phase-separated one. They also state that they "do not fully understand" the partial crystallization seen in several dispersions.

The AI writer's own judgement (1): the ranking holds at the ends, not in the middle. The authors describe "a good correlation" and a ranking that "correlates particularly well" with stability. The data support the two ends: E100 and PVP were stable everywhere, and PAA needed the most polymer. The middle does not follow. HPMCAS, ranked third by IR, gave the least stable naringenin dispersion under dry storage (56 days at 25 °C and 3 days at 50 °C, both partial), while HPMC, CMCAB and PAA, ranked lower, stayed amorphous for more than 450 days at 25 °C/0% RH. In the loading test, PVP, the strongest hydrogen bonder, allowed about 65% naringenin and 50% quercetin, the same as HPMCAS and less than HPMC (about 75% and 55%), so the Conclusions' statement that less polymer was needed for polymers that "developed stronger hydrogen-bonding interactions" is supported only by E100 versus PAA, and E100 is the polymer with the least hydrogen bonding. The authors' own wording, "In general", is the right strength (card notes X4, X6). The OH positions of E100, HPMC and CMCAB are within a few wavenumbers of each other, so the IR could not have resolved them anyway (X8).

The AI writer's own judgement (2): the design mixes several variables. The stored dispersions contained 30–80% polymer, so polymer identity and drug loading change together; the PAA dispersions in particular held only 20–30% polyphenol. The IR was measured on thin films at 30% polyphenol under dry air, not on the cryo-milled powders that were stored. Some compositions are also inconsistent: Table 3 lists quercetin–CMCAB at 30% polymer, which means 70% quercetin, whereas Figure 7 shows that only about 30% quercetin could be made X-ray amorphous with CMCAB; quercetin–HPMC at 40% polymer (60% quercetin) also exceeds the about 55% of Figure 7, and several pairs were stored at more polymer than the stated 40% although Figure 7 shows that 40% would have sufficed (card note X2). The storage end is given as "up to 350 days" in the methods and as ">450" days in Table 3 (X1), and Figure 8 does not match the Table 3 entries for the same naringenin–HPMC composition (X7). No replicates or statistics are reported for the IR positions or the days to crystallization.

The AI writer's own judgement (3): the mechanisms are proposed, not tested. The acid–base interaction of E100 is shown for quercetin only, through the loss of two amine bands; the naringenin spectra are described but not shown, no band of the ionized phenol is reported, and nothing in the design separates the ionic interaction from other properties of E100 (for example its polymer chemistry as a whole, tested here at a single composition per drug) as the cause of stability (X10). The explanations for PVP's humidity stability, suppressed water uptake or absence of phase separation, were not measured; no water content or sorption data are reported (X11). The 1661 cm⁻¹ carbonyl shoulder used for PVP overlaps naringenin's own absorption (X9).

Conclusion

The paper extends the group's IR approach to two flavonoids with intramolecular hydrogen bonds. Mid-IR band shifts in thin films separate PVP, which forms by far the strongest hydrogen bonds with naringenin and quercetin, from the other polymers, and a second IR marker, the loss of E100's non-ionized amine bands, identifies an acid–base interaction between the basic polymer and the phenolic drugs. The predicted order E100 > PVP > HPMCAS > HPMC ≥ CMCAB > PAA was then tested twice. E100 allowed the highest polyphenol content in an X-ray amorphous dispersion and PAA the lowest, and the E100 and PVP dispersions were the only ones that resisted crystallization at 84% RH, at 40 °C/75% RH and, for naringenin, at 90 °C. The authors conclude that mid-IR is an important tool for rational polymer selection.

Read with its data, the paper shows something narrower and still useful. IR identified the two polymers that work for these weak-acid polyphenols, one through a very strong amide acceptor and one through acid–base chemistry that the hydrogen-bond shifts alone would have missed, while the order of the cellulosic polymers was not reproduced by the stability data. Humidity, not dry heat, was the condition that separated good from poor polymers.