How dissolved polymers slow the conversion of anhydrous drugs to their hydrates in water

Background and motivation

Many drugs exist both as an anhydrous crystal and as a crystal hydrate. The solid form can change pharmacokinetics, bioavailability and chemical stability, and forms can change during manufacturing steps such as milling and wet granulation (Haleblian 1969; Brittain 1999; Byrn 1995) C1. When the starting material is the anhydrate, conversion to the hydrate during any unit operation that uses water is "not uncommon" (Morris 2001) C1. Drying may reverse the hydrate, but dehydration can leave amorphous content or crystal defects that harm physical and chemical stability C1. A formulator who wet-granulates an anhydrous drug therefore has a reason to want the hydrate not to form at all.

Conversion of an anhydrate to a hydrate in water does not happen inside the solid. It runs through the solution as a solvent-mediated transformation (SMT) in three stages: the anhydrate dissolves, the hydrate nucleates, and the hydrate crystals grow (Cardew and Davey 1985) C2. If one of these stages can be slowed, the whole conversion can be slowed or stopped, which would let a manufacturer keep control of the solid form C2. One way to do this is to add something that interferes with nucleation or growth. Additives from proteins to salts can do it, but common excipients would be the practical choice C3.

Several earlier studies had shown that polymers can act this way. Simonelli and co-workers found in 1970 that poly(vinylpyrrolidone) (PVP) reduced or completely stopped the growth of sulfathiazole single crystals in solution C3. Excipients including binders and surfactants either sped up or slowed the conversion of anhydrous carbamazepine to its dihydrate (Katzhendler 1998; Otsuka 2000; Rodriguez-Hornedo 2004; Tian 2007; Qu 2007) C3. Excipients also changed the formation of theophylline monohydrate in slurries and wet granulations, including in earlier work from the same group (Wikström; Airaksinen 2003; Jorgensen 2004) C3. In transdermal delivery, hydroxypropyl methylcellulose (HPMC) stabilised supersaturated solutions of hydrocortisone and ibuprofen (Iervolino 2000; Raghavan 2001) C3. What these studies left open, the authors write, is the mechanism: it "is often unclear" whether a polymer acts on nucleation, on growth or on both C24.

To answer that, kinetic data in water are needed, and Raman spectroscopy is well suited to collect them. Different crystal forms have distinct Raman spectra, a fiber-optic probe can sit in the slurry so that no sample has to be removed (unlike off-line X-ray powder diffraction), and water scatters Raman light weakly, so it barely disturbs the spectra C4. The group had already used in-line Raman to follow hydrate formation in slurries and wet granulation (Wikström 2005; Wikström thesis 2007) C4.

The paper therefore follows the anhydrate-to-hydrate conversion of three model drugs, caffeine (CAF), carbamazepine (CBZ) and sulfaguanidine (SGN), in aqueous slurries containing each of seven polymeric excipients, by in-line Raman. It then adds two separate tests for CAF and CBZ, the growth rate of seeded hydrate crystals and the induction time for nucleation from supersaturated solution, to work out which stage an inhibiting polymer acts on A1·A2·A3.

Background concepts

Solvent-mediated transformation. The anhydrate of these drugs is more soluble than the hydrate: 49.7 versus 21.8 mg/mL for CAF, 0.424 versus 0.139 mg/mL for CBZ and 1.38 versus 1.07 mg/mL for SGN, values the paper cites from a PhD thesis from the same group M1·M3·M6. When the anhydrate is put in water, it dissolves until the solution is supersaturated with respect to the hydrate. The hydrate then nucleates, and its growth lowers the drug concentration in solution, which lets more anhydrate dissolve; this goes on until all the anhydrate has gone and the solution sits at hydrate solubility C2. The overall rate depends on all three stages, so an additive can only change the kinetics by changing at least one of them C2. An earlier study of the same three drugs without additives concluded that dissolution was not the rate-limiting step for any of them, and the authors do "not anticipate" that their low polymer concentrations change dissolution, so in their reading a polymer must act on nucleation or on growth R15·C25.

Nucleation, induction time and growth. Nucleation is the birth of new crystals from a supersaturated solution; the induction time is the delay before such crystals become detectable. A polymer that slows nucleation lengthens the induction time. Growth is the enlargement of crystals that already exist; it can be measured on its own by placing seed crystals in a supersaturated solution and following their size. In a slurry the two are mixed: the first appearance of hydrate needs nucleation, but whether enough hydrate forms to be detected also depends on how fast the nuclei grow. Telling the two apart is the central design problem of the paper.

Reading a transformation profile. The Raman-derived percent hydrate plotted against time is usually sigmoidal: it stays near zero while the anhydrate dissolves and the hydrate nucleates, then rises steeply in a growth phase that combines dissolution, further nucleation and crystal growth, and finally levels off when conversion is complete C9. The authors summarise each profile by two times P4. The onset (delay) time runs from water addition to the first rise in hydrate; it is found by extrapolating the steep part of the profile back to the baseline. The growth time runs from that onset to 80% hydrate. The total transformation time is their sum, a relation the paper states in words:

In plain terms: the time the slurry needs to reach 80% hydrate is the waiting time before any hydrate appears plus the time the visible conversion then takes E1. A polymer that lengthens the onset is read as delaying nucleation ("appeared"), and one that lengthens the growth time as slowing crystal growth ("most likely") C10·C11.

Quantifying hydrate by Raman. Calibration samples were powder blends of anhydrate and hydrate in 10% (mol/mol) steps plus 5, 15, 85 and 95% P3. For CAF and SGN the authors used a bivariate calibration, which compares one band characteristic of the anhydrate with one characteristic of the hydrate (1655 and 1699 cm⁻¹ for CAF; 828 and 819 cm⁻¹ for SGN) R1·E2. The CBZ bands overlap too closely for that, so they used partial least squares (PLS) regression, a multivariate method that finds the combinations of intensities over a spectral region (1015–1050 cm⁻¹) that best predict the composition, after standard normal variate (SNV) scaling and mean centring R1·C8·E3·E4. Figure 1 shows the spectra of the pure forms R2.

Crystal habit. Habit is the outer shape of a crystal and is set by which faces grow fastest. CAF 4/5 hydrate and CBZ dihydrate grow as needles in pure water, fastest along the long axis R10·R11. In the authors' reasoning, a polymer that adsorbs selectively on the fast-growing face slows that face and changes the shape, whereas adsorption that is not face-specific slows growth without changing it C36·C37.

Methods in detail

The study has one main experiment and two follow-up tests. The slurry experiment, run for all three drugs, shows whether a polymer delays or slows hydrate formation. The seeded-growth test and the induction-time test, run for CAF and CBZ only, isolate growth and nucleation. The flow below shows how they fit together.

Materials

Anhydrous CAF and the SGN monohydrate came from Sigma-Aldrich and anhydrous CBZ from Hawkins, Inc. M1·M3·M5. Anhydrous SGN was made by heating the monohydrate at 105 °C for at least 12 h and was kept below 5% RH M6. For calibration, CAF 4/5 hydrate was made by storing the anhydrate at about 100% RH for at least two weeks, and CBZ dihydrate by recrystallising the anhydrate from hot water M2·M4. Table 1 assigns the forms: CAF anhydrate is the triclinic β-phase and the hydrate a monoclinic 4/5 hydrate; CBZ anhydrate is the monoclinic low-temperature form and the hydrate an orthorhombic dihydrate; SGN anhydrate is monoclinic form II and the hydrate a monoclinic monohydrate M1·M3·M6. Each drug came from one source throughout, to keep the particle size distribution the same C5. The seven polymers were HPMC E5 (Methocel), hydroxypropyl cellulose (HPC, Klucel LF), sodium carboxymethylcellulose (CMC, 12M31P), cross-linked poly(acrylic acid) (PAA, Carbopol 934P), PVP (Kollidon 30), poly(vinyl alcohol) (PVA, average molecular weight 47000) and poly(ethylene glycol) (PEG, average molecular weight 1000), dissolved or dispersed in double-distilled water M7–M14.

Slurry experiments with in-line Raman

Problem. The kinetics of conversion must be followed in the slurry itself, without removing samples that might convert or dry while being measured.

What they did. The slurries contained 2.5 g CAF in 50 mL of a 2.0 mg/mL polymer solution, 0.5 g CBZ in 50 mL of a 0.04 mg/mL solution, or 1.0 g SGN in 40 mL of a 5 mg/mL solution P1. Drug amounts followed the drugs' solubilities, and the polymer concentrations were set from preliminary experiments that showed drug-dependent effects C5. An overhead stirrer ran at about 300 rpm, and each slurry was run in triplicate; the paper does not state the slurry temperature P1·X10. A 785 nm Raman spectrometer with a sapphire-tipped immersion probe, angled toward the moving liquid about 5 mm above the impeller and delivering about 80 mW to the sample, took a spectrum every 30 s for CAF and CBZ (four 5 s accumulations) and every 5 or 10 s for SGN (one accumulation), because SGN converts so fast P2.

Calibration. Three spectra were taken of each of three replicate blends at each composition, with the probe moved between spectra to limit sub-sampling P3. The first CAF calibration predicted poorly: CAF is so soluble that the dissolved drug in the slurry broadened and shifted the bands compared with the dry blends C7. The authors fixed this by adding a weighted spectrum of saturated CAF solution to each calibration spectrum C7. The final models gave R² of 0.981 (CAF), 0.982 (CBZ) and 0.986 (SGN), and root mean square errors of calibration (RMSEC) of 5.6, 4.6 and 5.2% R1.

Seeded growth rates

Problem. A longer growth time in the slurry suggests slower crystal growth, but the slurry growth phase also contains dissolution and further nucleation C9. Growth has to be measured on its own.

What they did. Hydrate seed crystals of CAF or CBZ were placed in the well of a microscope slide with about 350 µL of supersaturated solution, 25 mg/mL CAF or 0.5 mg/mL CBZ, containing 2.0 or 0.04 mg/mL polymer (the slurry levels) P5. These concentrations were chosen because they gave reasonable growth rates without polymer C6. Under a cover slip, the crystals were photographed at regular intervals; length was plotted against time and the initial linear slope gave the growth rate of each crystal, averaged over at least five crystals P5. Where a polymer stopped growth along the needle (HPMC with CBZ), the growth in width was measured instead P5·R11. For CAF with PAA and PVA, a further test at 75 mg/mL checked what happens at higher supersaturation A4.

Induction times by light scattering

Problem. Nucleation needs to be measured separately from growth.

What they did. Supersaturated solutions of 35 mg/mL CAF or 0.35 mg/mL CBZ, with the same polymer levels, were placed in a 500 mL jacketed vessel and cooled from 60 to 25 °C over 40 min, then held at 25 °C P6. Initial tests made sure that nucleation did not begin before 25 °C P6. A focused beam reflectance measurement (FBRM) probe next to the stirrer counted particles larger than 1 µm; the appearance of particles was taken as the onset of nucleation, and the induction time came from extrapolating the signal-versus-time curve back to baseline P6·C6. The paper does not state how many runs went into each mean X10.

How the tests combine. If a polymer lengthens the slurry onset and also the induction time, it delays nucleation. If it lengthens the slurry times and slows seeded growth but leaves the induction time unchanged, it acts on growth. The comparison is the authors' main tool for assigning a mechanism C18·C20·C30.

Experimental analysis

Baseline: conversion in water alone

Without polymer, all three drugs converted quickly (Figure 2). The mean onset was 7.1 ± 2 min for CAF and 3 ± 0.6 min for CBZ, and the growth times were 2.6 ± 1 and 2.9 ± 0.5 min, giving totals of 9.8 and 5.9 min R4.a·R4.i. The single profiles in Figure 2 show onsets of about 8 and 4 min and growth of about 2.5 min R3. SGN behaved differently: hydrate was already present in the first spectrum, taken 5 s after water was added, so its onset was reported as zero and its growth time was 0.1 ± 0.1 min (the text gives about 5–10 s) R4.q·R8·R3. The CAF profile in Figure 2 rises above 100% hydrate, to about 117%, and sits at roughly 1–18% before conversion; the percent scale for CAF is therefore approximate, though the onset and growth times come from the shape of the curve R3·X12.

Table A collects the slurry results for every drug and polymer.

Table A. Slurry transformation times in minutes, mean ± SD (n = 3), from Table 4; SGN onset was zero for every polymer. "—": no hydrate formed during the run. Standard deviations of the totals are omitted. R4·X1·X2

Caffeine: PAA stops the conversion, PVA delays it by hours

Two polymers stood out for CAF (Figure 3). With PAA no hydrate formed during the run: more than 1150 min in Table 4, "more than 19 h" in the discussion, against 9.8 min in water R4.e·R14·R4.a. PVA delayed the onset to 212 ± 62 min and stretched the growth time to 16 ± 4 min, a total of 228 min R4.g. The authors call both effects "strong inhibition" C13. The three cellulose derivatives "consistently increased" the onset: HPC to 15.6 ± 2.3 min, CMC to 13.1 ± 1.7 min and HPMC to a value printed as 1.83 min R4.b–d·C13. That printed HPMC value cannot be right: its growth time is 3.4 min and its total 21.6 min, which puts the onset at about 18 min, and the HPMC onset bar in Figure 3b stands at about 18 min; the decimal point is most likely misplaced X1·R5. PVP (10.3 ± 1.8 min) and PEG (5.2 ± 0.8 min) did not change the onset within experimental error, and none of these five polymers changed the growth time, which stayed near 3 min R4.f·R4.h·C13. In the single profiles of Figure 3a, PVA converted between about 155 and 185 min and PAA stayed flat for the 250 min shown R5.

Carbamazepine: HPMC stops the conversion; HPC, PVP and PVA slow it

For CBZ the most effective polymer was HPMC, which kept the slurry free of hydrate until the runs were stopped (more than 1500 min in Table 4; "up to 24 h" in the text) R4.j·R14·C14. Three more polymers inhibited: HPC raised the onset from 3 to 56 ± 38 min and the growth time from 2.9 to 39 ± 25 min; PVP to 40 ± 20 and 39 ± 19 min; PVA to 23 ± 6 and 6.3 ± 1.7 min R4.i·R4.k·R4.n·R4.o. CMC, PAA and PEG showed "little to no inhibition"; PAA lengthened only the growth time, to 7.1 ± 2.4 min R4.l·R4.m·R4.p·C14. Figure 4a shows the PVA, HPC and PVP profiles rising at about 30–37, 55–80 and 65–88 min, and the HPMC profile flat for the 112 min plotted R6. The large spreads for HPC and PVP are, in the authors' view, "most likely" due to the random timing of nucleation: a polymer that delays nucleation will not delay it by the same amount in every run, and the number of nuclei formed also changes how fast the growth phase runs C12. Note that Table 4 prints a total of 90 min for PVP, whereas its onset plus growth is 79 min; the bars in Figure 4b (about 40 and 39 min) agree with the parts, not the printed total X2·R6.

Sulfaguanidine: nothing stops the conversion; PVP slows it most

SGN converted on contact with water whatever the polymer (Figure 5): onset was zero in every case R4.q–x·R8. The polymers differed only in how long the rest of the conversion took. PVP stretched it most, to 6.8 ± 0.5 min against 0.1 ± 0.1 min in water; HPMC (1.3 ± 0.4 min) and PVA (1.8 ± 0.2 min) slowed it slightly; HPC (0.4), CMC (0.1), PAA (0.1) and PEG (0.2 min) hardly at all R4.q–x·C15. Figure 5a shows that even with PVP about 35% hydrate had formed within about 0.3 min, and the remainder followed over about 8 min R7. Note by the AI writer (not discussed in the paper): with one spectrum every 5 or 10 s, growth times of 0.1–0.4 min span only one to a few spectra, so the small differences among the weak polymers are not resolved X11.

Crystal growth rates

Table B gives the seeded growth rates. For CAF, PAA and PVA gave no growth at all during the 5.5 h experiment, while the other polymers gave 0.22–0.68 µm/min against 0.62 ± 0.47 µm/min in water; the authors read these as "no effect" R9·R14·C16. When the CAF concentration was raised threefold to 75 mg/mL, whisker- and needle-like crystals grew even in PAA and PVA R10·C16. For CBZ, the rate in water was 4.9 ± 1.6 µm/min; PAA (4.4), PEG (2.9) and CMC (2.5 µm/min) had "little to no effect", PVA (0.4), HPC (0.08) and PVP (0.04 µm/min) "intermediate effects", and HPMC stopped growth along the needle so that only growth across it, 0.003 µm/min, could be measured R9·C17.

Table B. Linear growth rates of seeded hydrate crystals, mean ± SD of at least five crystals, from Table 5 (CAF 25 mg/mL; CBZ 0.5 mg/mL; polymer at the slurry levels). R9

The authors find that these results "correlate well" with the slurries: PAA and PVA, the strongest CAF inhibitors in the slurry, also affected CAF growth most; for CBZ, HPMC had the largest effect on both, and HPC and PVP were intermediate on both C18. The micrographs add shape information (Figures 6 and 7). In water both hydrates grew as needles; in HPC the CBZ dihydrate grew more slowly in length and wider, though the long axis was still the fastest direction; in HPMC length growth almost stopped, width became the fastest direction, and after 25 min the crystals were rectangular R10·R11·C17.

Two points need care. First, the authors' "no effect" covers some means well below the control: HPC 0.22 and CMC 0.25 µm/min against 0.62 for CAF, and CMC 2.5 and PEG 2.9 µm/min against 4.9 for CBZ, all within the large spread of the control X7. Second, note by the AI writer (not discussed in the paper): the micrograph scale bars do not fit the tabulated units. The CAF needle in Figure 6a lengthens from about 235 to about 430 µm in 2 min, about 100 µm/min against 0.62 in Table 5; the CBZ needle in Figure 7a from about 440 to about 1450 µm in 1.5 min, and the HPC crystal in Figure 7b from about 170 to about 330 µm in 5 min R10·R11·X3. The paper does not say whether the micrographs were taken under the Table 5 conditions, so the safest use of Table 5 is as a ranking of polymers within this paper X3.

Induction times

Table C gives the induction times. For CAF, PAA prevented nucleation for the whole 700 min run and PVA raised the induction time to 625 ± 150 min, against 100 ± 25 min without polymer; the other polymers gave 80–108 min R12·R14. The authors note that this "corresponded well" with the slurries, where PAA stopped and PVA strongly delayed conversion C20. For CBZ, only HPMC prevented nucleation (no particles detected); the other polymers gave 40–70 min against 55 ± 15 min without polymer, which the authors read as no significant increase R12·C21.

Table C. Induction times for nucleation from supersaturated solution on cooling to 25 °C (CAF 35 mg/mL, CBZ 0.35 mg/mL), mean ± SD, from Table 6; the number of runs is not printed. R12·X10

The light-scattering traces (Figure 8) explain the most interesting case. For CBZ with PVP and HPC the signal began to rise at about 60–70 min, close to the control, but climbed slowly; the steep rise came only at about 110 min with PVP and above 200 min with HPC R13. The authors read this as nuclei that form on time but grow so slowly that it takes a long time before enough particles reach detectable size ("most likely") C22. They extend the reading to the slurry: the long CBZ onsets with PVP and HPC do not mean delayed nucleation; "it appears" these polymers do not inhibit nucleation and only reduce growth, and it is "possible" that nuclei form early in the slurry but stay below the Raman detection limit until enough hydrate mass has accumulated C23. The CAF signal rose faster and to a higher level than the CBZ signal overall, which the authors attribute "most likely" to the much higher CAF concentration in solution R13·C19.

Putting the three tests together

Table D. The inhibitors and the stage each one acts on, combining Tables 4, 5 and 6 with the authors' interpretations. R4·R9·R12·C23·C26·C28·C30·C31

Table D is the core of the paper's argument. For CAF, PAA and PVA act on both stages, and the authors find it "reasonable to conclude" that they slow the SMT by delaying hydrate nucleation and, where nuclei do form, by slowing their growth C26. For CBZ, HPMC "appears" very effective at inhibiting nucleation (no particles in the induction test) and nearly stops growth, which makes it "a much more effective inhibitor", while HPC and PVP inhibit growth but have little effect on induction C28·C30. Once nucleation is blocked, the authors note, growth inhibition only matters when seed crystals are present C29.

Discussion

How the authors connect the findings. The argument runs from the three-stage model to three cases.

The first two steps rest on an earlier study and an assumption: dissolution was not rate-limiting for these drugs without additives, and the low polymer concentrations are "not anticipated" to change dissolution R15·C25. The SGN case is an explanation, not a measurement: the lack of inhibition is "most likely" due to the near-instant conversion, and it is "likely" that the conversion began before the polymer could interact with the drug C31.

Why a given polymer works for a given drug. The authors propose that "a specific interaction between the polymer and the crystal surface of the drug may be responsible", and in the conclusions that specific interactions governing polymer adsorption to individual hydrate crystal surfaces are the source of growth and nucleation inhibition ("the results suggest") C33. They point to earlier proposals: PVP adsorbing on sulfathiazole crystals as a net-like layer (Simonelli 1970); hydrogen bonding and structural matching between the cellulose ring of HPMC and the amine hydrogen and carbonyl oxygen of CBZ (Katzhendler 1998); and adsorption tied to polymer–CBZ hydrogen bonding and the polymer's solubility parameter, that is, its hydrophobicity (Tian 2007) C34. In this paper, the reduced CBZ growth rates with HPMC, HPC and PVP "indicate" adsorption that prevents growth C35. The change from needles to rectangles in HPMC points, in the authors' words, to selective adsorption of HPMC on the fast-growing face as "the most likely explanation", with the polymer–surface interaction "presumably" set by the chemical groups exposed on that face C36. CAF, by contrast, kept its needle shape with PAA and PVA even at 75 mg/mL, which "suggests" that adsorption on caffeine is less specific and not limited to the fast-growing faces C37. None of this was measured directly: the paper has no adsorption isotherms, no spectroscopy of polymer–drug interaction and no face indexing C33·C36.

No universal inhibitor. The most effective polymers were PAA for CAF, HPMC for CBZ and PVP for SGN, and "no single polymer is optimal for all compounds" C32. The rankings did not transfer: HPMC stopped CBZ for more than 1500 min but moved the CAF onset by about 11 min, and PAA, which stopped CAF, lengthened only the CBZ growth time and left the CBZ growth rate (4.4 versus 4.9 µm/min) and induction time (45 versus 55 min) essentially unchanged R4.b·R4.j·R4.m·R9.e·R12.e·X1. The authors conclude that it "is currently hard to predict a priori" which polymer will work best, and announce adsorption-isotherm and morphology studies C38.

What this suggests for formulation. The authors' practical message is that hydrate formation in water "can be manipulated" with common excipients, which could help keep the anhydrate during water-based processing C32·C2. Note by the AI writer (not discussed in the paper): three further points follow from the data. First, very little polymer can be enough: 0.04 mg/mL HPMC kept CBZ anhydrous for more than 1500 min, although no concentration series was run, so the minimum effective level is unknown P1·R4.j·X8. Second, a growth-only inhibitor buys time rather than protection: with HPC or PVP the CBZ slurry still reached 80% hydrate within about 1.5 h R4.k·R4.n. Third, a drug that converts within seconds of wetting, as SGN did, is unlikely to be protected by a polymer dissolved in the granulating liquid R8·C31.

Limitations

Limitations stated by the authors. The authors acknowledge that the mechanism of polymer effects on crystallisation "is often unclear" in the literature and that the best polymer cannot yet be predicted C24·C38. They note that replicate slurries with inhibiting polymers varied widely, "most likely" because nucleation is random C12. They recognise that the slurry onset and the induction time disagree for HPC and PVP with CBZ and offer slow growth below the Raman detection limit as "a possible explanation" C23. The first CAF calibration failed because dissolved CAF distorts the spectra, and needed a correction C7.

The AI writer's own judgement (1): what the induction test can and cannot show. The light-scattering probe counts particles above 1 µm, so an induction time measures nucleation plus the growth needed to reach detectable size P6. The authors use exactly this point to explain why HPC and PVP lengthen the slurry onset without lengthening the induction time C23. By the same reasoning, the absence of particles with HPMC, which nearly stopped CBZ growth, does not by itself show that HPMC blocks nucleation; it is consistent with blocked nucleation and with nuclei that never grow large enough to be counted X13·C28. The general statement that the polymers affecting the kinetics "also inhibited the nucleation and/or the crystal growth" holds for the strongest inhibitors but not for every polymer: for CAF, HPMC, HPC and CMC lengthened the slurry onset without lengthening the induction time, and HPMC left the CAF growth rate unchanged (0.67 versus 0.62 µm/min); for CBZ, PAA lengthened the growth time without slowing seeded growth or nucleation X6·C40.

The AI writer's own judgement (2): conditions and comparability. Each drug was tested at a single polymer concentration, and the levels differ 125-fold between drugs (0.04 to 5 mg/mL), so the "most effective polymer per drug" ranking compares polymers at different levels and cannot be read as an intrinsic order of polymer strength X8. The slurry and growth-test temperatures, the temperature of the cited solubilities and the pH of the polymer solutions (PAA and CMC carry acid groups) are not stated X10. The driving force also differs between tests: from the cited hydrate solubilities, the growth solutions were about 1.1 times (CAF) and 3.6 times (CBZ) supersaturated, the induction solutions about 1.6 and 2.5 times, and in the slurries the ceiling is the anhydrate solubility, about 2.3, 3.1 and 1.3 times for CAF, CBZ and SGN; the three tests are therefore best compared as rankings X9. The anhydrous SGN and the calibration hydrates were made in-house and characterised only by the Raman spectra of Figure 1 X14.

The AI writer's own judgement (3): errors in the printed record. Table 4 prints the CAF–HPMC onset as 1.83 min, which is most likely 18.3 X1. Its CBZ–PVP total, 90 min, is not the sum of onset and growth (79 min) X2. Table 5 growth rates do not match the micrograph scale bars, which show crystals lengthening by hundreds of micrometres within minutes X3. Smaller slips: sulfaguanidine is "SFG" in Table 1; the text cites "Tab. 1" for delay times that are in Table 4; one sentence reads "none of the polymeric excipients were unable to completely inhibit" where the data show none was able; the CBZ–HPMC run is ">1500" min in Table 4 but "up to 24 h" in the text; the CAF–PAA induction run is "700 min" in the text but its trace in Figure 8a runs to about 1000 min X4·X5. None of these changes the main conclusions, but anyone reusing Tables 4 and 5 should correct for them.

Conclusion

The paper shows, with triplicate in-line Raman kinetics, that common polymeric excipients dissolved in the water of a slurry can delay or stop the solvent-mediated conversion of an anhydrous drug to its hydrate, but in a strongly drug-specific way: PAA stopped caffeine, HPMC stopped carbamazepine, and nothing stopped sulfaguanidine, which PVP only slowed R4·C32. Separate growth and induction tests show that the effective polymers slow hydrate growth, lengthen the induction time for nucleation, or both; HPC and PVP slow carbamazepine growth without delaying its nucleation R9·R12·C30. The authors attribute the effects to specific adsorption of each polymer on the hydrate crystal surfaces, which remains an untested hypothesis C33.