Which polymer slows crystal growth best depends on the drug: bifonazole and nimesulide

Background and motivation

Many new drug candidates dissolve poorly in water. Turning such a drug into its amorphous form raises its apparent solubility and dissolution rate, but the amorphous solid tends to crystallize, and crystallization removes the advantage and can make the product fail (Introduction, refs 1–3). The usual remedy is to mix the drug with a miscible polymer to make a kinetically stabilized drug–polymer dispersion, an amorphous solid dispersion (ASD) (refs 4–7). Why the polymer helps has been attributed to several factors: the polymer's glass transition temperature (Tg), the Tg of the dispersion, molecular mobility, and specific drug–polymer interactions such as hydrogen bonds (refs 8–12). The authors point out that, despite this list, there was no clearly defined rationale for choosing the polymer that gives the most stable dispersion, and that more basic studies of crystallization from amorphous systems were needed C1.

Two earlier studies from the same group had started to link polymer chemistry to crystallization inhibition. In a screen by Van Eerdenbrugh and Taylor (ref 13), 7 chemically diverse polymers were tested against 8 readily crystallizing compounds. Acidic polymers behaved most extremely: they prevented crystallization of drugs with basic and amide groups but performed poorly with drugs bearing acidic groups C2. In the second study (Kestur and Taylor, ref 14), the linear crystal growth rate of felodipine was measured with low levels of several polymers. Polymers that formed stronger or more extensive hydrogen bonds with felodipine, judged by infrared (IR) spectroscopy, "appear to be" better crystal growth inhibitors, and there was no correlation with the Tg of the polymer or of the dispersion, or with the polymer's molecular weight C3.

Both studies point to drug–polymer chemistry, but the felodipine work used one drug. The present paper asks whether the conclusion holds when the drug's chemistry changes. The hypothesis is stated plainly: drug–polymer interactions play a predominant role in how well a polymer inhibits crystallization of an amorphous drug, so the best polymer for one drug need not be the best for another C4. To test it, the authors chose two drugs with different hydrogen bonding capacity (Figure 1):

• Bifonazole, an antifungal with an imidazole ring. Its imidazole nitrogen is a hydrogen bond acceptor, and it has no traditional hydrogen bond donor. It was tested with PAA, poly(vinylpyrrolidone) (PVP, grade K-29/32), the vinylpyrrolidone–vinyl acetate copolymer with 40% acetate side chains (PVP/VA, Kollidon VA64) and hydroxypropyl methylcellulose acetate succinate (HPMCAS, AS-MF grade) M2·M3·M5·M6·M7.
• Nimesulide, a non-steroidal anti-inflammatory drug with a sulfonamide NH donor and several acceptors (nitro, sulfonyl, ether oxygen). It was tested with a low-molecular-weight PVP (K-12), PVP/VA and poly(vinyl acetate) (PVAc) M1·M4·M5·M8.

PVP and PVP/VA carry only hydrogen bond acceptors; PAA and HPMCAS carry donors as well. The effectiveness of a polymer was measured as its ability to reduce the crystal growth rate of the drug from the undercooled melt, close to Tg C5.

Background concepts

Crystal growth from the undercooled melt. Above its Tg, an amorphous drug is a viscous supercooled (undercooled) liquid below its melting point. Once a crystal nucleus has formed, it grows into the surrounding liquid. In this paper the crystals grew as spherulites, round aggregates of radiating crystal fibres, whose size increased linearly with time at a fixed temperature (Figure 2) R2·R4. The slope of the spherulite diameter against time is reported as the linear growth rate in m/s P3. Note by the AI writer (not discussed in the paper): the paper does not say whether the diameter slope was halved to a radial rate, so its absolute rates may differ by a factor of 2 from radius-based rates in other papers; ratios within the paper are unaffected X12.

Why growth slows so steeply near Tg. General background, not discussed in the paper: as an undercooled liquid is cooled toward Tg, molecular mobility drops by many orders of magnitude, and growth, which needs molecules to move to the crystal surface and attach, slows accordingly. The paper reports the result rather than the theory: between the highest temperature and 18 °C the growth rates of both drugs fell by roughly five to seven orders of magnitude (Figures 3 and 4) R7·R9.

Hydrogen bond donors and acceptors. A hydrogen bond forms between a donor, a hydrogen attached to an electronegative atom (N–H, O–H; C–H is a weak donor), and an acceptor, an atom with a lone pair (the imidazole nitrogen, a carbonyl oxygen, a nitro group). The authors rank donors by the electronegativity of the atom carrying the hydrogen and by neighbouring electron-withdrawing groups: the O–H of a carboxylic acid, next to its carbonyl, is "a much stronger donor" than a C–H group (ref 16) C15. Acceptors also differ in strength; the authors cite a scale (ref 27) by which the vinylpyrrolidone carbonyl of PVP and PVP/VA is a stronger acceptor than the acetate carbonyl of PVAc C26.

Self-association. Molecules hydrogen bond to their own kind, and a polymer has to compete with these contacts. In the bifonazole crystal the molecules are joined only by weak C–H···N hydrogen bonds forming chains along the b axis (ref 15) R18. Nimesulide uses its single NH donor twice in the crystal: in an intramolecular bond to the oxygen of its own ether group and in a weak intermolecular bond to the nitro group of a neighbour (ref 19); the intramolecular motif is common in crystals (ref 20) R19. PAA forms several self-associated structures, including carboxylic acid dimers (ref 17), although carboxylic acid self-association can be disrupted relatively easily (ref 18) R22.

Intramolecular versus intermolecular hydrogen bonds in IR. In IR spectroscopy, a hydrogen bonded N–H stretch appears at a lower wavenumber than a free one. In a non-polar solvent, an intermolecular hydrogen bond breaks up as the solution is diluted, so the NH band shifts; an intramolecular bond does not depend on concentration, so the band stays where it is (refs 22, 23). The authors use this dilution test on nimesulide in dichloromethane C21.

Antiplasticization. Adding a polymer with a higher Tg to a drug usually raises the Tg of the mixture. If the polymer stabilized the drug simply by raising Tg (antiplasticization), the dispersion Tg should track the inhibition; the authors measured it to test this C12.

DFT geometry optimization. Density functional theory (DFT) finds the lowest-energy shape of a single molecule. Here it was done in vacuum at the B3LYP/6-311++G(3d2f, 2p) level, after a first optimization with the 6-31G* basis set, to see whether nimesulide adopts the intramolecular hydrogen bond on its own P11.

Methods in detail

The study has two parts. The first measures how each polymer changes the crystal growth rate of each drug over a wide temperature range, and checks whether Tg or molecular weight can explain the ranking. The second probes the hydrogen bonds that nimesulide forms by itself and with the polymers, to explain why its polymers differ so little.

Contribution 1: growth rates of two drugs with polymers of different chemistry

The problem. To compare polymers fairly, the drug, the polymer level and the thermal history have to be the same in every sample, and the growth rate has to be measured over a temperature range wide enough to reach the slow regime close to Tg, where differences between polymers matter most for storage.

What they did. All polymers were vacuum-dried for 48 h. Each polymer was mixed at 5% w/w with the crystalline drug in a cryogenic mill (Spex 6750 freezer mill) P1. For each growth sample, 3–5 mg of drug or drug–polymer mixture was melted on an 18 × 18 mm cover glass, covered with an 18 mm round cover glass to give a transparent film, and cooled rapidly to 50 degrees below Tg. Under crossed polarizers the films showed no birefringence, so they were amorphous P2·R1. The films were stored over phosphorus pentoxide (0% relative humidity) at ambient temperature so that nuclei could form. The nuclei were then grown at set temperatures: a Linkam THMS 600 hot stage held 40–100 °C, and where growth was too slow to follow in one microscopy session, samples were kept in temperature-controlled ovens or coolers and taken out periodically for time-lapse pictures. Bifonazole was measured at 18 °C and from 30 to 100 °C in steps of 10 °C; nimesulide at 18 °C and from 30 to 80 °C. Each condition was run in triplicate, and the growth rate was the slope of the linear plot of spherulite diameter against time P3. Raman microscopy (783 nm laser) checked the polymorph of the grown crystals P4.

Polymer grades. PVP K-29/32 (MW 45,000–55,000 g/mol) was used with bifonazole and PVP K-12 (MW 2,000–3000 g/mol) with nimesulide; PVP/VA had MW 45,000–47,000, HPMCAS 18,000, PAA 1800 and PVAc 100,000 g/mol (Materials and methods) M3–M8. The PVP used with bifonazole is labelled "K-29/32" in the Materials and Discussion but "K 29/30" in the Results and Table 1; it is presumably the same material X5.

Contribution 2: testing Tg and molecular weight as explanations

The problem. If a polymer worked only by raising Tg (antiplasticization) or because of its own high Tg or large molecular weight, the ranking of the polymers would follow these properties, and drug chemistry would not matter.

What they did. DSC (TA Q2000, nitrogen purge, 3–5 mg in pin-holed Tzero pans) measured the onset Tg of each drug, each polymer and each 5 wt% dispersion on heating at 10 °C/min, after the cryomilled mixtures had been melted and cooled at 10 °C/min to 50 °C below Tg and the thermal history erased by a heat–cool–reheat cycle P5. The polymer molecular weights are those of the suppliers. The authors then compared the growth-rate rankings with these numbers (Tables 1 and 2) C12·C13.

Contribution 3: what hydrogen bonds nimesulide forms

The problem. For felodipine, the strength of drug–polymer hydrogen bonding ranked the polymers. If nimesulide's NH donor is already engaged, the polymers' acceptors have less to bond to, and the ranking could flatten.

What they did. Four complementary measurements P6–P11:

• DFT: the geometry of one nimesulide molecule was optimized in vacuum (Figure 5).
• Solution IR, dilution: nimesulide at 100, 50, 25, 10, 5, 2.5 and 1 mg/ml in dichloromethane, in a variable path cell with a 100 µm path; solvent bands removed by weighted subtraction (Figure S1).
• Solution IR, competitors: 50 mg/ml nimesulide with acetic acid (a hydrogen bond donor) or 1-ethyl-2-pyrrolidinone (an acceptor) added at a 1:1 molar ratio, to see whether a free donor or acceptor changes the NH band (Figure S4).
• Solid-state IR: crystalline nimesulide in a KBr pellet; amorphous nimesulide spin-coated from a 150 mg/ml dichloromethane solution onto a KRS5 disc (Figure 6). Drug–polymer films were cast from 1:1 ethanol:dichloromethane onto a heated KRS5 disc, dried at 90 °C and measured in a nitrogen glove box; the NH region was compared at 30 wt% polymer (Figure 7) and the polymer carbonyl regions at 20, 30, 40 and 50 wt% polymer (Figure 8, Figures S2 and S3). Higher polymer levels than in the growth samples were used so that the IR method could pick up differences, and to follow trends with concentration. Spectra: Bio-Rad FTS 6000, 128 scans, 4000–400 cm−1.

Note by the AI writer (not discussed in the paper): the IR films contain 20–50 wt% polymer and are made by solvent casting, while the growth samples contain 5 wt% and are melt-quenched. The IR shows what interactions can form, not how much of the drug is bonded in the growth samples X10.

Experimental analysis

Common ground: same crystals, single Tg, steep temperature dependence

Both drugs grew as spherulites at every temperature, with and without polymer (Figure 2). Their texture changed with temperature: bifonazole spherulites had rough, fibre-like edges at 18 °C, smoother edges as the temperature rose and sharper, less spherulitic edges at 90 and 100 °C; nimesulide spherulites had needle-like features at 18–40 °C, zig-zag edges at 40–60 °C and a smooth interface above that R2. Raman spectra showed the same polymorph at every temperature, with and without polymer, and the same as the starting material, so the growth rates always refer to one crystal form R3. Every 5 wt% dispersion showed a single Tg R5.

Growth was strongly temperature dependent. Without polymer, bifonazole grew at about 4.7 × 10⁻⁶ m/s at 100 °C and about 1.7 × 10⁻¹² m/s at 18 °C; nimesulide at about 8.5 × 10⁻⁶ m/s at 80 °C and about 3.7 × 10⁻¹¹ m/s at 18 °C (read from Figures 3 and 4) R7·R9. The text gives the bifonazole drop as "approximately 5.5 orders" of magnitude without polymer and 6 orders with the most effective polymer; the plotted points span about 6.4 and 6.9 orders. The half-order difference between the two agrees with the stated polymer factors of 5 and 15, but the absolute spans in the text are smaller than in the figure X6. For nimesulide the text gives 5–6 orders, in line with the plot (about 5.4 without polymer, 6.0 with PVP) R9.

Bifonazole: PAA stands out

All four polymers slowed bifonazole growth at 5 wt% (Figure 3). PVP and HPMCAS made it 2–6 times slower, with overlapping curves, and their effect was "somewhat" larger at low temperature; PVP/VA was reported to be similar to them, but its data are not shown C8·R8·X14. PAA lay below all other curves at every temperature: the authors give its effect as a factor of 5 at 100 °C and 15 close to Tg at 18 °C R7. Their ranking is PAA > PVP ~ HPMCAS ~ PVP/VA, and the abstract calls PAA "notably better" C9. Table A gives the readings at selected temperatures.

Table A. Bifonazole linear growth rate (m/s) at 5 wt% polymer, read from Figure 3 (approximate); ratio = rate without polymer divided by rate with polymer, computed by the AI from the readings except where the text states it R7.

Nimesulide: a smaller spread

For nimesulide (Figure 4), the three polymers lay close together at 60–80 °C, within about a factor of 2 of each other; below 50 °C they separated, with PVP and PVP/VA lowest and PVAc between them and the polymer-free curve R9. The text says that nimesulide grows "2–10 times faster" without polymer. The plot agrees for PVAc (about 2–4.5-fold across the range) but shows larger effects for PVP and PVP/VA close to Tg: about 21-fold and 16-fold at 18 °C, and about 11-fold for PVP at 40 °C X7. The authors' ranking at 18 °C is PVP ~ PVP/VA > PVAc; they write that all polymers "appear to somewhat be effective" and that the differences were "more modest" than for bifonazole C10. They also note that this is "somewhat in contrast" to felodipine, where the same three polymers at 3 wt% differed much more, in the order PVP > PVP/VA > PVAc (ref 14) C11·R21.

Table B. Nimesulide linear growth rate (m/s) at 5 wt% polymer, read from Figure 4 (approximate); ratios computed by the AI from the readings R9.

Note by the AI writer (not discussed in the paper): the lowest temperature, 18 °C, lies below the onset Tg of nimesulide (19.5 °C) and of its dispersions (19.9–20.8 °C), so the 18 °C nimesulide rates were measured at or just below Tg, whereas for bifonazole (Tg 15.2 °C) 18 °C is just above it X13.

Tg and molecular weight do not track the ranking

With 5 wt% polymer the onset Tg rose by only 0.4–2.2 °C (Table C), which the authors describe as no significant change; antiplasticization therefore cannot account for the inhibition R6·C12. The Tg of the pure polymer did not rank the inhibitors either: PAA (Tg 103.0 °C) beat PVP K-29/32 (Tg 163.9 °C) for bifonazole. Nor did molecular weight: for bifonazole the order of molecular weight is PVP K-29/32 > PVP/VA > HPMCAS > PAA, yet PAA, the smallest, was best; for nimesulide, PVP K-12 (Mw 2500 g/mol) beat PVAc, whose molecular weight is 40 times larger (Mw 100,000) C12·C13. Both results match the felodipine study.

Table C. Onset Tg from Tables 1 and 2 (standard deviations within ± 1 °C, stated under Table 2); molecular weights as supplied (Materials and methods) R6·M1–M8. The Results text gives the drug Tg values as 20 °C (nimesulide) and 15.5 °C (bifonazole), slightly different from the tables X4.

Note by the AI writer (not discussed in the paper): two details qualify "no correlation". For nimesulide, the ranking PVP ~ PVP/VA > PVAc does follow the polymers' Tg (102.9 and 102.6 °C against 38.7 °C), so the Tg counterexample comes from bifonazole alone. And in both drugs the dispersion with the highest Tg was made with the best inhibitor (PAA–bifonazole 17.4 °C; PVP–nimesulide 20.8 °C), although the differences are close to the stated ± 1 °C X8.

Nimesulide holds its NH donor in an intramolecular hydrogen bond

Three pieces of evidence point to the intramolecular NH···O bond. The DFT-optimized geometry places the sulfonamide hydrogen next to the ether oxygen, the same intramolecular bond seen in the crystal (Figure 5) R10·C20. In dichloromethane, the NH band stayed at 3345 cm−1 at every concentration from 1 to 100 mg/ml (Figure S1); because an intermolecular bond would shift on dilution, the authors conclude ("it can be concluded") that nimesulide forms an intramolecular hydrogen bond in solution R11·C21. Finally, the NH band of the amorphous drug lies close to that of the crystal (about 3276 against 3288 cm−1 in Figure 6, with a broader band for the amorphous form), which the authors read as "a similar hydrogen bond environment": the supercooled liquid is "expected" to keep the intramolecular bond and "possibly" to form additional intermolecular ones R12·C22. The abstract puts this more strongly: IR and DFT "strongly suggested" that amorphous nimesulide formed an intramolecular hydrogen bond C22.

Nimesulide–polymer hydrogen bonds are present but weak

In cast films with 30 wt% polymer, the nimesulide NH band changed only slightly: its maximum moved from about 3281 cm−1 without polymer to about 3258–3265 cm−1 with each of PVP, PVP/VA and PVAc (read from the NH spectra, see below), which the authors call "small changes" "suggesting some extent of drug-polymer interaction" R13·C23. In the PVP carbonyl region, the band of pure PVP (about 1684 cm−1) moved to about 1680 cm−1 in the dispersions and grew a shoulder near 1657 cm−1, which the authors associate with drug–polymer hydrogen bonds, "confirming" interactions in nimesulide–PVP films (Figure 8) R14·C23. PVP/VA showed similar changes in its vinylpyrrolidone carbonyl (Figure S3), while the acetate carbonyl of PVAc barely changed (about 1739 to 1738 cm−1, Figure S2) R15·R16·C25. Adding acetic acid or 1-ethyl-2-pyrrolidinone 1:1 to a nimesulide solution left the NH band at about 3345 cm−1 (Figure S4) R17.

Table D. IR band positions. The 3345 cm−1 value and the felodipine shifts are stated in the text; all other positions are read from the spectra (approximate) R11–R17·R20.

All these shifts, the authors write, are "much smaller" than in systems with similar groups, such as felodipine–PVP, where the felodipine NH band shifted by about 45 cm−1 and the PVP carbonyl by about 20 cm−1 with a more distinct shoulder (ref 25) C24·R20. The paper gives no shift values for nimesulide; the readings above (an NH shift of roughly 16–23 cm−1 and a carbonyl shift of roughly 4 cm−1 plus the shoulder) fit the qualitative statement X15.

A figure problem to know about. In the published article the images of Figures 7 and 8 are swapped relative to their captions. The image printed above the Figure 7 caption shows the carbonyl region (1720–1630 cm−1) with an arrowed shoulder, which is what the Figure 8 caption describes; the image above the Figure 8 caption shows the NH region (3400–3150 cm−1) with four line styles, which is what the Figure 7 caption describes. The text cites the figures in line with the captions X1. The text also cites Figures S2 and S3 for NH-region changes, although they show polymer carbonyl and acetate regions, and Figure S4 for the PVAc carbonyl, which is in Figure S2 X2.

Discussion

The authors' explanation: free hydrogen bonding groups decide

Having ruled out Tg and molecular weight, the authors turn to the chemistry of drug and polymer together. Their starting point, from earlier work, is that polymers "may" inhibit growth by slowing the drug–drug interactions needed for molecules to attach to a growing crystal (Taylor and Zografi, ref 10) C19. A polymer should therefore work best when it offers the drug a hydrogen bonding partner that beats the drug's own self-association C15. The chain splits by drug:

Bifonazole. Because the bifonazole crystal holds together only through weak C–H···N bonds (ref 15), any polymer with a donor stronger than C–H "should be capable" of forming more favourable hydrogen bonds with the imidazole nitrogen than those in the pure drug, disrupting drug–drug interactions and delaying crystallization. "This is indeed the case with PAA": its carboxylic acid O–H is a much stronger donor than C–H. PAA's own self-association has to be broken first, but the imidazole nitrogen "would be expected" to be a better acceptor than the PAA carbonyl, and carboxylic acid self-association is disrupted relatively easily (refs 17, 18) C15·C16·R18·R22. PVP has acceptors only and "can at most" form very weak hydrogen bonds with the acidic C–H of bifonazole, or interact through dipole–dipole or non-specific forces C17. HPMCAS has donors, but the drug must compete with the many HPMCAS acceptors, HPMCAS carries far fewer carboxylic acid groups than PAA (1 against 13.9 mmol/g), and steric factors "may" limit access, so HPMCAS "may be hydrogen bonding with bifonazole but not as efficiently as PAA" C18·M6·M7. In the abstract the authors put the PAA effect cautiously: it "may be related to formation of specific interactions between the drug and the polymer" C15.

Nimesulide. Here the drug's single donor is busy. The intramolecular NH···O bond, present in the crystal, in solution and, by inference, in the amorphous drug, "considerably diminished" the potential for drug–polymer hydrogen bonding; the unchanged NH band when a strong donor or acceptor is added to a solution lends support to this, in the authors' words ("lend support"). They conclude that this "most likely" explains why the differences between the polymers are smaller than anticipated, and that the remaining edge of PVP and PVP/VA over PVAc "may be related" to their stronger acceptor (ref 27) C27·C26. In the abstract they write that drug–polymer hydrogen bonding "did occur, but was hampered by the partial loss of the hydrogen bond donor to the intramolecular hydrogen bond" C27.

The general statement. The authors conclude that the choice of polymer to delay crystallization "clearly depends" on the chemistry of the drug and on the competition between drug and polymer groups C29; that the result "highlights the importance" of considering intramolecular interactions C28; and, in the abstract, that the effectiveness of a polymer "appears to depend not only on the chemistry of the drug, but also on the availability of hydrogen bonding groups" C30. In the Conclusions, polymer effectiveness "could be rationalized" by the likely intra- and intermolecular interactions in the pure systems and in the dispersions C14.

How it compares with felodipine

The authors cite felodipine–PVP as a system with similar functional groups: there the felodipine NH band shifted by about 45 cm−1 and the PVP carbonyl by about 20 cm−1 (ref 25), and at 3 wt% the same three polymers used here for nimesulide gave clearly different growth rates in the order PVP > PVP/VA > PVAc (ref 14) R20·R21. Nimesulide, with its NH partly engaged intramolecularly, shows smaller IR shifts and a smaller spread between the same polymers. Note by the AI writer (not stated in the paper): read together, the two drugs suggest that a drug donor free to bond gives polymers with strong acceptors something to work with, and a locked donor does not; but this comparison crosses two papers, two drugs and two polymer levels (3 and 5 wt%), so it is suggestive rather than a controlled test.

What it suggests for formulation

The authors draw a general lesson rather than a rule: polymer selection needs better rules, with "an improved emphasis on considering the chemistry of the drug" C32, and a quantitative study with more drugs combined with an in silico approach "should provide a larger database and better selection rules" C31.

Note by the AI writer (not stated in the paper): in practice the two cases translate into a simple screening question. For a drug whose only strong hydrogen bonding site is an acceptor, such as a basic nitrogen, a polymer with acidic donors deserves an early place in the screen; here PAA beat polymers with higher Tg and higher molecular weight by a wide margin at the same 5 wt% level. For a drug whose donor is tied up within the molecule, expect little discrimination between acceptor polymers, and choose on other grounds such as processing or dissolution. In both cases the rankings come from growth in a dry melt near Tg at 5 wt% polymer; nucleation, humidity, higher polymer loadings and dissolution behaviour were outside the study.

Limitations

Limits the authors state. The authors say that the differences between polymers for nimesulide were smaller "than would be anticipated" and attribute this to the intramolecular hydrogen bond C27. They call for a larger, quantitative data set with an in silico approach before general selection rules can be made C31. They do not discuss other limits of the design.

The AI writer's own judgement.

• The bifonazole mechanism was not measured. No IR or other spectroscopy was run on bifonazole or its dispersions; the PAA–imidazole hydrogen bond, the breaking of PAA dimers and the weaker HPMCAS bonding all rest on the crystal structure and general hydrogen bond chemistry X9. The growth data show that PAA is best, not why.
• IR and growth samples differ. IR films held 20–50 wt% polymer and were solvent cast; growth samples held 5 wt% and were melt-quenched X10.
• The acceptor-strength argument is not resolved by the NH data. In the NH spectra the maximum moves about as far with PVAc as with PVP and PVP/VA (about 3258–3265 against 3281 cm−1); the polymers differ mainly in intensity at 3150–3200 cm−1 X16.
• Text and figures disagree in places. The bifonazole temperature span (5.5 orders in the text, about 6.4 in Figure 3), the nimesulide polymer factor (2–10 in the text, about 16–21 at 18 °C in Figure 4), the drug Tg values (text 15.5 and 20 °C, tables 15.2 and 19.5 °C) and the PVP grade label (K-29/32 or K 29/30) X6·X7·X4·X5. Figures 7 and 8 have swapped images, the Figure 2 caption labels both rows (a)–(c), and two supporting-figure citations point to the wrong figure X1·X3·X2.
• "No Tg correlation" rests on one pair. For nimesulide the ranking follows polymer Tg, and in both drugs the dispersion with the highest Tg came from the best inhibitor, although by little X8.
• Design details. Different PVP grades were used for the two drugs (K-29/32 and K-12) X11; the bifonazole PVP/VA data are not shown X14; the growth rate is the diameter slope, with no statement on halving X12; the 18 °C nimesulide point lies at or below Tg X13; and only two drugs at one polymer level were studied.

Overall, the growth-rate observations (every polymer helps; PAA dominates for bifonazole; nimesulide's polymers differ modestly; Tg and molecular weight do not explain the order) are direct and consistent. The intramolecular bond in nimesulide is well supported. The explanations that connect chemistry to inhibition are plausible but, for bifonazole, untested, and for nimesulide rest on IR at higher polymer levels.

Conclusion

This paper extends the felodipine study from the same group to two drugs with different hydrogen bonding chemistry. At 5 wt%, every polymer slowed crystal growth from the undercooled melt, by factors that grew toward Tg. For bifonazole, an acceptor-only drug, PAA was by far the best inhibitor (5-fold at 100 °C, 15-fold at 18 °C), ahead of PVP, HPMCAS and PVP/VA (2–6-fold). For nimesulide, whose NH donor is largely held in an intramolecular hydrogen bond, PVP and PVP/VA were somewhat better than PVAc at low temperature, and the differences were smaller than for bifonazole or for felodipine. Polymer Tg, dispersion Tg and molecular weight did not explain either ranking. The authors' reading is that polymer effectiveness "appears to depend not only on the chemistry of the drug, but also on the availability of hydrogen bonding groups" C30: a polymer helps most when it can offer a free group on the drug a better partner than the drug offers itself.