Which polymers slow the crystal growth of supercooled flutamide, and why hydrogen bonding does not predict its nucleation

Background and motivation

An amorphous drug has a higher free energy than its crystal, which gives it a higher transient solubility and faster dissolution, but it also tends to crystallize; slowing that transformation is the main challenge in making amorphous dosage forms. Two factors set the rate: the thermodynamic driving force, which grows as a melt is cooled further below its melting point, and molecular mobility, which falls steeply on cooling and drops sharply at the glass transition temperature, $T_g$. C1 Storing the glass far below $T_g$ would work in principle, but many drugs have a low $T_g$ and some crystallize well below it, so the practical route is to add a polymer, described in the introduction as the most effective strategy. C1

Three mechanisms had been proposed for how a polymer inhibits crystallization: it can raise the $T_g$ of the dispersion and so lower mobility at the storage temperature (Van den Mooter 2001), act as a diluent that lowers the chemical potential of the drug, or form specific drug–polymer interactions that restrict drug mobility locally (Taylor and Zografi 1997). C2 Earlier work from the same group on felodipine (Kestur and Taylor 2010) had found that, among these factors, crystal growth rates "appear" to correlate with drug–polymer hydrogen bonding; the authors note that confirming this requires drugs of more diverse chemistry. C3

The authors also point to a weakness in how polymers are usually compared. Most studies use equal weight percent, which gives polymers with heavy repeat units fewer functional groups per gram, and the common pharmaceutical polymers carry several kinds of hydrogen-bonding groups, which makes any interaction hard to interpret. Their proposal is to use chemically simple polymers, each with one kind of group, and to compare them at a defined molar ratio of repeat units. C4

The paper therefore tests one hypothesis: the polymer that can form the strongest or most numerous hydrogen bonds with the drug will be the best inhibitor of crystal growth in the supercooled liquid. C5 The model drug is flutamide, 2-methyl-N-[4-nitro-3-(trifluoromethyl)phenyl]-propanamide, a small molecule (276.2 Da) with one hydrogen-bond donor, the amide NH, and two acceptors, the amide carbonyl and the nitro group (Figure 1). M1 Six simple vinyl polymers and one small molecule, the structurally related nilutamide, were tested at a fixed molar ratio; three pharmaceutical polymers and PVP were tested at 5% w/w; and the effect of the same additives on nucleation was evaluated, by the authors' own description, "qualitatively". A1·A2·C6

Background concepts

Supercooled liquid, nucleation and growth. A supercooled liquid is a melt held below its melting point without crystallizing. Crystallization from it needs two steps: nucleation, the birth of stable crystal nuclei, and growth, the advance of the crystal front into the liquid. Nucleation begins with small clusters (prenuclei) that have short-range, crystal-like order; most redissolve because the favourable free energy of the new volume is outweighed by the cost of the new interface, and only clusters larger than a critical radius become stable nuclei (Mullin 2001). C46 The two steps usually peak at different temperatures: nucleation lower, closer to $T_g$, and growth higher, closer to the melting point $T_m$. Figure 5 of the paper draws these as a nucleation zone ($J$) and a growth zone ($U$) between $T_g$ and $T_m$. R10 Flutamide is an extreme case: its $T_g$ is about −3 °C and it melts at 110 °C, so at the growth temperatures used here (40–70 °C) it is a mobile liquid 43–73 °C above $T_g$. R7.a·P2

Growth rate from spherulites. Flutamide crystallizes from the melt as spherulites, round polycrystalline domains visible under crossed polarizers. The paper follows the diameter of a spherulite over time at a fixed temperature; the diameter grows linearly and its slope is the "diameter growth rate", in µm/s. R2·P2 This is twice the radial growth velocity that many other papers report, which matters when comparing numbers across papers. X13

Reading nucleation from a DSC exotherm. When a glass is heated in a DSC, it crystallizes at some temperature and gives an exothermic peak, whose maximum is called $T_{max}$. Where this peak appears depends on how many nuclei are present: with many nuclei, the sample crystallizes early, at a low temperature where growth is still slow; with few nuclei, the crystals must grow larger, which happens later, at a higher temperature where growth is faster (Hernández Sánchez 2005; Masirek 2007). C17 The paper uses this logic in reverse: knowing growth rate as a function of temperature for each sample, it reads off the growth rate at $T_{max}$. If an additive raises the growth rate at $T_{max}$ compared with the pure drug, the sample had fewer nuclei, so the additive lowered the nucleation rate. The graphical abstract draws this idea. C17·P7

Hydrogen bonds and infrared shifts. A hydrogen bond links a donor group (NH, OH) to an acceptor (a carbonyl oxygen, a nitro oxygen). In an infrared spectrum the stretching band of the donor moves to a lower wavenumber when its hydrogen bond becomes stronger, and the acceptor's carbonyl band also moves down. Infrared spectroscopy has long been used this way to detect drug–polymer interactions in amorphous dispersions (Taylor and Zografi 1997; Andrews 2010). C24 When a mixture's spectrum is simply the sum of the two pure spectra, the components are not interacting, which is typical of immiscible blends (Rumondor 2009). C29

Intramolecular hydrogen bonds and partial charges. A hydrogen-bond acceptor can be engaged inside its own molecule. In crystalline flutamide (Cambridge Structural Database code WEZCOT), the amide carbonyl oxygen makes a bifurcated contact: one CH···O bond to a neighbouring molecule and one intramolecular CH···O bond to an aromatic hydrogen, which keeps the molecule planar (Cense 1994). R26 Such a contact uses up part of the oxygen's negative charge. The paper estimates acceptor strength with natural bond orbital (NBO) partial charges from density functional theory (DFT): the more negative the carbonyl oxygen, the better the acceptor. P9·R25

Flutamide's two polymorphs. Earlier thermal work (Ceolin 1995) proposed that flutamide glass first crystallizes as a metastable form that then rapidly converts to the stable form, the two events overlapping in a single DSC exotherm at 10 °C/min and separating at slower heating; the two forms were suggested to be built from different conformers. C12 This becomes relevant for some of the polymers below.

Methods in detail

The study has three strands that run in parallel on the same set of additives: crystal growth rates from the melt, a DSC and microscopy reading of nucleation, and a spectroscopic and computational reading of drug–additive hydrogen bonding. The comparison at the end sets the growth ranking against the interaction measures and against the change in $T_g$.

Samples: equal molar ratio instead of equal weight

Problem. If polymers are compared at equal weight, a polymer with a light repeat unit offers more functional groups than one with a heavy repeat unit, and any ranking mixes chemistry with group count. C4 What they did. Flutamide was mixed with each of six chemically simple vinyl polymers, each carrying one kind of side group (Figure 1), and with nilutamide, always at 9 flutamide molecules per polymer repeat unit (or per nilutamide molecule), i.e. 10 mol% additive. P1·A1 The six polymers were poly(N,N-dimethyl acrylamide) (PDMA, tertiary amide, M_w 100 000), poly(vinyl acetate) (PVAc, ester, M_w 350 000), poly(acrylic acid) (PAA, carboxylic acid, M_w 450 000), poly(4-vinylphenol) (PVPH, phenol), poly(N-isopropylacrylamide) (PIPA, secondary amide) and poly(vinylpyrrolidone) K12 (PVP, lactam, M_w 2000–3000). M3·M4·M5·M6·M7·M8 On a weight basis this corresponds to 2.8% PAA, 4.3% PVP, 3.8% PDMA, 4.6% PVPH, 3.3% PVAc, 4.6% PIPA and 11.3% nilutamide. P1 Card check: with the N-isopropylacrylamide repeat unit drawn in Figure 1 (113.16 g/mol, not printed), 10 mol% PIPA is 4.35 wt%, not 4.6%; the other six values reproduce, and 4.6% equals the PVPH value, so a copy error is likely (note X1). X1 As a check closer to practice, hydroxypropyl methylcellulose (HPMC, type 2910), HPMC acetate succinate (HPMCAS, grade AS-MF), PVP–vinyl acetate copolymer (PVPVA, Kollidon VA64) and PVP K12 were tested at 5% w/w. M9·M10·M11·P1·A2 All mixtures were blended in a mortar and pestle and then cryomilled for 10 min. P1

Crystal growth rates

About 3 mg of each mixture was melted between two coverslips at 120–125 °C and quenched to room temperature, giving a clear film that showed no birefringence under crossed polarizers. Because nuclei appeared and grew slowly even at room temperature, each film was measured immediately. P2·R1 On a hot stage at 40, 50, 60 or 70 °C, the growing spherulites were photographed at intervals; the diameter rose linearly with time, and its slope gave the diameter growth rate; each condition was run in triplicate. P2·R2 Raman spectra of the crystallized films identified the polymorph that grew. P3·R6 Why it works. At equal molar ratio the dilution of the drug is the same with every polymer, so differences in growth rate should come from what each polymer does to the drug, not from how much of it there is. C32

Thermal analysis and nucleation

DSC (TA Q2000, nitrogen purge, sealed pans with a pinhole) gave the onset $T_g$ (melt cooled at 15 °C/min, then reheated at 10 °C/min) and the melting temperature of each 10 mol% mixture (Table 1). P5·R7 For nucleation, melts were cooled at 5 °C/min to a series of temperatures and then reheated at 10 °C/min to see whether crystallization had been initiated; the range of temperatures that produced nuclei is the nucleation zone. P6·R10 The average temperature of the crystallization exotherm maximum ($T_{max}$, two runs) was then combined with each sample's growth rate–temperature data to give the growth rate at $T_{max}$, and each additive was compared with pure flutamide at P = 0.05 (the statistical test is not named). P7·R11·X4 As a direct visual check, films melted between coverslips were cooled at 15 °C/min to −30 °C, reheated at 10 °C/min to about 40–60 °C, and a small region was photographed (Figure 6). P8·R12

Interactions: FTIR and DFT

For infrared spectroscopy, flutamide and each additive were dissolved in ethanol at a 1:1 molar ratio (drug to repeat unit), spin-coated at about 2000 rpm onto ZnSe, and measured in transmission (128 scans, 4000–400 cm⁻¹, 0% RH). The NH stretching band of flutamide (3500–3200 cm⁻¹) and the carbonyl region were compared with the spectra of the pure components. P4 Crystalline and amorphous flutamide were also compared, to see how the drug bonds to itself in each state (Figure 7). R13 With DFT (B3LYP, SPARTAN'08), the authors found the lowest-energy conformation of a single flutamide molecule in vacuum, computed its energy as the bond between the phenyl ring and the amide group was rotated in 20° steps (Figure 12), and computed NBO charges on the carbonyl oxygen of each polymer repeat unit (Table 4). P9·R23·R24·R25 Note by the AI writer (not discussed in the paper): the FTIR samples are 1:1 films made from solution at room conditions, while growth was measured on 9:1 melts at 40–70 °C, so the infrared results rank how strongly each polymer can bind the drug rather than describe the growth samples themselves. X9

Experimental analysis

Result 1: carbonyl polymers slow growth most, donor polymers hardly at all

Pure flutamide grows very fast from its melt: about 77 µm/s at 70 °C, falling by nearly two orders of magnitude to about 1.4 µm/s at 40 °C (diameter growth rates read from Figure 2). R3.a With 10 mol% additive the rates spread over about one order of magnitude, and the additives kept the same order at every temperature (Table A). R3

Table A. Growth rates read from Figure 2 (approximate; triplicate), NH band positions from Table 3, NBO charges on the carbonyl oxygen from Table 4. The columns are measured or computed separately; placing them side by side is the AI writer's arrangement. R3·R14·R25

The authors rank the additives PVP > PDMA = PIPA > nilutamide > PVPH > PVAc > PAA. PVP lowered growth by about one order of magnitude; PDMA and PIPA had very similar effects; PVAc had very little effect and PVPH a slightly better one; nilutamide beat both and was about half as effective as PVP; and PAA, the least effective, gave a slightly higher average rate than the pure drug, a difference the authors call not significant. C7 The PAA melt was also the only 10 mol% sample to crystallize partly while being cooled. R4 Raman showed that the same crystal form (WEZCOT) grew in every sample except one with HPMC, so the growth comparison is not complicated by a change of form. R6 From this ranking the authors draw their first reading: "It thus appears that polymers with strong acceptor groups are more effective" than polymers with donor groups, although flutamide itself has both a donor and acceptors. C8

Reading Figure 2. The plot is on a log scale with many overlapping markers. At 40 and 60 °C the pure-drug marker is hidden behind other markers, and the caption's colour names for PVPH ("orange solid square") and nilutamide ("red solid square") correspond to a small light-red and a larger dark-red square, so the values for those series are less certain than the others. X10

Result 2: the pharmaceutical polymers follow the same pattern

Table B. Growth rates read from Figure 3 (approximate); summaries from the text. R5·C9

At 5% w/w all four pharmaceutical polymers slowed growth, with PVP again best, by more than an order of magnitude (Table B). R5·C9 PVPVA was weaker than PVP; the authors call this "expected", since its vinyl acetate units resemble PVAc, which was a poor inhibitor, and explain it "presumably" by the much weaker interactions of vinyl acetate (Table 4). C9·C41 The cellulose polymers were least effective, "presumably" because their acceptor groups are less basic (Van Eerdenbrugh and Taylor 2011). C41 With HPMC another polymorph grew, and DSC showed its metastable form strongly stabilized, with a melting onset of 90 °C; with HPMCAS, flutamide partly crystallized while the melt was cooled. R6·R9·C10·C14 Card check: the Figure 3 caption also lists a "5% w/w PAA (red solid square)" series, which the methods and text do not mention; its values (about 1.6, 9.0, 27 and 89 µm/s at 40–70 °C) coincide with the 10 mol% PAA series of Figure 2, so it is probably those data replotted, and the text's "all the polymers were effective" does not apply to it (note X2). X2

Result 3: the glass transition barely moves

Table C. Onset T_g and T_m at 10 °C/min from Table 1 (standard deviations within 1 °C); the "rise" column is computed by the AI writer. R7

The polymers raised $T_g$ by only a few degrees, from −3.1 °C to between −1.07 and 0.6 °C, with little difference between them; nilutamide, whose own glass has a $T_g$ of about 32 °C, raised it most, to 2.5 °C, and gave a eutectic rather than a flutamide melting point. R7·M2·C11 The melting point stayed at 108–110 °C. R7 The authors use these numbers to exclude two alternative explanations before turning to hydrogen bonding. Dilution is the same for every polymer by design. A higher $T_g$ lowers mobility and so growth, but the differences between the dispersions are too small to explain the growth differences, and nilutamide, with the largest $T_g$ effect, was not the best inhibitor, "suggesting that effects on mobility as manifested by changes in Tg cannot completely explain effects on crystal growth rate". The nearly constant $T_m$ means the degree of supercooling is unchanged. C32·C33 Note by the AI writer (not discussed in the paper): molecular mobility at the growth temperatures was not measured directly; the argument rests on $T_g$ as a proxy, which is reasonable for differences of 2–4 °C at temperatures 43–73 °C above $T_g$. C33

Result 4: the NH shift follows the growth ranking

Amorphous flutamide bonds to itself more strongly than the crystal does. In the crystal the NH band is sharp at 3360 cm⁻¹ and the carbonyl band sharp at 1716 cm⁻¹; in the supercooled liquid the NH band broadens and moves down to 3315 cm⁻¹, and the carbonyl band splits into a stronger peak at 1681 cm⁻¹ and a shoulder at 1710 cm⁻¹ (Figure 7). R13 The authors read this as stronger hydrogen bonding "on average" in the glass ("suggests"): in the crystal the NH binds the weak nitro acceptor, while in the glass it "appears" to bind the amide carbonyl, and the shoulder is a population of carbonyls that keep the crystal-like, non-NH-bonded state. C21·C22 They add, as a possibility, that this more extensive bonding "might contribute" to flutamide forming a glass at all, since a molecule this simple would be expected to crystallize on cooling (Baird 2010). C23

With polymers, Table 3 gives the NH band in 1:1 films (Table A). PVP, PDMA and PIPA move it down to 3287, 3289 and 3293 cm⁻¹; the largest shift, 28 cm⁻¹, is with PVP. R14 The polymer carbonyl bands move down too, from 1679 to 1658 cm⁻¹ for PVP (Figure 8) and from 1641 to 1620 cm⁻¹ for PDMA (Figure 9), and with PDMA the flutamide carbonyl shoulder near 1710 cm⁻¹ grows, which the authors read as fewer drug–drug and more drug–polymer bonds ("likely"). R15·R16·C25·C26 Their conclusion: "it appears that the CO groups in the polymers act as competitive acceptors for the NH group in flutamide". C26

The other additives behave differently. PVAc raises the NH band slightly, to 3322 cm⁻¹, although it has a carbonyl; the authors explain that its carbonyl is attached directly to an oxygen, which lowers its negative charge and makes it a weak acceptor, and they state that the extra intensity near 1710 cm⁻¹ cannot be assigned. R17·C27·C28 PAA changes no band at all, and the mixture spectrum is the sum of the two pure spectra, "typical of immiscible systems" (Figure 11). R18·C29 PVPH leaves the flutamide carbonyl unchanged and shows small changes in the NH/OH region, "indicating some level of interaction". R19·C30 Nilutamide moves its own carbonyl band from 1717 to 1726 cm⁻¹, flutamide's from 1680 to 1686 cm⁻¹, and the NH band up to 3321 cm⁻¹; "Clearly there is some interaction", but it does not appear stronger than the bonds within each drug. R20·C31

Putting the columns of Table A together, the authors find that among the carbonyl polymers a larger downward NH shift goes with stronger growth inhibition, and that the computed carbonyl oxygen charges, from −0.598 (PVAc) to −0.65 (PVP), "also match" the extent of growth reduction. "Both of these pieces of evidence strongly support our hypothesis that specific interactions between the drug and the polymer are extremely important in determining the extent of crystal growth inhibition." C39·C40 Card check: the charge order fits the four carbonyl acceptor polymers, but PAA (−0.606) is more negative than PVAc (−0.598) while being the weaker inhibitor; and PIPA and PDMA, equal in growth, differ by 0.007 in charge and by 4 cm⁻¹ in NH shift. The correlation is a qualitative ranking of four polymers (note X8). X8

Result 5: flutamide's carbonyl is a poor acceptor

The puzzle is why the two polymers with the best donor groups, PAA and PVPH, failed: flutamide's amide carbonyl looks like the best acceptor in the molecule, better than any acceptor in those polymers, and both PAA and PVPH are very effective growth inhibitors for acetaminophen, another amide drug (Trasi and Taylor 2012; Miyazaki 2004). C34·R27 The authors draw two possibilities: either the hypothesis is wrong, or these polymers do not form good hydrogen bonds with the flutamide carbonyl; the unchanged carbonyl band in FTIR is "consistent with" the second. C34

The crystal structure and the calculation explain why. In crystalline flutamide the carbonyl oxygen is engaged in a bifurcated contact, one of whose arms is an intramolecular CH···O bond that keeps the molecule planar (Cense 1994). R26 The lowest-energy conformation of a single molecule in vacuum is also planar with this contact. R23 Rotating the amide group away costs energy: the profile in Figure 12 has its minimum with the carbonyl next to the hydrogen on C(5), opposite the CF₃ group, a second minimum about 2.7 kJ/mol higher when the amide is turned by 180° (where it "presumably" faces the hydrogen on C(3)), and barriers of about 23 kJ/mol between them (values read from Figure 12, approximate). R24·C36 The authors attribute the stronger C(5) contact to a higher positive charge on that hydrogen, set by the positions of the two electron-withdrawing groups. C36 From this they conclude that the intramolecular bond is "highly likely" kept in the amorphous state, that part of the oxygen's negative charge is used up inside the molecule, and that "the intramolecular hydrogen bond renders the amide carbonyl a much less favorable acceptor group". C35 PAA, which forms hydrogen-bonded dimers with itself (Dong 1997), then "very likely" has no reason to bond to the weaker flutamide carbonyl. C37·R28 Consistent with this, flutamide–PAA blends with more PAA appear immiscible: their IR spectra are summed spectra, and a 50 wt% blend has the same $T_g$ as flutamide. R21·C38

Note by the AI writer (not discussed in the paper): the intramolecular contact in the glass is inferred from a single-molecule calculation in vacuum and from the crystal structure; no measurement on amorphous flutamide isolates it. PAA also crystallized flutamide on cooling and is immiscible at higher content, so its failure could partly reflect poor mixing with the melt rather than the intramolecular bond alone; the paper does not separate the two. C35·X12

Result 6: nucleation does not follow hydrogen bonding

Table D. Nucleation zones as labelled on the schematic Figure 5; T_max, growth rate at T_max and the verdicts from Table 2 (T_max standard deviation within 1 °C, n = 2; the meaning of the values in parentheses is not stated). R10·R11·R4

Pure flutamide nucleates between 10 and 40 °C. R10.a The authors summarize Figure 5 as showing that "the only major effect" of the polymers was a slight narrowing of this zone; PAA, which crystallized on cooling, "seemed to promote nucleation", and nilutamide "appeared to be an effective nucleation inhibitor", crystallizing very little in DSC, "most likely" because its nucleation rate fell too low for nuclei to form at the DSC rates. C15·C16 The growth rate at $T_{max}$ (Table D) adds a statistical reading: PDMA, PVPH and PVP gave growth rates at $T_{max}$ not significantly different from the pure drug's 4.4 µm/s, so they "did not appear" to change the nucleation rate, whereas PVAc (7.7 µm/s) and PIPA (7.0 µm/s) lowered it, leaving fewer nuclei that crystallized at a temperature where growth is faster. R11·C18 Figure 6 shows the contrast most directly: after cooling to −30 °C and reheating, pure flutamide and the PIPA film are covered with spherulites, while the nilutamide film shows a single nucleus, the "drastic" effect the authors describe. R12·C19

The authors' conclusion is that hydrogen bonding between drug and additive has "a very clear effect" on growth, but "a similar correlation is not observed in this study with respect to the nucleation behavior", as in their earlier felodipine work (Konno and Taylor 2006). C45 Card check: the two polymers classed as lowering nucleation, PIPA and PVAc, are one of the best and one of the weakest growth inhibitors, so the data show no consistent relation in either direction; the conclusion that polymers "effective at reducing crystal growth rates were not found to have a similar impact on nucleation" does not hold for PIPA, and the abstract's "only two polymers" leaves out PAA and HPMCAS, which crystallized on cooling (note X5). The image of the PIPA film in Figure 6 does not show fewer nuclei than pure flutamide (note X6). X5·X6

Result 7: nilutamide acts by other routes

Nilutamide slowed growth less than PVP but more than PVPH (Table A), raised $T_g$ most, and by FTIR did not appear to bond to flutamide more strongly than each drug bonds to itself. R3.e·R7.h·R20·C31 The authors write that its effect on growth "can be explained by the combination of a dilution effect and an increase in Tg": corrected for both, its growth curve overlaps the pure-drug data (the corrected curve and the correction method are not shown). C42·R22·X11 For nucleation they propose, with explicit hedges, that nilutamide, being structurally similar to flutamide, "may" be incorporated into flutamide prenuclei by molecular recognition and, being larger, that it is "quite possible" it then blocks further growth, a mechanism proposed for impurities crystallizing from solution (Weissbuch 1995; Anwar 2009). C47 The abstract states the cause differently, as "most likely due to its low molecular size". C48·X7 Neither version was tested.

Discussion

The authors' chain of reasoning. The argument runs in four steps. Equal molar ratios make dilution equal, and the small $T_g$ changes rule out mobility as the main cause, so the remaining explanation is specific interaction. C32·C33 Flutamide offers its NH as a donor and its carbonyl and nitro groups as acceptors, but the carbonyl is partly engaged inside the molecule. C20·C35 Polymers that bring an acceptor therefore find a free partner in the NH, while polymers that bring a donor find none. C26·C34·C51 Nucleation follows different rules. C45

The causal chain as the authors state it (Discussion and Conclusions). The intramolecular contact in the glass is their inference ("highly likely"), not a measurement. C33·C35·C39·C51

What sets the outcome. The authors' closing generalization is that the study "highlights the importance of considering not only the molecular structure of the crystallizing compound but also the availability of functional groups to participate in intermolecular interactions", and "points to" drug–polymer interactions as important for reducing crystallization rates in amorphous formulations. C53 The comparison with acetaminophen makes the point concrete: there the amide carbonyl is free and the donor polymers PAA and PVPH are effective growth inhibitors; in flutamide the same group is engaged inside the molecule and the same polymers fail. R27·C34·C35

Scope. The conclusions rest on one drug, crystallizing as one polymorph in all but one sample, at 40–70 °C, which is 43–73 °C above its $T_g$, with additives at 10 mol% (simple polymers) or 5% w/w (pharmaceutical polymers). R6·R7.a·P1·P2 The growth result holds across both polymer series. The nucleation result is described by the authors themselves as qualitative. R3·R5·C6

Implications for formulation (note by the AI writer, not stated in the paper). Three practical points follow from the data. First, a structural drawing can mislead polymer selection: a carbonyl that looks available may be engaged intramolecularly, so a quick look at the crystal structure, the preferred conformer or the infrared spectrum of the glass can tell which groups a polymer can actually reach. Second, for a drug whose free group is a donor, acceptor-rich polymers are the logical first choice; here the 5% w/w ranking was PVP > PVPVA > HPMCAS > HPMC. R5 Third, growth inhibition and nucleation inhibition were separate properties in this study, and the only strong nucleation inhibitor was a close structural analogue, which suggests that a formulation relying on a polymer to stop crystallization should be tested for nucleation separately. R10·R11·R12

Limitations

Limitations the authors state. The nucleation behaviour was investigated "more qualitatively" than growth. C6 The factors that govern nucleation in supercooled liquids "are not well understood", and additives add complexity; most impurity studies concern inorganic compounds or crystallization from solution. C44 For PVAc, it "is impossible to determine" whether the stronger 1710 cm⁻¹ band reflects bonding to the polymer, disrupted drug–drug bonds, or both. C28 The growth–hydrogen bond correlation needed "additional compounds with more diverse chemistry"; this paper adds one drug. C3

The AI writer's own judgement (1): how far the correlation reaches. The growth data themselves are solid: triplicate rates at four temperatures, the same order at each, and the same pattern with the pharmaceutical polymers. R3·R5 The link to hydrogen bonding is carried by four carbonyl polymers (PVP, PDMA, PIPA, PVAc), with infrared measured at 1:1 rather than at the 9:1 ratio of the growth tests, and PAA's computed charge out of order. X8·X9 The explanation for the donor polymers rests on an intramolecular contact inferred from the crystal and a vacuum calculation; for PAA, crystallization on cooling and immiscibility at higher content offer a second route that the paper does not separate. C35·X12 The ranking is therefore well supported; the mechanism is plausible but indirect.

The AI writer's own judgement (2): nucleation and nilutamide. The nucleation verdicts come from two DSC runs per sample, a growth rate at $T_{max}$ obtained from the growth data by an unstated fit, an unnamed statistical test and one micrograph per sample. X4·X6 Within these data, PIPA, a strong growth inhibitor, did lower nucleation, so the summary that growth inhibitors had no similar effect on nucleation is stronger than Table 2. X5 The two nilutamide explanations, dilution plus $T_g$ for growth and incorporation into prenuclei (or, in the abstract, small size) for nucleation, are not shown or tested. X11·X7 Some printed details are also wrong or unclear: the PIPA weight percent (4.6% instead of about 4.4%), a PAA series in the Figure 3 caption that matches the 10 mol% data, and Figure 4 peak temperatures for PDMA and PVP (about 64 and 72 °C) that differ from the Table 2 averages (60.6 and 68.4 °C) by more than the stated standard deviation. X1·X2·X3 None of these changes the growth ranking.

Conclusion

For supercooled flutamide, the paper shows that polymers compared at equal molar ratio differ by about an order of magnitude in how much they slow crystal growth, that the best ones (PVP, PDMA, PIPA) are those whose carbonyl groups accept a hydrogen bond from the drug's NH, and that donor polymers fail, "most likely" because the drug's carbonyl is held in an intramolecular hydrogen bond. R3·R14·C39·C51 Changes in $T_g$ were too small to explain the ranking. R7·C33 The same polymers did not affect nucleation in a way that follows their hydrogen bonding, while the analogue nilutamide nearly stopped nucleation. C43·C45·R12 The study's lasting point is methodological as much as chemical: compare polymers per functional group, and check which of the drug's groups are actually free before expecting a polymer to bind them. C4·C53