Heat versus water: how temperature and moisture change the mixing of felodipine and PVP in amorphous dispersions

Background and motivation

An amorphous solid dispersion (ASD) keeps a poorly soluble drug in a high-energy glassy state by mixing it at the molecular level with a polymer. Its value lasts only as long as the drug does not crystallize, and during manufacture, storage and accelerated stability testing the dispersion meets both humidity and temperature excursions (Introduction). Earlier work had shown that water raises nucleation rates of drugs from amorphous dispersions (Miyazaki 2004, ref 1), acts as an antiplasticizer at very low content and a plasticizer at higher content (refs 2–4), may change the thermodynamic driving force for crystallization (Marsac 2008, ref 5), and in small amounts strongly changes crystallization rates (refs 1, 6, 7). Water may also compete with the drug for hydrogen-bonding sites on the polymer. Heat, in turn, can destabilize amorphous systems through molecular mobility (Aso 2004, ref 8). C2·C3

The authors' starting point is that these effects had been described phenomenologically, as faster or slower crystallization, but not as changes in the miscibility of drug and polymer. They argue that the stability of a molecular-level dispersion depends on how well its components mix, which is set by the thermodynamics of mixing: the entropy of mixing is always favourable, so the enthalpy of mixing decides, and that enthalpy reflects the balance between cohesive interactions (drug–drug, polymer–polymer) and adhesive interactions (drug–polymer). C1·C4 Later, in the Discussion, they add the reason why miscibility matters for stability: whatever its mechanism (lower drug chemical potential, a higher kinetic barrier, hindered transport of drug to the crystal), a polymer can only inhibit crystallization if it is intimately mixed with the drug. C30·C31

Felodipine with poly(vinyl pyrrolidone) (PVP) K29/32 was a convenient model. The group had already shown by infrared (IR) spectroscopy that solvent-evaporated felodipine–PVP dispersions of all compositions show drug–polymer interactions at room temperature (Konno 2006, ref 10; Marsac 2006, ref 19), had predicted the pair to be miscible from glass transition and mixing-thermodynamics data (Marsac 2006, ref 25), and had estimated a slightly negative Flory–Huggins interaction parameter for it (Marsac 2009, ref 43). R23·R22 Felodipine is a hydrophobic calcium channel blocker, and PVP a hydrophilic polymer, so water should interact very differently with the two. M1·M2

The paper therefore asks, at the molecular level, how temperature and relative humidity (RH) affect the structure and miscibility of this dispersion. IR spectroscopy of hydrogen bonding is the main tool; differential scanning calorimetry (DSC), atomic force microscopy (AFM) and transmission electron microscopy (TEM) are used to confirm what the spectra suggest.

Background concepts

The NH stretch of felodipine as a mixing probe. Felodipine has one NH group that donates a hydrogen bond to a carbonyl. In the pure amorphous drug the acceptor is the carbonyl of another felodipine molecule; in a dispersion it can also be the carbonyl of PVP. The two situations give separate IR bands: at 25 °C the drug–drug band is centred at 3341 cm−1 and the drug–polymer band at 3290 cm−1, and PVP itself absorbs little in this region. R1 The assignment comes from the group's earlier work (Konno 2006, ref 10), and later in the paper the two bands are referred to as the 3345 and 3290 cm−1 peaks. C5·R1 The logic of the whole paper rests on this assignment: if drug and polymer are mixed, much of the drug's NH is bonded to PVP and the 3290 cm−1 band is strong; if the drug gathers in its own phase, the drug–drug band grows. A second, simpler use of the bands is their position: a hydrogen bond that weakens shifts its stretching band to higher wavenumber. C6

Amorphous–amorphous phase separation. A miscible amorphous dispersion is one glass. If it becomes immiscible it can split into two glasses of different composition, here a drug-rich and a polymer-rich phase, without any crystals forming. Such a separation is hard to see: both phases are amorphous, the domains can be small, and if water caused the separation, the water itself can obscure the measurement. C14·C38

Critical solution temperatures. Temperature can make two liquids mix or demix. Phenol and water have an upper critical solution temperature of 66.8 °C: above it they form one phase at every composition, below it two phases can coexist. Polystyrene–cyclohexane and nicotine–water show both an upper and a lower critical solution temperature, with a closed region of immiscibility between them. R24 The authors use these examples to argue that a drug–polymer dispersion "might undergo mixing or demixing as a function of temperature", provided there is a driving force and the kinetics are not prohibitively slow. C32

Flory–Huggins parameters and the Δχ effect. The Flory–Huggins interaction parameter χ summarizes how favourably two components mix: a negative value means favourable interactions, a large positive value unfavourable ones. In a ternary mixture of a solvent (1) with two components (2 and 3), polymer scientists found that a solvent can split two components that are miscible on their own when it interacts very differently with each of them (Pouchly 1976, ref 39; Robard 1977, ref 40). The paper writes this asymmetry as: C36·E1

In plain terms: the larger the difference between the solvent's interaction parameter with component 2 (${\chi }_{12}$) and with component 3 (${\chi }_{13}$), the stronger the push to separate 2 from 3 when the solvent is present. Here the solvent is water (1), component 2 is felodipine and component 3 is PVP; the felodipine–PVP parameter is ${\chi }_{23}$. E1 The authors take ${\chi }_{13}$ for water–PVP as around 0.5 (Hancock 1993, ref 44) and ${\chi }_{12}$ for water–felodipine as around 3.3, obtained by fitting the group's earlier moisture-sorption data (ref 5) with Flory–Huggins theory as described in ref 44. R22·E3 They print no Δχ value; with these inputs the difference is about 2.8. E1

Methods in detail

The study is a set of experiments on one drug–polymer pair, organized around the IR probe. The flow below shows the stages; the boxes in one row were run in parallel.

Samples

Felodipine (a gift from AstraZeneca) and PVP K29/32 (Sigma–Aldrich) were the only formulation components. M1·M2 For IR, the two were dissolved together in ethanol or a 1:1 (w/w) ethanol–dichloromethane mixture inside a glovebox below 10% RH, checked visually to be a single clear solution, and one or two drops were spin-coated onto a rotating ZnSe window; the film was then heated to 90–100 °C for several minutes to drive off solvent, leaving an optically transparent film. P1 The same procedure on glass slides gave AFM films and on carbon-coated copper grids TEM films, both heated to 90 °C. P3·P4 Bulk powders for DSC and sorption were made by dissolving felodipine with PVP (dried over phosphorus pentoxide for at least a week) in ethanol and removing the solvent on a rotary evaporator, followed by at least 12 h under vacuum. P9 Compositions are given throughout as weight percent PVP in the dry dispersion; the paper used 30, 37, 46, 54 and 74 wt% PVP for the temperature series, 30.3 and 46.4 or 46.5 wt% for most moisture work, 50% for AFM and the remixing test, and 30, 50 and 70 wt% for sorption. A2 The higher composition is labelled 46.4 wt% in the Figure 7 caption and 46.5 wt% elsewhere; the AI writer treats these as one nominal composition. X4

Variable-temperature IR

Spectra were recorded in absorbance mode from 500 to 6000 cm−1 at 4 cm−1 resolution with at least 64 scans, with the spectrometer purged by conditioned air. P5 For the temperature series the film sat in a Specac variable-temperature cell, a background was collected at each temperature, and spectra were taken from 0 °C up to as high as 180 °C in steps of 5–15 °C (Figure 1). Peak positions of the drug–drug and drug–polymer bands were located with a 21-point Savitzky–Golay second derivative; Figure 2 shows the averages of three trials. A 37 wt% PVP film was also heated and then cooled to test reversibility (Figure 3). P6 The DSC glass transition temperatures (Tg) of the 30 and 54 wt% PVP dispersions were compared with the temperature at which the peak-position slope changed. R5

Humidity and water

Films and powders were stored in desiccators over saturated salt solutions giving 33% (magnesium chloride), 43% (potassium carbonate), 58% (sodium bromide), 75% (sodium chloride), 84% (potassium chloride) and 94% RH (potassium nitrate); "0% RH" meant phosphorus pentoxide. Unless stated otherwise, samples were dried under vacuum or under dry nitrogen before analysis. P7 Two kinds of IR experiment followed. In the first, a 30.3 wt% PVP film was kept for 1 week at 33, 58 or 75% RH and measured in a humidity-controlled cell (Figure 4), then dried for 1 week over phosphorus pentoxide and measured again (Figure 5). P8·R7·R8 In the second, 30.3 and 46.4 wt% PVP films were stored for 1 day at 75, 84 or 94% RH, dried, and compared with crystalline and amorphous felodipine (Figure 7). R11 A separate set of films was cast from solvent to which 0.1 mL of water had been added per 1.9 mL, then heated to 90 °C for up to a minute (Figure 8). P2 Water uptake of 10–15 mg bulk samples was weighed in a symmetrical gravimetric analyzer at 25 °C after drying for 6 h at 40 °C under nitrogen and 5 h exposure to 75, 85 and 95% RH (Table 1); Figure 6 shows a full isotherm for 70 wt% PVP. P10·R9·R10

DSC, AFM and TEM

DSC used 4–6 mg of 46.5 wt% PVP powder in open pans, stored for about 10 to 12 h at 75, 84 or 94% RH, dried in a vacuum oven and then purged with nitrogen in the cell for 1 h, or for 17 h in the most rigorously dried sample, before a first and a second heating scan (Figure 9). P11 In a cycling run, a sample kept 1 day at 94% RH and dried 15 h in the DSC was heated in successive cycles to 60, 70, 80, 90, 100 and 110 °C and finally to 175 °C (Figure 10). P12 The heating rate is not printed. X6

AFM was run in alternating-current (tapping-like) mode with silicon probes of about 2 N/m spring constant and tip radius below 10 nm, recording height and phase over 5 µm × 5 µm areas at two to three places per sample. P14 Films of 50% PVP were stored at 94% RH for 1, 2, 4 or 16 h and dried under vacuum for at least 24 h before imaging. P3 TEM at 80 kV with electron diffraction examined amorphous felodipine, PVP and "30% PVP-Felodipine" films stored at 0% or 94% RH and dried overnight. P4·P15 The storage time of the 94% RH TEM sample is not printed. X11

Remixing kinetics

To test whether a phase-separated film can remix, a 50% PVP film was stored for about 4 h at 94% RH, dried with a dry-air purge for 1 h and held at 125 °C in the variable-temperature cell, with spectra taken from 5 to 1500 min (Figure 11). P13·R16 The progress of remixing was expressed as one number: E2

In plain terms: the height of the drug–polymer band ($h_{3290}$) divided by the height of the drug–drug band ($h_{3345}$); it rises as more felodipine molecules bond to PVP. The paper plots it against the natural logarithm of time in seconds and compares it with the ratio of a film of the same composition that was never exposed to moisture (Figure 12). E2·R17

Experimental analysis

Heating weakens the bonds but keeps the dispersion mixed

Figure 1 shows the NH region of five compositions from 0 °C to the highest temperature. In pure amorphous felodipine there is one main band, near 3340 cm−1 in the coldest spectrum and near 3350 cm−1 in the hottest. In every dispersion with 30 to 74 wt% PVP the band near 3290–3300 cm−1 is present in every spectrum, and with more PVP it becomes the main band while the drug–drug band shrinks to a shoulder. R2 The authors' reading is that drug–polymer hydrogen bonding "persist[s] up to the melting point of the drug". C7 The paper does not print felodipine's melting temperature.

Figure 2 quantifies the band positions. Both bands move to higher wavenumber on heating, which the authors read as a gradual weakening of the hydrogen bonds. R3·R4·C6 Table A gives values at selected temperatures.

Table A. NH peak positions (cm−1) at selected temperatures, read from Figure 2 (approximate; averages of three trials). R3·R4

The text gives the end points as 3337 cm−1 at 5 °C to 3350 cm−1 at 160 °C for the drug–drug band of pure amorphous felodipine, and 3288 to 3296 cm−1 between 5 and 160 °C for the drug–polymer band of the 30 wt% PVP dispersion. R3·R4 The second pair matches Figure 2b. The first does not quite: in Figure 2a the pure-felodipine series ends at 120 °C, at about 3346 cm−1, so the value at 160 °C is not shown. X1

A second feature of Figure 2 is a change of slope. For the 30 wt% PVP dispersion both bands bend at about 60 °C, close to its DSC Tg of 57 °C (±5 °C); for the 54 wt% PVP dispersion the bend lies near 90 °C, close to its Tg of 89 °C (±5 °C). R5 The authors conclude that "it appears" both adhesive and cohesive interactions weaken further once Tg is exceeded, as the group had seen for amorphous felodipine alone (Tang 2002, ref 11). C8

The decisive test is reversibility. For a 37 wt% PVP film the heating and cooling points superimpose over 0–160 °C: the drug–drug band runs from about 3344 to about 3351–3353 cm−1 and the drug–polymer band from about 3291.5 to about 3298 cm−1, with no hysteresis (Figure 3). R6 The authors call this "an important result": the effect of heating on the interactions is reversible, and the system is in "equilibrium" with respect to maximizing adhesive and cohesive interactions during the runs. C9 In the Discussion they add that the spectral changes "can be readily accounted for by thermal expansion of the system", that the reversibility indicates the system is well mixed at all temperatures explored, and that neither an upper nor a lower critical solution temperature was seen "over the timescales of the experiments", so this is "a region of miscibility". C34·C35 That passage cites the reversibility as "Figure 8", which shows a different experiment; Figure 3 is meant. X2

Note by the AI writer (not discussed in the paper). At the lowest temperature the drug–drug band sits higher the more PVP the dispersion contains (about 3337, 3340 and 3344 cm−1 at 0, 30 and 54 wt% PVP), and the drug–polymer band likewise (about 3288 and 3290 cm−1 at 30 and 54 wt% PVP). R3·R4 The paper does not comment on this composition effect.

Water uptake rises steeply above about 75% RH

Before looking at what water does to the bonds, the authors measure how much water the dispersions take up. The isotherm of a 70 wt% PVP dispersion at 25 °C (Figure 6) rises gently to about 10.3% at 65% RH and then turns upward: about 15.0% at 75%, 22.9% at 85% and 55.5% at 95% RH, on a dry basis. R9 Table 1, transcribed as Table B, shows the same upturn for all three compositions; the authors state that it held for all dispersions studied. R10·C11

Table B. Water uptake (weight percent of the initial mass) after 5 h at 25 °C; range of duplicate measurements in parentheses (from Table 1). R10

Two details matter later. At 75% RH the 50 wt% PVP dispersion took up slightly less water (4.8%) than the 30 wt% one (5.6%), the only point that does not rise with PVP content. R10·X3 And the method describes only the 5 h steps at 75, 85 and 95% RH; how the lower-RH points of Figure 6 were collected is not stated. X12

Humid storage replaces drug–polymer with drug–drug bonds

Figure 4 shows a 30.3 wt% PVP film after 1 week at 33, 58 and 75% RH, measured in the humidity-controlled cell. A broad water OH band between about 3450 and 3650 cm−1 grows with RH. In the NH region the drug–polymer bands fall relative to the drug–drug bands; the change is small at 33 and 58% RH and "quite distinct" at 75% RH. R7·C10 Because the water band overlaps the NH region, the same films were dried for a week over phosphorus pentoxide; the changes remained (Figure 5). R8

Figure 7 makes the effect larger. After 1 day at 75, 84 or 94% RH and drying, the spectra move toward that of pure amorphous felodipine, without the sharp band of crystalline felodipine near 3362 cm−1. R11·C12 Table C summarizes the IR observations across the humidity experiments.

Table C. IR observations after humid storage at 25 °C (Figures 4, 5 and 7, and text; band positions read from Figure 7, approximate). R7·R8·R11·R12

The authors read the spectra in two steps. First, the number of drug–polymer interactions has been "irreversibly compromised" by moisture, increasingly so at 75% RH and above ("These results suggest"). Second, the rise in drug–drug relative to drug–polymer hydrogen bonding "strongly suggest[s]" amorphous–amorphous phase separation into a drug-rich and a polymer-rich amorphous phase. C13·C14 In the Discussion they grade the extent: separation is "far from complete" at 75 or 84% RH, where drug–polymer interactions persist at a reduced level, whereas after 94% RH the drug's NH band "closely resembles" that of the pure amorphous drug, so the drug-rich phase contains little polymer and the polymer-rich phase "must be" drug poor. C39

In that passage the authors attach water contents to the three humidities, "4.8 and 9.9 wt% water" at 75 and 84% RH and "40.3 wt% water" at 94% RH, "as shown in Table 1". C39 These are the Table 1 values for 50 wt% PVP after 5 h at 75, 85 and 95% RH, not measurements on the 30.3 and 46.4 wt% films stored for 1 day over salts; they indicate an order of magnitude, not the water content of the IR samples. X3·R10

Water in the casting solvent has the same effect

When 0.1 mL of water was added per 1.9 mL of casting solvent, the 30 and 50 wt% PVP films lost most of the drug–polymer band: instead of a maximum near 3300 cm−1 (30 wt%) or 3293 cm−1 (50 wt%) with a second band near 3228 cm−1, they peaked at about 3340 cm−1 with only a weak band near 3230 cm−1 (Figure 8, read approximately). R13 The authors had expected ethanol and dichloromethane to evaporate first during spin coating, concentrating the water; they conclude that water "disrupted the potential for drug–polymer hydrogen bonds to be formed", and since the spectra "strongly resemble" those of dry-cast films after high RH, that the bonds can be disrupted "either by exposure to atmospheric moisture or by the presence of liquid water". C15·C16 Two practical caveats: the panels of Figure 8 carry no a/b labels (the upper pair is read as 30 wt%, the lower as 50 wt%), and the water-cast films were heated more briefly (90 °C, up to a minute) than the dry-cast ones, so some residual water at the time of measurement cannot be excluded. X5

DSC sees a drug-rich phase

All 46.5 wt% PVP powders stored at 75, 84 or 94% RH for about 12 h and dried showed, on the first scan, a "glass transition-like event" at about 45–50 °C, close to the Tg reported for pure felodipine, whereas a single Tg around 80 °C is expected for this composition (Figure 9). R14·R21·M1 The authors take this to suggest a drug-rich amorphous phase. C17 The most rigorously dried sample (94% RH plus 17 h of drying in the DSC) also showed an event with onset at about 124 °C. The authors offer two readings: it "may represent a second glass transition" of a phase that holds much drug from the start or drug that remixed during the scan ("it is likely"), since pure PVP K29/32 has a Tg of about 150 °C; or it may be the depressed melting point of felodipine, although no recrystallization exotherm was seen. R14·C18·M2 On second scans every sample showed one Tg around 80 °C, which the authors read as remixing on heating ("it appears"). R14·C20

The cycling run (Figure 10) supports this. The first cycle showed strong enthalpy recovery, largely absent later; with each cycle to a higher maximum temperature the thermal event moved to a higher temperature, "likely reflecting the remixing"; and the final scan to 175 °C showed an exotherm near 118 °C followed by an endotherm near 130 °C, "probably indicating" crystallization and melting (both temperatures read approximately from Figure 10). R15·C21 The authors caution that the non-isothermal nature of DSC makes these high-temperature events hard to interpret. C19

In the Discussion this becomes "the DSC results which show two Tg events", the lower close to amorphous felodipine. C40 The results section supports a drug-rich phase but not a measured Tg of the polymer-rich phase: the higher event appears in one sample, has an alternative explanation the authors themselves give, and resembles the endotherm they assign to melting in Figure 10. The figure also does not label which curve is the first scan. X7·X6

AFM and TEM see two phases, then crystals

AFM gives a time series on 50% PVP films at 94% RH (Figure 13). Table D lists what the text reports.

Table D. AFM of 50% (w/w) PVP films stored at 94% RH and dried before imaging (from the text and Figure 13). R19

The authors read the 1 h phase contrast as regions that "may have different chemical and/or mechanical properties", the 4 h birefringence as crystals, and the 16 h shapes as "presumably" surface drug crystallization; together, "it appears that the system first undergoes amorphous-amorphous phase separation followed by crystallization of the drug rich regions". C23·C24·C25 TEM agrees qualitatively (Figure 14): amorphous felodipine, PVP and the dispersion stored at 0% RH are featureless, while the dispersion stored at 94% RH shows two distinct regions and no diffraction rings, so it "appears to have undergone" amorphous–amorphous phase separation; longer storage gave diffraction rings. R20·C27 From the micrographs the authors estimate that the dispersed phase forms islands of about 300–500 nm. C41

The timing of crystallization differs between techniques: the AFM film was birefringent after 4 h at 94% RH, while the IR films after 1 day at 94% RH showed no significant crystallization. The authors attribute this "most likely" to the stochastic nature of nucleation, small differences in sample preparation and the different sensitivity of the techniques to surface and bulk crystals. R19·R11·C26 The compositions and substrates also differed (50% PVP on glass for AFM, 30.3 and 46.4 wt% on ZnSe for IR). X10

Only heat brings the phases back together

The dried, phase-separated films did not remix at room temperature: the IR changes persisted through a week of drying (Figure 5) and after the 1 day storage series (Figure 7). R8·R11 At 125 °C they did. In Figure 11 the spectrum of a 50% PVP film that had spent about 4 h at 94% RH starts with the drug–drug band near 3340 cm−1 dominant; over 1500 min the band near 3290–3297 cm−1 grows until the spectrum approaches that of a never-exposed film. R16 Table E gives the peak-height ratio at selected times.

Table E. Peak-height ratio during the hold at 125 °C, read from Figure 12 (approximate); ln time computed from the times in the Figure 11 caption. R17

The ratio rises quickly at first and then slowly, roughly linearly in the logarithm of time. The authors state that extrapolation "indicates that the maximum level of drug–polymer hydrogen bonding will be reached after approximately 39 h". R18·C22 In the Discussion they describe remixing at 125 °C, well above the Tg of felodipine and of the one-phase dispersion (about 80 °C), as occurring "over a period of several hours", and take it to support the "supposition that amorphous felodipine and PVP are miscible in the absence of moisture". C47·R21 The extrapolation method is not given. After 25 h the ratio was still below the reference, and a straight continuation of the last plotted points reaches the reference only after about 2.5–3 days; "several hours" therefore describes the bulk of the change, not its completion. X8

Discussion

Why water splits the pair and heat does not

The authors' explanation runs as follows. Without water, felodipine and PVP are miscible: the drug–polymer bonds persist and recover through heating, the earlier miscibility predictions agree, and the felodipine–PVP parameter ${\chi }_{23}$ is slightly negative, which they say indicates favourable interactions "and confirm[s]" miscibility. C33·C35·C42·R22 The moisture-induced immiscibility therefore "probably" comes from altered thermodynamics once water enters: water interacts extremely unfavourably with the hydrophobic felodipine (${\chi }_{12}$ around 3.3) and much better with the hydrophilic PVP (${\chi }_{13}$ around 0.5), so "there is clearly a large Δχ for this system", the situation in which solvents are known to split compatible polymer blends (refs 39, 40, 46–51). C42·C43·R22·R25 Water "most likely" has a second role as well: it raises molecular mobility, so that molecules can actually diffuse into drug-rich and polymer-rich regions. C44 Once the water is removed, the miscible state is again the lower-free-energy one, but "it appears that the kinetics of mixing at room temperature is prohibitively slow", "presumably" because each separated phase is glassy at room temperature (felodipine Tg around 46 °C, PVP about 150 °C). C45·C46·M1·M2 Only well above the Tg, as at 125 °C, does remixing proceed. C47

Only the end points of this chain were measured: the loss of drug–polymer bonding, the drug-rich phase, the persistence after drying and the recovery at 125 °C. The Δχ argument rests on cited χ values, and no drug–polymer–water phase behaviour was measured. C43

How much water is enough

The extent of separation follows the amount of sorbed water: little change at 33 and 58% RH, a clear change at 75% RH after a week, "far from complete" separation at 75 and 84% RH and a drug-rich phase with little polymer after 94% RH, where the sorption isotherm has turned steeply upward. C10·C11·C39·R9 The abstract condenses this into the statement that "it appears that felodipine-PVP solid dispersions are susceptible to moisture-induced immiscibility when stored at a relative humidity ≥75%", while the dispersions "remained miscible on heating". C51 The 75% RH figure rests on the 1-week experiment with 30.3 wt% PVP; after 1 day at 75% RH both compositions stayed close to their initial spectra. X9·R11

What it suggests for formulation

The authors draw two practical points. First, "the water content of solvents used during the processing operation should be controlled", since the solvent composition can decide whether the dispersion ends up miscible or not. C48 Second, exposure to high RH "may influence the crystallization tendency" by changing the number of metastable phases: in a water-induced drug-rich phase the polymer's stabilizing influence is reduced, and crystallization kinetics are "likely to be very different" from those of a miscible dispersion. C49 They close by saying that more work is needed on which drug–polymer combinations are susceptible to moisture-induced immiscibility and on which analytical techniques best detect it. C50

Note by the AI writer (not discussed in the paper). The IR probe responds within hours to a day at high RH, before crystals are seen in the IR samples, which makes the NH region a sensitive early indicator for this pair. A humidity-exposed dispersion that has been dried at room temperature is not back in its original state; here only storage well above the dispersion's Tg restored it, and slowly. The Δχ argument would predict a similar risk for other hydrophobic drugs in hydrophilic polymers, but this paper tests only felodipine with PVP K29/32.

Limitations

Limits the authors state. They note that the amorphous–amorphous immiscibility was "challenging to characterize experimentally in the presence of the absorbed moisture", which is why spectra were also taken after drying and why several techniques were combined. C38 They limit the temperature result to "the timescales of the experiments and the range of temperatures investigated". C35 They warn that "the nonisothermal nature of the DSC measurement makes interpretation of the high temperature events difficult", and that it is hard to infer the composition of the polymer-rich phase from DSC because remixing is suspected during the scan. C19·C40 They explain the different crystallization times seen by IR and AFM by stochastic nucleation, sample preparation and technique sensitivity. C26

The AI writer's own judgement. The temperature result is solid within its scope: direct, repeated spectra at several compositions, with clean reversibility. R3·R4·R6 The moisture result is consistent across four techniques, but each technique used a different composition, sample form and storage time (IR on 30.3 and 46.4 wt% films, DSC on 46.5 wt% powder, AFM on 50% films, TEM on "30%" films of unstated storage time), so no single sample was followed by all of them. X4·X10·X11 The water contents attached to the IR samples come from a different composition and protocol. X3 The polymer-rich phase is inferred rather than measured; DSC shows the drug-rich phase but no clear second Tg, and the Discussion's "two Tg events" is stronger than the results. X7 The ≥75% RH threshold depends on storage time and composition. X9 Remixing at 125 °C had not reached the never-exposed level after 25 h, and the 39 h extrapolation has no stated method. X8 Two figure–text mismatches are minor: the pure-felodipine series in Figure 2a ends at 120 °C although the text quotes 160 °C, and the reversibility is cited to Figure 8 instead of Figure 3. X1·X2 The error bars of Figure 2 are described only as the "average of three trials". X13 Finally, the mechanism (Δχ, water-enhanced mobility, kinetic trapping in the glass) is the authors' interpretation built on cited parameters, not a result of this paper. C42·C43·C44·C46

Conclusion

The paper separates two environmental stresses that are often lumped together. Temperature, from 0 °C to as high as 180 °C, weakens felodipine's hydrogen bond to PVP but never removes it, and the change reverses on cooling; the authors conclude that the dispersion stays miscible over this whole range. R2·R6·C33 Sorbed water does something qualitatively different: at room temperature it breaks the drug–polymer bonds and splits the glass into a drug-rich and a polymer-rich amorphous phase, an effect that grows with humidity, also appears when water is present in the casting solvent, is followed by crystallization of the drug-rich regions, and is "effectively irreversible at low temperature due to the hindered kinetics of the system". C14·C16·C25·C53 The authors explain it by the very different interactions of water with the drug and the polymer and by slow remixing in the glass, which is why only heating well above Tg brought the phases back together. C43·C46·C47 For formulators the message is that miscibility measured on a dry dispersion does not guarantee miscibility after humid storage or after processing from wet solvents. C48·C49