Polymers against water: how PVP, HPMC and HPMCAS hold back crystallization of amorphous felodipine in humid storage

Background and motivation

Amorphous solid dispersions raise the apparent (kinetic) solubility of poorly water-soluble drugs, but the amorphous state is metastable, and if the drug crystallizes during storage the solubility gain is lost (Chiou 1971; Ford 1986; Serajuddin 1999; Leuner 2000) C1. Polymers such as poly(vinylpyrrolidone) (PVP) and the cellulose derivatives hydroxypropylmethylcellulose (HPMC) and HPMC acetate succinate (HPMCAS) are used to stabilize the amorphous drug. Two explanations had been offered for their effect: antiplasticization, since a dispersion usually has a higher glass transition temperature ($T_g$) than the amorphous drug, so that a lower molecular mobility "may be achieved"; and specific drug–polymer interactions such as hydrogen bonds, which "it has also been suggested" contribute to stability (Taylor and Zografi 1997; Matsumoto and Zografi 1999; Miyazaki 2004) C2.

The starting point is the authors' own earlier study (Konno and Taylor 2006, ref 11). With water rigorously excluded, PVP, HPMC and HPMCAS each inhibited nucleation of amorphous felodipine, and the three could not be told apart: at a given polymer content each lowered the nucleation rate by the same extent C3·R1. Real formulations, however, are never perfectly dry. Amorphous solids absorb water into their bulk as well as onto their surface, so they are more hygroscopic than crystals; absorbed water plasticizes the solid, raises molecular mobility and can trigger crystallization (Zografi 1988; Ahlneck and Zografi 1990; Yoshioka 1995; Shamblin and Zografi 1999; Oksanen and Zografi 1990) C4. How much water a material absorbs depends on its polar groups: PVP takes up more water than poly(vinylacetate), in line with the stronger hydrogen bond between water and the pyrrolidone group (Taylor, Langkilde and Zografi 2001) C5.

This creates a conflict that the dry study could not see. The drug in a dispersion is usually hydrophobic and the polymer hydrophilic, so the water a dispersion takes up "is likely" to come mainly from the polymer, and the dispersion "may actually" hold more water than the drug alone at the same relative humidity (RH) C6. In the authors' words, the question is how well a polymer competes as a crystallization inhibitor against the crystallization-promoting effect of the water it brings in, and how this relates to the polymer's hygroscopicity. The discussion adds that very few studies had measured crystallization in such ternary drug–polymer–water systems, although single-component amorphous drugs exposed to moisture and dry binary dispersions had both been studied widely C7, and that the ternary case matters in practice because it is "virtually impossible" to exclude moisture completely from a formulation C8.

To answer the question the authors measured nucleation rates of felodipine in thin films containing 0–25% of each polymer after storage at five humidities, and characterized matching bulk dispersions by differential scanning calorimetry (DSC, for $T_g$), gravimetric water vapour sorption and Fourier transform infrared spectroscopy (FTIR, for drug–polymer interactions) A1·A2·A3·P1·P2·P5·P6·P7.

Background concepts

Nucleation rate. Crystallization from an amorphous solid has two steps: nuclei form (nucleation) and then grow. This paper measures only the first step. The nucleation rate is the number of new nuclei that appear per unit volume and per unit time; here it is given in m⁻³ s⁻¹ and plotted on a logarithmic axis, because it changes by orders of magnitude between samples P4·R1·R3. A nucleus is counted only once it has grown large enough to be seen under the polarized light microscope, about 1 µm in this setup P4·C38.

From sites per area to sites per volume. The microscope counts nuclei in an area of film, while the rate is wanted per volume. The authors multiply the number of sites per area by the depth of field of the objective, the thickness of sample that is in focus, given by Eq. 1 (from a handbook of optics, ref 18) E1·C38:

Here $D_{\mathrm{t}\mathrm{o}\mathrm{t}}$ is the total depth of field, $\lambda$ the wavelength of the light (550 nm), $n$ the refractive index of the medium between sample and lens (air, 1.000), $\mathrm{N}\mathrm{A}$ the numerical aperture of the objective, $e$ the smallest distance the detector can resolve (14 µm) and $M$ the lateral magnification (10×) E1. In plain terms: the in-focus slice has one part set by diffraction (the first term) and one set by the detector's resolution (the second term), and the authors obtain 0.0144 mm for their 10× objective E1·P4. The paper does not print the numerical aperture; a value of 0.25, typical for a 10× objective, reproduces the stated 0.0144 mm exactly (8.8 µm + 5.6 µm), so the printed number is consistent with the equation E1. Film thickness is not given, so whether the film is thicker or thinner than this slice is not stated X7.

Plasticization and antiplasticization. Mixing two amorphous components gives a single $T_g$ between those of the pure components. Adding a component with a lower $T_g$ lowers the mixture's $T_g$ (plasticization); adding one with a higher $T_g$ raises it (antiplasticization). Water, with a very low $T_g$, is a strong plasticizer; the three polymers have dry $T_g$ values far above that of felodipine (about 168, 141 and 114 °C for PVP, HPMC and HPMCAS against about 43 °C for the drug, from the earlier study) R7.a·R7.h. A lower $T_g$ is usually read as higher molecular mobility at the storage temperature, and higher mobility as faster crystallization; this is the antiplasticization explanation of polymer stabilization that the paper puts to the test C2·C25.

Reduced temperature $T/T_g$. To compare samples with different $T_g$ at one storage temperature, crystallization rates are often plotted against the ratio of the storage temperature to $T_g$, both in kelvin; a ratio near 1 means the sample is stored close to its glass transition. The authors use this plot (Figure 11) to ask whether $T_g$ alone orders the nucleation rates P9·R13. As discussed later, the values plotted in this figure do not correspond to a kelvin ratio X1.

Hygroscopicity and water sorption isotherms. A sorption isotherm records the equilibrium weight gain of a dry sample at a series of RH values at constant temperature. It shows how much water a material holds at a given storage RH, which is the quantity that links storage conditions to plasticization P7·R8.

Infrared markers of drug–polymer hydrogen bonding. Felodipine has an NH group and two carbonyl groups; PVP has one carbonyl group per repeat unit M1·M2. In the NH-stretching region of a felodipine–PVP dispersion the band has a component near 3345 cm⁻¹, assigned in the earlier study to felodipine–felodipine hydrogen bonds, and a lower-wavenumber shoulder at 3,291 cm⁻¹ assigned to felodipine–PVP hydrogen bonds; in the carbonyl region a shoulder at around 1,655 cm⁻¹ is assigned to PVP carbonyl groups hydrogen bonded to the drug R17. Crystalline felodipine has its NH stretch at 3,373 cm⁻¹, so a shift of the band to that position marks crystallization R17.

Methods in detail

The study combines four kinds of measurement on two kinds of sample. Nucleation and FTIR were done on thin spin-coated films; $T_g$ and water uptake were measured on bulk dispersions of the same compositions made by rotary evaporation. The nucleation rates were then replotted against $T_g$ and against water content to see which of the two orders them P1·P2·P8·P9.

Contribution 1: nucleation rates in films under controlled humidity

Problem. The dry study had shown that the three polymers inhibit nucleation equally. The open question is whether that equality survives when the films take up water, and how the protection changes with humidity and polymer content C3.

What they did. Felodipine and polymer were dissolved together in dichloromethane:ethanol 1:1 and dropped onto a glass cover slip spinning at about 2,500 rpm; the film was heated to 90 °C for a few seconds to drive off solvent, and the whole preparation was done in a nitrogen-purged glove box below 10% RH to keep water out P1. Films held 0, 3, 7, 10, 15, 20 or 25% (w/w) polymer P1. They were stored at 22 °C in desiccators over phosphorus pentoxide (0% RH) or saturated solutions of lithium chloride (11%), magnesium chloride (33%), sodium bromide (58%) or sodium chloride (75%) P3·M7. At each sampling time a film was taken out for about 5 min, the nuclei in 12 areas were counted under the polarized light microscope, and the film was put back; the same films were followed throughout, and each experiment was done in triplicate P3·P4. Sites per area were converted to sites per volume with the 0.0144 mm depth of field (Eq. 1) E1·P4. In some films nuclei formed and grew preferentially at the edge; these sites were left out R16.

What is not stated. The paper does not give the sampling times, the total observation period, the film thickness, or how the rate was obtained from the counts over time (for example the slope of site density against time) X7. The 0% RH data are not new: they are the results of the earlier study, "reproduced" in Figure 1a R1·X10.

Contribution 2: $T_g$ and water uptake of matching bulk dispersions

What they did. Bulk dispersions with 0–100% polymer were made by dissolving drug and polymer in dichloromethane:ethanol 1:1 and removing the solvent on a rotary evaporator at 60 °C, followed by several hours under vacuum P2. For $T_g$, about 5 mg was held for 2 days at the chosen RH at 25 °C, sealed hermetically and heated at 20 K/min; the onset $T_g$ was taken from a second heating after the sample had been heated 20 K above $T_g$ and cooled 50 K below it, in triplicate with fresh samples P6. For water uptake, 7–10 mg were dried at 50 °C under nitrogen in a gravimetric sorption analyser and then exposed to 5–95% RH in 10% steps at 25 °C P7. HPMC and HPMCAS samples moved to the next step once the weight changed by less than 0.01% in 15 min (or after 90 min), whereas PVP samples took longer to equilibrate and were held 1,000 min at each step P7. The water contents at the storage RHs used for the nucleation study were estimated from these isotherms P8.

Why it matters for the argument. These two measurements give the two candidate explanations of the nucleation data: if mobility set by $T_g$ controls nucleation, the rates should follow $T_g$; if the polymer acts mainly through the water it brings in, the rates should follow water content C25·C30.

Contribution 3: replotting nucleation against $T/T_g$ and against water content

Figure 11 plots the nucleation rates of the 7, 15 and 25% polymer films against $T/T_g$, with $T$ given in the caption as 22 °C and $T_g$ from DSC P9·R13. Figure 12 plots the rates of the 0, 7, 15 and 25% polymer films against water content, with one regression line for each polymer content, pooling the three polymers P8·R14. These two plots carry the paper's main interpretation C26·C30.

Contribution 4: FTIR of films after humid storage

What they did. Films were spin-coated onto ZnS discs in the dry glove box, stored at 75% RH (25 °C in the figure captions) for 1, 3, 4, 5, 6 and 7 weeks (25% polymer) or 2, 4, 6, 8 and 14 days (3% polymer), and dried in dry nitrogen before each spectrum; 128 scans were collected at 4 cm⁻¹ resolution with the optics purged by dry nitrogen P5. Because the water is removed before measuring, any lasting change in the bands reflects a change in how drug and polymer are arranged, not water itself P5·C21. The 3% polymer films are used, in the authors' words, to show that the changes seen in the 25% PVP film "cannot be accounted for by crystallization": in a film that does crystallize, the NH band moves to 3,373 cm⁻¹ R12·R17·C24.

Experimental analysis

Experimental setup. One drug, felodipine, was combined with three polymers: PVP K29/32, HPMCAS (AQOAT AS-MF) and HPMC (Pharmacoat 606) M1·M2·M3·M4. Nucleation was followed in films with up to 25% polymer at five humidities; $T_g$ and water uptake were measured on bulk dispersions with 0–100% polymer; spectra were taken on 3% and 25% polymer films after humid storage A1·A2·A3·A4. All values below are read from the figures and are approximate. Table A collects the nucleation rates at three humidities.

Table A. Nucleation rate of felodipine (m⁻³ s⁻¹, 22 °C) in spin-coated films, read from Figure 1 (approximate; n = 3, error bars omitted). The 0% RH column is the dry data of the earlier study reproduced in Figure 1a. R1·R2·R3

Main result 1: water speeds nucleation, polymer slows it, and polymer wins. For felodipine alone the rate rose about 20-fold from 0% to 75% RH, from about 1.4×10⁵ to about 2.9×10⁶ m⁻³ s⁻¹, with intermediate values of about 1.6×10⁵ (11% RH), 2.7×10⁵ (33%) and 9.8×10⁵ (58%) R1.a·R4.a·R2.a·R4.h·R3.a. The authors call the rise "anticipated" C11. At every humidity, every polymer film nucleated more slowly than the drug alone; at 75% RH the 25% polymer films were three to four orders of magnitude slower R3.g·C13. This holds even though the polymer films contain more water: at 75% RH the 25% polymer dispersions held about 1.7–4.7% water against about 0.7% for felodipine alone (Table B) R9.a·R9.d·C13. Polymer content also changed how strongly humidity mattered. With 25% HPMCAS the rate stayed between about 2.9×10² and 3.5×10² from 11% to 75% RH, and with 25% HPMC it rose only from about 2.9×10² to 7.2×10²; the authors describe the slope of rate against RH for these films as close to zero R4.g·R2.g·R4.n·R3.g·C12. With 3% PVP, in contrast, the rate rose about 300-fold between 0% and 75% RH (about 2.9×10³ to 8.6×10⁵), a steeper rise than the drug alone, though at every RH it stayed below the drug's R1.b·R3.b·C12.

Main result 2: the polymers part ways only when water is present. Dry, 3% of any polymer gave about 2.5–2.9×10³ m⁻³ s⁻¹ and 15% gave about 0.8–1.0×10³, with little to choose between them R1·C9. In humid air PVP fell behind. At 33, 58 and 75% RH PVP gave the fastest rate at every polymer content, and HPMC was usually slightly faster than HPMCAS; at 11% RH the three were still close R2·R3·R4. The gap was largest at low content: at 75% RH, 3% PVP lowered the rate only about 3-fold relative to the drug alone, while 3% HPMC and 3% HPMCAS lowered it about 145-fold and about 220-fold R3.a·R3.b·C10. With more polymer the gap closed: at 75% RH the PVP/HPMCAS ratio fell from about 66 at 3% to about 11 at 15% and about 7 at 25% R3.b·R3.e·R3.g. The authors see the divergence "most clearly" in Figure 1c and describe the polymers as converging with increasing content C10.

Table B. Water uptake at 75% RH and 25 °C (pure materials from Figure 5, 25% dispersions from Figure 6) and onset $T_g$ of the 25% dispersions (Figure 3a), all approximate; dry felodipine $T_g$ as stated in the text. Figure 5a gives a lower value, about 2.3%, for the 25% PVP dispersion. R8.v·R8.a·R8.o·R8.h·R9.d·R5.a·R5.e·R7.a·X2

Main result 3: the polymer brings in the water, PVP most. Amorphous felodipine took up very little water, about 0.7% at 75% RH and about 1.4% at 95% RH, while the pure polymers took up about 33.5% (PVP), 10.9% (HPMC) and 6.2% (HPMCAS) at 75% RH R8.v·R8.a·R8.o·R8.h·C17. For every polymer the water uptake of the dispersions rose with polymer content over the whole RH range, which the authors read as polymer "increases the hygroscopicity of the system" R8·C18. At 75% RH the PVP dispersions held the most water at every content, from about 1.7% at 7% PVP to about 11.1% at 70% PVP, with the gap to the cellulose dispersions growing at high polymer content R9·C18. A caution applies to the low-PVP values: the sorption isotherms of Figure 5a give about 1.2, 1.7 and 2.3% water at 75% RH for 7, 15 and 25% PVP, whereas Figure 6 and the later plots use about 1.7, 2.6 and 4.7%; the HPMC and HPMCAS values agree between figures R8.e·R8.f·R8.g·R9.b·R9.c·R9.d·X2.

Main result 4: water lowers $T_g$, but $T_g$ does not order the rates. The $T_g$ of the 25% dispersions fell steadily with storage RH: for PVP from about 54.9 to 31.1 °C, for HPMCAS from about 44.2 to 35.6 °C and for HPMC from about 43.1 to 34.4 °C between 0 and 75% RH (Figure 3a) R5·C14. At about the same water content PVP lowered $T_g$ more, for example by about 13.3 °C at about 1.1% water against about 6.2 °C for HPMCAS at about 1.1% (Figure 3b), which the authors express as water being "a more efficient plasticizer" for the PVP system R6·C14. Because it started higher, the 25% PVP dispersion kept the highest $T_g$ at low RH and ended slightly lower than the cellulose dispersions at 75% RH; the authors place the crossing at about 60% RH, while the plotted values already put PVP below HPMCAS at 58% RH R5.d·R5.e·C15·X12. Figure 4 extends this to all compositions: dry $T_g$ (from the earlier study) rose with polymer content, but after 75% RH every dispersion with up to 25% polymer sat at about 30–35 °C, below dry felodipine (about 43 °C) R7·C16. For PVP dispersions stored at 75% RH, $T_g$ peaked near 70% polymer (about 51.8 °C) and dropped to about 10.3 °C for pure PVP R7.f·R7.h·C16. The drop in $T_g$ between dry and 75% RH grew with PVP content from about 15 to about 158 °C; the authors call it "obvious" from Figure 6 that plasticization by water is directly related to the polymer's hygroscopicity R9·C19·C20. For the cellulose polymers the drop did not grow steadily (HPMCAS about 12 °C at 7%, about 3.8 °C at 50%, about 24 °C for the pure polymer), so the steady rise is a PVP result R9.b·R9.e·R9.h·X3.

The key point for the argument follows from Table A and Figure 4 together. Every humid film with up to 25% polymer has a $T_g$ below that of dry felodipine, yet the cellulose films and the PVP films with more than 7% polymer nucleate more slowly than dry felodipine, so the polymers "are still able to act as crystallization inhibitors even when $T_g$ is reduced" R7·R3·R1.a·C27. For 10% PVP the margin is small, about 1.0×10⁵ against about 1.4×10⁵ m⁻³ s⁻¹ R3.d·R1.a·X9. In Figure 11, rates at the same $T/T_g$ differ by more than an order of magnitude; at about 0.68–0.69 they range from about 2.9×10² (25% HPMC) to about 9.4×10³ (7% PVP), and the authors conclude that "No correlations could be drawn" between nucleation rate and $T_g$ R13·C26. They find the polymer differences "difficult to rationalize" on $T_g$: without water the low-polymer dispersions have $T_g$ values close to that of felodipine, they absorb little water, and the $T_g$ values with the three polymers are not very different from each other C25.

Table C. The authors' regression lines of nucleation rate (m⁻³ s⁻¹) against water content (% w/w) in Figure 12, one per polymer content with the three polymers pooled; line ends read from the figure (approximate), fold-rise computed by the AI writer. R14

Main result 5: against water content, polymer content matters more than polymer type. Figure 12 replots the rates of Figures 1 and 2 against the water content of each film, estimated from the sorption data R14·P8. Three features stand out. First, felodipine alone takes up the least water but is the most sensitive to it: its line rises about 23-fold over less than 0.7% water, which the authors liken to the moisture-enhanced nucleation of indomethacin (Andronis and Zografi 2000) R14.a·C28. Second, with more polymer, more water is absorbed at a given RH but the rate depends less on it; at 25% polymer the line rises only about 10-fold over about 4.7% water (Table C) R14·C29. Third, at each polymer content the PVP, HPMC and HPMCAS points lie near one line. From this the authors conclude that, "within experimental error", each polymer's effect "appears to be dependent on the water content of the system", so that the differences between the polymers "appear to be reasonably explained" by their different water uptake, "although it must be conceded that there is considerable scatter in the nucleation rate data" C30. They go on to "suggest" that polymer chemistry matters to stabilization only insofar as it changes moisture sorption, and judge it "unlikely" that the stabilizing ability in the presence of water comes from specific drug–polymer interactions, the same conclusion they had reached for the dry films C31·C32.

The scatter is visible in the data. At about the same water content (1.0–1.3%), 7% PVP nucleated at about 2.9×10⁴, 7% HPMC at about 1.2×10⁴ and 7% HPMCAS at about 6.4×10³ m⁻³ s⁻¹, a 2.5–4.5-fold spread within one polymer content, while 25% polymer films at the same water content nucleated at about 2.9–4.9×10² R14.b·R14.h·R14.e·R14.d·R14.g·R4.j·R3.c·R2.g·R4.n·X4. So polymer content clearly matters more than polymer type at a given water content, but the claim that type matters not at all rests on overlap within this spread X4.

Main result 6: humid storage disrupts felodipine–PVP hydrogen bonds. In the 25% PVP films the drug–PVP NH shoulder at 3,291 cm⁻¹ lost intensity relative to the drug–drug component over 1–7 weeks at 75% RH, and the 1,655 cm⁻¹ carbonyl shoulder of hydrogen-bonded PVP decreased as well (Figure 7) R10·R17. The NH band maximum stayed at about 3333–3339 cm⁻¹, with no band at 3,373 cm⁻¹ R10. Since the spectra were recorded after drying, the authors conclude that "absorbed moisture disrupts drug–polymer interactions and that this disruption persists even after moisture is removed" C21. The 25% HPMCAS and 25% HPMC films showed no notable change in either region (Figures 8 and 9) R11·C22. The 3% films confirm what crystallization looks like in this measurement: in 3% PVP the NH maximum moved from about 3341 to about 3372 cm⁻¹ over 14 days (the text gives 3,345 to 3,373 cm⁻¹ after 2 weeks), the position of crystalline felodipine, while in 3% HPMCAS it stayed at about 3343–3345 cm⁻¹, and 3% HPMC behaved like HPMCAS; this matches the nucleation rates of Figure 1c R12·R17·R3.b·C23. The 25% PVP changes are therefore not crystallization C24. One inconsistency in the text: the methods state that no crystallization was seen in the FTIR films after moisture exposure, which can only refer to the 25% films R15·X8.

Role of the controls. Three comparisons act as controls. The dry data show that the polymers are equivalent when water is absent, so any difference in humid air must come with water R1·C9. The cellulose FTIR spectra show that the hydrogen-bond changes are specific to PVP R11·C22. The 3% films show that the FTIR method does detect crystallization when it happens, so the absence of the 3,373 cm⁻¹ band in 25% PVP is meaningful R12·C24.

Discussion

The authors' chain of reasoning. The argument can be laid out in steps, with the case split that decides whether a film is well or poorly protected C13·C17·C18·C26·C30.

In this reading the polymer is a nonspecific inhibitor whose net benefit in humid storage is set by how much water it lets into the film C31·C32. The authors present this as an interpretation, with hedges ("appears to be reasonably explained", "These results suggest", "it is unlikely"), and they flag the nonspecific stabilization by chemically different polymers as something that "clearly needs to be investigated more thoroughly" C30·C31·C32·C33.

The PVP side story. The FTIR results add a second effect that is specific to PVP. The authors read the loss of drug–PVP hydrogen bonding, without crystallization, as a sign that the 25% PVP dispersion "maybe undergoing a partial phase separation induced by the absorbed moisture", and the fact that it does not recover on drying as "indicating" that the system is "kinetically trapped as a partially phase separated system" C34·C35. They offer two possible reasons why PVP, and not the cellulose polymers, behaves this way: it "may" simply absorb more water, or there may be "more complex reasons" relating to the relative polarities of drug, polymer and water C36. They expect phase separation to affect the performance of a solid dispersion and call for more work C37. In the abstract the same finding is worded as molecular-level changes "which may be indicative of phase separation", and in the conclusions as "evidence of a tendency to phase separate" C34.

Where the conclusions apply. The data cover one drug (felodipine), three polymers at 0–25% in films and 0–100% in bulk, storage at 0–75% RH and 22 °C for nucleation, and 25 °C for $T_g$, sorption and FTIR storage A1·A2·A3·P3·P6·P7. Only nucleation was measured, not crystal growth P4. The comparison of polymers at equal water content is made within this one drug, so the statement that polymer chemistry matters only through water uptake is a statement about felodipine with these three polymers C31.

Note by the AI writer (not discussed in the paper): what this suggests for formulation. For a hydrophobic drug stored humid, the polymer that looks best dry is not necessarily best in use. Here the three polymers were indistinguishable dry, but at 3% loading and 75% RH the cellulose polymers held nucleation about 40–70 times below PVP; at 25% loading the difference shrank to 3–7-fold R1·R3.b·R3.g. Water uptake of the dispersion at the intended storage RH, which is cheap to measure, ranked the polymers better than dry $T_g$ did R9·R13·R14. Raising polymer content reduced the water sensitivity of nucleation for all three polymers R14. Whether a drug that hydrogen bonds more strongly to the polymer would follow the same water-content ranking is not tested here.

Limitations

Limitations stated by the authors. The authors concede "considerable scatter in the nucleation rate data", so their water-content explanation is offered "within experimental error" and with the hedge "appears to be reasonably explained" C30. They state that the polymer differences are "difficult to rationalize" on $T_g$, and that the nonspecific stabilization they infer still "needs to be investigated more thoroughly" C25·C33. They present phase separation in PVP dispersions as a possibility ("maybe", "may") and as a phenomenon "that needs to be explored in more detail" C34·C36·C37. They also note that edge nucleation was excluded from the counts R16.

The AI writer's own judgement (1): the water-content explanation. The central claim depends on the water contents assigned to the PVP films, and these differ between figures. For 7, 15 and 25% PVP at 75% RH the sorption isotherms of Figure 5a give about 1.2, 1.7 and 2.3% water, while Figure 6, Figure 3b and Figure 12 use about 1.7, 2.6 and 4.7%; Figure 3b says its values were estimated from Figure 5, and the HPMC and HPMCAS values agree across figures X2. With the lower PVP values the PVP points in Figure 12 would move to lower water content at the same rates and would no longer overlap the cellulose points, so the explanation would weaken. Even with the higher values, 7% PVP lies 2.5–4.5 times above 7% HPMC and HPMCAS at similar water content, and the regressions pool the three polymers with $R^2$ of 0.85–0.89 X4·X5. The regression form is not stated; the lines are straight on the logarithmic rate axis, so log rate appears to have been regressed on water content X5. The water contents and $T_g$ values come from bulk samples at 25 °C, while nucleation was counted in films at 22 °C, so neither was measured on the films themselves X6. The sorption protocol also differed: PVP samples were held 1,000 min per RH step, the cellulose samples only until a 15-min stability criterion or 90 min X11.

The AI writer's own judgement (2): the $T/T_g$ plot and the rate method. The caption of Figure 11 gives $T$ = 22 °C, but the plotted abscissa reproduces 25 °C divided by $T_g$ in °C: for the 25% PVP film, 25/54.9 = 0.455 and 25/31.1 = 0.804, against plotted values of about 0.456 and 0.804; the 25% HPMC and HPMCAS series agree in the same way X1·R13·R5. A ratio of kelvin temperatures would lie near 0.90–0.98 for these samples. Because the plotted quantity still increases monotonically as $T_g$ falls, the conclusion that rate does not follow a single trend with $T_g$ is unaffected, but the axis is not a reduced temperature and cannot be compared with $T/T_g$ plots in other work X1. The nucleation method is described only partly: sampling times, observation period, film thickness and the way a rate was obtained from the counts are not given, and the per-volume conversion uses the 0.0144 mm depth of field as the sampled thickness X7. The dry rates and dry $T_g$ values come from the earlier paper X10.

The AI writer's own judgement (3): phase separation. The phase-separation reading rests on IR band changes measured after drying. Figures 3a and 4 report a single $T_g$ for each sample, and the paper does not say whether a second transition was looked for in the humidified PVP dispersions; no imaging or other direct test of phase separation was made X14. The proposal is plausible as an explanation of the spectra, but in this paper it is a hypothesis, which is how the authors word it C34.

Smaller points. The ΔTg between dry and 75% RH grows steadily with polymer content only for PVP X3. The 25% PVP $T_g$ curve crosses HPMCAS between 33 and 58% RH and HPMC between 58 and 75% RH, rather than at "about 60% RH" for both X12. The RH of individual points in Figures 3b, 11 and 12 is not labelled; the assignments used in this article were made by matching rates and $T_g$ drops with Figures 1–3 X13.

Conclusion

The paper shows that, for amorphous felodipine, polymer and absorbed water act in opposite directions on nucleation, and that the polymer wins at every humidity tested: every polymer-containing film nucleated more slowly than the drug alone at the same RH, although it held more water C13·R1·R3·R9. It also shows that PVP, HPMC and HPMCAS, equivalent when dry, differ in humid storage, PVP being the weakest and the difference being largest at low polymer content C9·C10·R3. The authors' interpretation is that this difference "appears to be reasonably explained" by how much water each polymer brings into the film, not by $T_g$, which does not order the rates, and that polymer chemistry matters only through moisture sorption C26·C30·C31. In 25% PVP films, humid storage disrupts drug–polymer hydrogen bonding in a way that survives drying, which the authors read as possible partial phase separation C21·C34.

The firm results are the nucleation rates themselves, the humid-air ranking of the polymers and the much higher water uptake of PVP R1·R2·R3·R4·R8·R9. The explanation by water content is suggestive rather than proven, given the scatter, the pooled regressions and the disagreement between figures on the PVP water contents X2·X4. Later work could test it by measuring water content on the films themselves and by comparing polymers at matched water content rather than matched RH.