When a PEG carrier crystallizes first: how it slows, ignores or speeds up drug crystallization

Background and motivation

Solid dispersions, a drug dispersed in an inert, water-soluble carrier, are a standard way to raise the dissolution rate of poorly water-soluble drugs (Chiou 1971, ref 1; Leuner 2000, ref 2; Serajuddin 1999, ref 3; Craig 2002, ref 4). Two kinds of carrier are common. Amorphous polymers such as polyvinylpyrrolidone (PVP) are used to inhibit drug crystallization and give an amorphous dispersion. Partly crystalline polymers such as polyethylene glycol (PEG) are different: in a PEG dispersion the drug can be amorphous, crystalline or semicrystalline, and the phase behaviour of the system can be very complex. C8

How both carrier and drug crystallize decides the microstructure of the dispersion, and with it dissolution and bioavailability. The literature had shown that the outcome varies: nifedipine was crystalline in a PEG dispersion (Lin 1996, ref 8), oxazepam amorphous (Ginés 1996, ref 9), and in lipid–PEG dispersions the lipid governed how PEG crystallized (Unga 2010, ref 10). C9 Most studies, however, described only the final state of the drug in PEG. A systematic study of crystallization kinetics in semicrystalline polymer dispersions was missing, although the mutual crystallization of drug and carrier sets the microstructure. The authors expected that PEG could either delay or promote drug crystallization, and that the drug might in turn change how PEG crystallizes. C10·C11

The paper builds on two earlier studies by the same group. In those, the crystallization tendency of the pure drug and its interaction with PEG decided where the drug was located in the PEG matrix (Zhu 2010, ref 11; Zhu 2011, ref 12). C12 Phase diagrams from that work and from Baird and Taylor (2011, ref 23) supplied the properties of the four drug–PEG pairs used here: estimated solubility in supercooled liquid PEG, phase diagram type and, from infrared spectra, whether the drug hydrogen-bonds with PEG. C20 From polymer science came one more expectation: a second component that interacts favourably with PEG can slow the growth of PEG spherulites (Talibuddin 1996, ref 13). C14

The hypothesis tested is stated directly: the crystallization tendency of the pure drug, combined with its ability to interact with the polymer, will affect how the drug crystallizes in the dispersion. C13 The authors also asked the reverse question, whether the drug changes PEG crystallization. To separate the two properties they chose four drugs: two that crystallize within seconds from the melt (haloperidol, HLP; benzocaine, BZC) and two that crystallize slowly and erratically (ibuprofen, IBP; fenofibrate, FNB), with one drug in each pair that hydrogen-bonds with PEG (BZC, IBP) and one that does not (HLP, FNB). R1·R2·R3·M1·M2·M3·M4 All dispersions were 20/80 w/w drug/PEG 3350, made by melting the two together. P1·A1

Background concepts

Induction time and stochastic nucleation. When a melt is cooled below its melting point, crystals do not appear at once. The time until the first crystals are detected is the induction time, here written $t_{\mathrm{i}\mathrm{n}\mathrm{d}}$. In Table 1 a negative $t_{\mathrm{i}\mathrm{n}\mathrm{d}}$ means crystallization had already begun before the sample reached the set temperature. For some drugs the induction time varies so much between identical samples that a single value is meaningless. Following an approach developed for solution crystallization (Jiang 2009, ref 14), the authors therefore followed 50 samples of each slow drug and recorded $P(t)$, the fraction of samples that had started to crystallize by time $t$. P3 The spread itself is read as the stochastic nature of nucleation. C16

Relative crystallinity. For PEG, the two crystalline diffraction peaks (at scattering vectors $q$ = 1.35 and 1.65 Å⁻¹, which correspond to 19.1° and 23.2° 2θ for Cu Kα radiation) were integrated, and their share of the total diffracted area gave the apparent crystallinity ${\Phi }_{\mathrm{p}\mathrm{e}\mathrm{g}}$. P4 The paper normalises it to its final value (Eq. 1):

In plain terms: $X(t)$ is the fraction of the final PEG crystallinity reached at time $t$, so every PEG curve runs from 0 to about 100% whatever the absolute crystallinity of the sample. E1 For the drugs the paper uses the intensity of Raman bands that appear on crystallization, again as a relative crystallinity; how the band intensities were normalised is not printed. P6

The Avrami equation. The PEG curves were fitted with the Avrami model (Eq. 2):

In plain terms: $z$ is a rate constant that depends on the crystallization conditions, and the exponent $n$ reflects the growth geometry. The paper's reading scheme, taken from ref 19, is that with all nuclei formed at one instant, $n$ = 1 means linear growth, $n$ = 2 platelike growth and $n$ = 3 polyhedral growth; if nuclei form at a constant rate during crystallization, each value is one higher. E2

Eutectic and monotectic systems, melting point depression and supercooling. When a drug mixes with molten PEG, each component can lower the melting point of the other. In a eutectic system the liquidus falls from both sides to a minimum, the eutectic; a mixture at that composition melts at the eutectic temperature, well below either pure component. In a monotectic system, which HLP forms with PEG, the drug's melting point is barely lowered. C40 Melting point depression matters for crystallization because the driving force depends on supercooling, the distance between the melting point and the temperature at which the sample is held. A drug whose effective melting point in PEG is lowered from about 89 °C to about 48 °C, as for BZC, is much less supercooled at 25 °C than the pure drug. M3·C32

Heterogeneous nucleation. Nuclei that form on a foreign surface, such as a crystal of another substance, need less energy than nuclei that form in the bulk liquid. Polymers had been shown to act as such heteronuclei for organic molecules, even selecting the polymorph, but in solution rather than melt crystallization (Lang 2002, ref 27). C35 The authors note that most studies of drugs in polymers report the opposite effect: the polymer retards crystallization. C36

Spherulites. PEG crystallizes as spherulites, spherical aggregates of crystalline lamellae that grow outward from a nucleus. Under the polarized microscope the front of a spherulite can be followed, and its speed is the linear growth rate. P8

Methods in detail

The study has no model; it is a set of crystallization experiments arranged so that the drug and PEG in the same dispersion can be followed on their own time scales. PEG crystallizes in seconds, the slow drugs over hours, so different instruments were needed for each. The flow below shows the stages.

Contribution 1: crystallization of the pure components

The problem. To say that PEG changes a drug's crystallization, one needs the behaviour of the pure drug and of pure PEG under the same conditions. For two of the drugs this is easy (they crystallize at once); for the other two it is not, because the induction time is erratic.

What they did. The pure drugs were melted on a hot plate on a glass coverslip, then held at 25 °C in a sealed Linkam THMS 600 hot stage and watched under a Nikon Eclipse E600 polarized light microscope. P2 For IBP and FNB, 50 samples each were followed and $P(t)$ recorded for 3 weeks (504 h). P3·R2·R3 Pure PEG 3350 was studied by in situ synchrotron X-ray diffraction at beam station 5-ID-D of the Advanced Photon Source (17 keV, λ = 0.73 Å, sample-to-detector distance 228 mm, q range 0.6–4.6 Å⁻¹, calibrated with lanthanum hexaboride). The sample sat in an aluminum DSC pan in a sealed Linkam DSC600 stage, was melted until all diffraction peaks had disappeared and was then quench cooled to 25 or 40 °C. P4 The two temperatures were chosen to represent room-temperature storage and a temperature between room temperature and the melting point of PEG. C15

Why it works. The synchrotron gives patterns fast enough to follow a crystallization that is over in under 2 min; the statistical approach turns an erratic induction time into a measurable probability.

Contribution 2: following drug and PEG in the same dispersion

The problem. In a 20/80 dispersion, the drug signal is weak and the drug crystallizes over hours, while PEG crystallizes in seconds.

What they did. For PEG, physical mixtures of drug and PEG were melted above the drug's melting point in the synchrotron stage and quench cooled to 25 or 40 °C, as for pure PEG. P4 For the drug, 20/80 physical mixtures were comelted in an aluminum holder for a laboratory diffractometer (Shimadzu XRD6000, Cu Kα, 40 kV, 30 mA, 2θ 4–30°, 0.04° steps at 4° per min), cooled, stored in a desiccator over phosphorus pentoxide and measured at intervals. The time needed for complete melting was set by hot-stage microscopy. P5 Because the BZC peak in XRD was too weak for kinetics, Raman spectroscopy was added (Kaiser RamanRXN2 with a PhAT-785 probe, 400 mW, 1890–150 cm⁻¹), with samples prepared and stored as for laboratory XRD. P6·C22 The marker signals were: BZC, the diffraction peak at 20.1° 2θ and Raman bands at 1573 and 1680 cm⁻¹; IBP, 20.2° 2θ and 746 and 784 cm⁻¹; FNB, 16.2° and 16.8° 2θ and 1419 and 1648 cm⁻¹. P5·P6 For the Raman kinetics of BZC, time zero is the moment the sample reached 25 °C. P6 Moisture was excluded by sealing the samples in the analysis chamber or storing them over desiccant. P1

How it is summarised. For each dispersion the authors report two numbers for the drug and two for PEG: $t_{\mathrm{i}\mathrm{n}\mathrm{d}}$, and $t_{50}$, the time from onset to 50% relative crystallinity (the induction time not included). R23

How it differs from earlier work. The group's earlier studies and most of the literature described the end state of the dispersion; here both components are followed in time, on the time scale each one needs. C10·C12

Contribution 3: seeding and spherulite growth

Seeding. To test whether crystalline PEG nucleates the slow drugs, a small piece of crystalline PEG was placed in supercooled liquid IBP or FNB at 25 °C and watched under the polarized microscope. P7 This test uses the pure supercooled drug, not the comelted dispersion.

Spherulite growth. To see how each drug affects PEG crystal growth, the front of a PEG spherulite was followed at 40 °C in pure PEG and in each 20/80 dispersion; at 25 °C PEG grows too fast for good data. P8·C30

Supporting information. For FNB, the one pair not covered by earlier phase diagrams, the SI gives the FNB/PEG phase diagram, built from the temperature at which the last crystal melted on heating at 1 °C/min under the microscope and from the DSC offset melting temperature at 1 °C/min. P9·P10 It also compares the ATR-FTIR spectrum of molten FNB/PEG 20/80 with a spectrum calculated from the pure components (reference bands 1655 cm⁻¹ for FNB and 1094 cm⁻¹ for PEG): if the two match, the components do not interact. P11

Experimental analysis

The pure drugs: two fast, two slow and erratic

Pure HLP and BZC crystallized before the samples even reached 25 °C and were completely crystalline within a few seconds, too fast to follow. R1 Pure IBP and FNB behaved very differently (Figure 1). Of 50 IBP samples, 2 (0.04) had started to crystallize within 1 h and 10 (0.2) after 3 weeks; $P(t)$ rose in steps, to about 0.12 after about 72 h, about 0.14 after about 168 h and 0.20 after about 336 h. R2 FNB rose faster at first, from 4 of 50 (0.08) within 1 h to about 0.20 by about 8–14 h and about 0.34 by about 96 h, and then stayed at 17 of 50 (0.34) to 3 weeks. R3 The authors conclude that the chance of crystallization starting is low over hours to days and grows over weeks, and that pure IBP either starts within a few hours (the less likely case) or stays amorphous for many days. C17

Pure PEG: crystallized within 2 minutes

Pure PEG crystallized in less than 2 min at both temperatures, more slowly at 40 °C (Figure 2). At 25 °C the relative crystallinity passed 50% between about 22 and 24 s and reached about 96% at 40 s; at 40 °C it stayed near zero for about 12 s, reached about 51% at about 64 s and about 89% at 90 s. R4·C18 The induction time was −10 s at 25 °C, so PEG began to crystallize before the set temperature was reached, and 16 s at 40 °C. R5 The fitted Avrami exponent was below 1 at both temperatures; the authors take this to mean that growth "tended to be one-dimensional", as reported for very fast crystal growth (refs 20–22). R6·C19 Card check: on the paper's own reading scheme, $n$ = 1 is already the lowest value (linear growth with instantaneous nucleation), so an exponent below 1 lies outside it; the fitted $n$ and $z$ are not printed (story note X9).

Card check: the pure-PEG curve at 40 °C is not the same in all figures. In Figure 2 it reaches about 51% at about 64 s; in Figures 11 and 15, with the same legend, it reaches about 51% at about 48 s and about 89% at about 67 s. The offset of about 15 s is close to the 16 s induction time, although Figure 2 is said to start at the onset of crystallization (story note X1). Comparisons below therefore use the pure-PEG curve printed in the same figure as the dispersion.

Haloperidol: PEG has no effect

When HLP/PEG 20/80 was cooled to 25 °C, HLP crystallized first and completely within a few seconds, and PEG began to crystallize immediately afterwards (synchrotron patterns not shown). R7 The authors conclude that PEG did not affect HLP crystallization. C2 Table 1 lists $t_{\mathrm{i}\mathrm{n}\mathrm{d}}$ = −100 s and $t_{50}$ = 7 s for HLP. R23 Card check: the text says the HLP kinetics could not be captured by XRD, and the method behind the two Table 1 values is not stated (story note X7). HLP has negligible solubility in PEG (less than 1% w/w in the text, 0 in Table 1) and forms a monotectic system with it. M1·C40·X10

PEG crystallized as fast in HLP/PEG as alone: the authors call the kinetics "essentially the same" (Figure 3). C21 The curve with HLP in fact reaches 50% at about 15 s (Table 1: $t_{50}$ 15 s) against about 23 s for pure PEG, and the two meet at about 30 s. R8·R23 Card check: the early stage is faster with HLP; only completion time and final crystallinity are the same (story note X3).

Benzocaine: delayed and slowed

With 20% BZC in PEG, BZC first stayed amorphous after cooling to 25 °C and began to crystallize only after PEG had completely crystallized. R9 Its diffraction peak at 20.1° 2θ was absent at 30 min and present at 180 and 300 min (Figure 4), giving an induction period of about 2–3 h. R9 In the Raman spectra (Figure 5) a new band at 1680 cm⁻¹ appeared at 180 min and grew rapidly up to 300 min, together with the band at 1573 cm⁻¹; both changes matched the spectrum of crystalline BZC. R10·C22 The two bands gave the same kinetic curve, which the authors call "extremely consistent" (Figure 6): about 20% at about 200 min, 50% at about 290 min, about 92% at about 530 min and about 97–98% at 1000 min. R11·C23 The text gives about 10 h to reach 90%. R11 Card check: on the plotted curve 90% is reached at about 500 min, about 8 h (story note X12). Table 1 gives $t_{\mathrm{i}\mathrm{n}\mathrm{d}}$ = 120–180 min and $t_{50}$ = 110–170 min after onset. R23 Against a pure drug that crystallizes within seconds, the authors call the drop in rate substantial. C1

PEG in turn was hardly affected at first. In BZC/PEG at 25 °C, PEG followed the pure-PEG curve up to about 58% at about 26 s, then levelled off lower: about 83% at 40 s and about 94% at 120 s, against about 96% and about 99% for pure PEG (Figure 7). R12 The authors describe this as no significant difference in the early stage and a reduced rate in the later stage. C24

Ibuprofen: crystallization brought forward

In IBP/PEG 20/80, the IBP diffraction peak at 20.2° 2θ was absent at 20 min and present from 120 min (Figure 8); the text states that IBP started to crystallize "consistently within 2 h" and was completely crystallized within 4 h. R13·C25 Card check: the text attributes this to synchrotron XRD, but Figure 8 is plotted on the 2θ scale of Cu Kα radiation (PEG peaks at about 19.1° and 23.3°), and the captions of Figures 4, 8 and 12 do not name the instrument (story note X8). New Raman bands at 746 and 784 cm⁻¹ gave an induction time of about 100 min and more than 90% crystallinity within 4 h (Figure 9): about 57–61% at 180 min and about 91–94% at 240 min. R14·C25 Table 1 gives $t_{\mathrm{i}\mathrm{n}\mathrm{d}}$ ≈ 100 min and $t_{50}$ ≈ 70 min. R23 Compared with pure IBP, of which only 10 of 50 samples had started after 3 weeks, the onset in PEG is earlier and far more reproducible. R2·C3

The authors' first explanation was that crystalline PEG supplies heterogeneous nucleation sites ("was thought due to"). To test it they placed a piece of crystalline PEG in supercooled liquid IBP at 25 °C. After a few minutes an IBP crystal had formed at the interface between the molten IBP and the PEG, and no nuclei were visible in the bulk liquid (Figure 10, which is also the graphical abstract). R15·C4

PEG crystallization in IBP/PEG depended on temperature (Figure 11). At 25 °C it was close to pure PEG: about 50% at about 27 s (Table 1: $t_{50}$ 26 s, $t_{\mathrm{i}\mathrm{n}\mathrm{d}}$ 8 s) and about 93% at 80 s. At 40 °C it was much slower: PEG started after 40 s, reached about 47% at about 112 s ($t_{50}$ 114 s) and only about 62% at the last point, about 132 s, whereas pure PEG in the same figure passed 50% at about 48 s. R16·R23 The authors read the similarity at 25 °C as "probably" due to the very fast crystallization of PEG at that temperature, and note that PEG crystallization was in any case far faster than IBP crystallization. C26·C27

Fenofibrate: brought forward in the same way

In FNB/PEG 20/80, PEG again crystallized before the drug, and FNB started earlier than as pure drug: on average after about 2–3 h, when diffraction peaks appeared at 16.2° and 16.8° 2θ (Figure 12). R17·C28 Raman bands at 1648 and 1419 cm⁻¹ gave the same curve (Figure 13): about 8% at about 180 min, about 54% at about 330 min and about 92% at about 510 min. R18 The text gives about 9 h to 90%, with an induction time similar to XRD. R18·C28 Card check: the plotted curve reaches 90% at about 495 min, a little over 8 h (story note X12). Table 1 gives $t_{\mathrm{i}\mathrm{n}\mathrm{d}}$ = 120–180 min and $t_{50}$ = 140–200 min. R23 A PEG particle placed in supercooled FNB at 25 °C nucleated FNB immediately next to it within a few minutes, with no crystallites elsewhere in the liquid (Figure 14). R19

PEG in FNB/PEG behaved much like pure PEG at both temperatures (Figure 15): $t_{\mathrm{i}\mathrm{n}\mathrm{d}}$ 1 s and $t_{50}$ 21 s at 25 °C, $t_{\mathrm{i}\mathrm{n}\mathrm{d}}$ 7 s and $t_{50}$ 52 s at 40 °C, the curve passing about 50% at about 52 s against about 48 s for pure PEG in the same figure. R20·R23·C29

The SI explains why FNB was chosen as a non-interacting drug. Its phase diagram with PEG is eutectic: the liquidus falls from about 82 °C for pure FNB to a minimum of about 58 °C at 15–20 wt% FNB, so the 20/80 dispersion lies at the eutectic composition (Figure S1). R24·C48 The infrared spectrum of the molten 20/80 mixture has the same bands as the spectrum calculated from the pure components, which the authors take to mean no FNB–PEG interaction (Figure S2). R25·C49 The main text gives the FNB melting point depression as approximately 20 °C. M2 Card check: Figure S1 gives about 24 °C between the pure-FNB liquidus and the minimum (story note X11).

The summary table

Table A. Drug crystallization in 20/80 w/w drug/PEG 3350 at 25 °C, from Table 1 of the paper, with the pure-drug behaviour from the text and Figure 1. Solubilities are estimates extrapolated to 25 °C with the Van't Hoff equation from earlier phase diagrams; hydrogen bonding is from FTIR in earlier work (FNB also in Figure S2). R1·R2·R3·R23·M1·M2·M3·M4

Table B. PEG crystallization alone and with each drug: dispersion values from Table 1, pure-PEG induction times from the text, pure-PEG times to 50% read from Figures 2, 11 and 15 (approximate), growth rates from the text. R4·R5·R16·R20·R21·R22·R23

PEG spherulite growth

At 40 °C the front of a pure PEG spherulite advanced across the microscope field within 10 s (Figure 16); the text gives a linear growth rate of 10 µm/s. R21 With HLP or FNB the rate was about 8 µm/s, "slightly slower"; with IBP and BZC it was 3 and 1 µm/s, "much slower" (Figure 17). R22·C31 Card check: in Figure 17 the pure-PEG line reaches about 470 µm at about 24 s, a slope of about 20 µm/s, while the lines of the four dispersions agree with the text values (about 175 µm at about 22 s for HLP, about 420 µm at about 54 s for FNB, about 215 µm at 80 s for IBP, about 85 µm at 80 s for BZC). Whether the text or the figure is wrong cannot be decided from the paper (story note X2).

Discussion

The authors' framework: crystallization tendency and interaction

Returning to their hypothesis, the authors conclude that both the crystallization tendency of the drug and its ability to interact with PEG influence crystallization in the dispersion, "with additional complicating factors" where PEG promotes heterogeneous nucleation ("the results of this study suggest"). C38 Their chain of reasoning, as given in the discussion, is:

For the fast crystallizers the rule is stated plainly: if the drug crystallizes fast relative to PEG and has no favourable interaction with it, PEG has a negligible effect; if it crystallizes fast but interacts favourably, the onset is delayed and the rate reduced. C39 For the slow crystallizers, the relationship between the speed of the pure drug and its speed in PEG "was found to break down", because of the "apparent promotion" of heterogeneous nucleation by PEG. C41 All systems except HLP/PEG, the only monotectic one, are eutectic, with 20/80 close to the eutectic composition. C40

Why benzocaine is slowed

The decrease in BZC crystallization rate "is thought to be caused by" entropic mixing and favourable BZC–PEG interactions, which depress the BZC melting point. At 20:80 the composition lies below the eutectic composition, so the relevant melting point is that of the eutectic, reported as around 48 °C (ref 12), against around 89 °C for pure BZC. The supercooling at 25 °C is therefore very different for the two, and pure BZC has a much larger thermodynamic driving force. C32 Since the Tg of pure BZC is −31 °C (Baird 2010, ref 26), far below 25 °C, the authors consider it "reasonable" that the difference is thermodynamic rather than kinetic. C33 Neither the supercooling of BZC in the solidified dispersion nor its mobility was measured in this paper. C32·C33

Why ibuprofen and fenofibrate are faster

Both form eutectics with PEG, IBP interacting more strongly (larger IR peak shift, ref 12 and Figure S2), with melting point depressions of about 30 °C (IBP, ref 12) and approximately 20 °C (FNB, Figure S1). The driving force is therefore reduced for both, yet the induction time was "consistently reduced", suggesting to the authors that both drugs "nucleate more readily in the dispersion". C34 The seeding tests point to the cause: PEG "appears to be able to promote heterogeneous nucleation", acting as a nucleation surface for both drugs. C4 The authors find this "extremely interesting" because most studies of drugs in polymers find that the polymer retards crystallization, and because earlier demonstrations of polymer heteronuclei concerned solution crystallization (ref 27). C35·C36

Why PEG is slowed

PEG crystallization varied more between the dispersions at 40 °C than at 25 °C, where the authors see negligible differences. C43 At 40 °C, PEG crystallized more slowly in IBP/PEG than alone, but almost as fast in FNB/PEG; the difference is "probably due to" the different ability of the two drugs to hydrogen-bond with PEG (ref 13): IBP can hydrogen-bond with PEG (ref 12), while FNB has no donor groups and PEG has mainly acceptor groups (Figure S2). C44 The spherulite growth rates, slowest with BZC and IBP, which both hydrogen-bond with PEG, are taken as further confirmation. C45 From a practical standpoint, PEG still crystallizes within a few minutes in all systems, and how a changed PEG crystallization rate affects the microstructure of the dispersion "is not known". C46

Card check: the "negligible difference" at 25 °C covers visible differences: PEG with BZC levels off at about 94% against about 99% at 120 s, PEG with HLP starts faster and PEG with IBP lags slightly (story notes X3, X4).

What it suggests for formulation

The authors draw two practical points. First, the drug's crystallization kinetics in the formulation matter for bioavailability: for consistent delivery the drug "should either be fully amorphous or fully crystalline" when it is administered, not a mixture of forms. C37 Second, because nucleation rates strongly influence particle size distributions (refs 28–30), the promotion of nucleation by PEG "could be an important factor" for the dissolution performance of solid dispersions. C42 The abstract concludes that any effect of PEG on crystallization "should be investigated" to obtain a dispersion with consistent properties. C7

Note by the AI writer (not discussed in the paper). In these PEG melt dispersions a drug that is slow to crystallize on its own, and so might be expected to stay amorphous, crystallized within a few hours at 25 °C once PEG had crystallized. For BZC, IBP and FNB the solid form kept changing over roughly the first 10 h after solidification, so a sample characterized or tested within that window may hold a mixture of amorphous and crystalline drug, the situation the authors warn against. R11·R14·R18

Limitations

Limits the authors state

• PEG spherulite growth could be measured only at 40 °C, because PEG grows too fast at 25 °C. C30
• The HLP kinetics were too fast to be captured by XRD. R7
• The BZC peaks in XRD were too weak for a kinetic study, so Raman was used for BZC. C22
• How changing the PEG crystallization rate influences the dispersion microstructure "is not known". C46
• The earlier demonstrations of polymer heteronuclei concerned solution, not melt, crystallization. C35

The AI writer's own judgement

• Where the drug nucleates in the dispersion is not shown. The seeding tests used pure supercooled drug. In the dispersions PEG crystallizes first, which leaves the drug concentrated in the remaining non-crystalline fraction; by Table 1 the estimated solubilities of BZC (28%) and IBP (25%) in supercooled liquid PEG exceed the 20% loading, yet both crystallized. Enrichment of the residual melt could shorten the induction time as well as nucleation on PEG crystals, and the paper does not separate the two. X13·M3·M4
• Pure-drug statistics are censored. $P(t)$ was followed for 3 weeks, when 40 of 50 IBP and 33 of 50 FNB samples had not crystallized; "hours to months" for the pure drugs is therefore not measured. The pure-drug numbers come from 50 microscope samples, the dispersion numbers from XRD and Raman runs whose number is not stated. X5·X6
• The explanations rest on cited values. The BZC melting points and Tg and the hydrogen bonding of BZC and IBP with PEG come from earlier papers; bulk PEG kinetics at 40 °C were measured with two drugs only. C32·C33·X14
• Reference values are inconsistent. The pure-PEG curve at 40 °C differs by about 15 s between Figure 2 and Figures 11 and 15, and the pure-PEG spherulite rate is 10 µm/s in the text but about 20 µm/s in Figure 17. The HLP values in Table 1 have no stated method, and the instrument for Figures 4, 8 and 12 is unclear. X1·X2·X7·X8
• Scope. One composition (20/80), one PEG grade (3350), one temperature for the drug kinetics (25 °C), one or two drugs per behaviour and no dissolution data. A1·A3·X6

Overall, the AI writer rates the evidence as moderate: the onset times and curves are measured directly by two methods that agree, and the seeding images are direct, but the mechanisms are inferred or cited. C23·R15·R19

Conclusion

In 20/80 w/w drug/PEG 3350 comelts held at 25 °C, the same carrier produced three different outcomes. Benzocaine, which crystallizes within seconds as a pure melt, started only after 120–180 min and needed hours more to finish; haloperidol still crystallized within seconds, before PEG; ibuprofen and fenofibrate, which mostly do not crystallize within 3 weeks as pure melts, started after about 100 and 120–180 min. R1·R2·R3·R23 The authors explain the delay by melting point depression lowering the driving force ("is thought to be caused by") and the earlier onset by crystalline PEG acting as a heterogeneous nucleation surface ("appears"), supported by seeding tests in pure supercooled drug. C32·C4·R15·R19 PEG itself always crystallized within a few minutes, but more slowly with the drugs that hydrogen-bond with it, most clearly for ibuprofen at 40 °C ($t_{50}$ 114 s against 52 s with fenofibrate) and in spherulite growth (1 and 3 µm/s with benzocaine and ibuprofen). R22·R23·C44·C46 The authors' message is that PEG "can delay, promote or have no influence" on drug crystallization, so the effect must be checked drug by drug. C6·C7