Does a less water-hungry polymer keep a drug–polymer glass mixed in humid air?

Background and motivation

Many new drug candidates dissolve poorly; the introduction cites an estimate that as many as 40% of drug molecules have a suboptimal aqueous solubility (Lipinski 2002). Making the drug amorphous raises its apparent solubility, dissolution rate and bioavailability (Hancock and Zografi 1997; Hancock and Parks 2000), but an amorphous drug can crystallize during manufacture and storage and lose that advantage. A second component, usually a water-soluble polymer, is therefore added to inhibit crystallization, and the result is called an amorphous solid dispersion (ASD) (Ford 1986).

The polymer has a cost: it is hygroscopic, so an ASD takes up more water from humid air than the amorphous drug alone. Water raises molecular mobility and can promote crystallization (Shamblin and Zografi 1999; Saleki-Gerhardt and Zografi 1994; Andronis and Zografi 1998). Work from the Taylor group had shown a further effect. In some molecularly mixed dispersions of a hydrophobic drug with the hydrophilic polymer poly(vinylpyrrolidone) (PVP), absorbed water produced separate drug-rich and polymer-rich amorphous domains before any crystals appeared. This moisture-induced amorphous–amorphous phase separation had been reported for felodipine–PVP (Konno and Taylor 2008; Marsac et al. 2010), nifedipine–PVP, pimozide–PVP and droperidol–PVP (Rumondor et al. 2009) and griseofulvin–PVP (Vasanthavada et al. 2005), but not for indomethacin–PVP or ketoprofen–PVP. The systems that separated also crystallized much sooner (Rumondor et al. 2009). The authors therefore suggest that moisture-induced changes in miscibility "may be an underlying cause" of the poorer stability of some dispersions under accelerated stability conditions, and that it matters to know which systems are susceptible.

Two factors were already on the table. The strength of drug–polymer interactions "seems to be important" in deciding whether water causes unmixing (Rumondor et al. 2009), and the same earlier study had found that raising mobility by water plasticization was not enough on its own to unmix PVP dispersions, so the thermodynamics of the system had to be considered. Polymers also differ in how much water they take up and in the interactions they form with a drug (Konno and Taylor 2006), so the same drug at the same ratio may hold different amounts of water with different polymers (Konno and Taylor 2008), with different effects on mobility and thermodynamics (Marsac et al. 2008).

The paper tests one hypothesis: dispersions prepared with less hygroscopic polymers have a decreased susceptibility to moisture-induced changes in miscibility. Three polymers of falling hygroscopicity were chosen: PVP (the most hygroscopic), poly(vinylpyrrolidone-co-vinyl acetate) (PVPVA, intermediate) and hypromellose acetate succinate (HPMCAS, the least hygroscopic). They were combined with four drugs chosen to cover different chemistries and possible interactions: felodipine, quinidine, pimozide and indomethacin (Figure 1). The main tool is mid-infrared (IR) spectroscopy, supported by differential scanning calorimetry (DSC).

Background concepts

Amorphous–amorphous phase separation. A freshly made ASD is ideally a single amorphous phase in which drug and polymer are mixed at the molecular level. If water makes this mixture unstable, it can split into two amorphous phases, one richer in drug and one richer in polymer, without any crystals forming. The paper also calls this drug–polymer demixing or immiscibility. It matters because a drug-rich amorphous domain behaves more like the pure amorphous drug and is more likely to crystallize.

Hygroscopicity. The amount of water a solid takes up from humid air at a given relative humidity (RH) is described by its water sorption isotherm. Figure 11 shows isotherms of the three polymers at 25 °C, taken from earlier work: at about 85% RH, PVP holds about 50 g water per 100 g dry polymer, PVPVA about 24 g and HPMCAS about 8 g (values read from the plot, approximate). At the highest RH plotted, about 95%, PVP reaches about 75 g per 100 g, PVPVA about 33 g and HPMCAS about 11 g.

How IR reads mixing. The frequency at which a chemical group absorbs infrared light depends on its surroundings. When a hydrogen-bond donor on the drug (an NH, OH or COOH group) bonds to an acceptor on the polymer (for example the carbonyl of the vinylpyrrolidone ring), the band of the bonded group moves, usually to a lower wavenumber (a red shift), and a new band or shoulder appears. In a molecularly mixed dispersion these drug–polymer bands are present, so its spectrum differs from that of a physical mixture of amorphous drug and polymer. If water later breaks drug–polymer bonds and the drug molecules bond to each other again, the drug–polymer band loses intensity relative to drug–drug or free bands. The authors set out two expected cases (Section "Miscibility of Solid Dispersions Following Exposure to Moisture"). Severe demixing would give a nearly pure drug phase and a nearly pure polymer phase, and the spectrum would approach that of a physical mixture. Partial demixing would give a drug-rich and a polymer-rich phase that each still hold much of the other component, and the spectrum would look like a combination of two dispersions of different drug:polymer ratio. Table 1 lists the band assignments used throughout.

How DSC reads mixing. An amorphous solid shows a glass transition: a step in heat flow at the glass transition temperature, Tg. A molecularly mixed two-component glass has one Tg, usually between the Tgs of its components. Two amorphous phases of different composition give two Tgs. For severe demixing the two Tgs would lie close to those of the pure drug and pure polymer (assuming the phases do not remix while the sample is heated in the DSC); for partial demixing they would lie between them. A glass that has been stored can also show an endothermic peak at its Tg on reheating, called enthalpic recovery: it reflects relaxation of the glass during storage, not a second phase.

Ternary Flory–Huggins theory. To discuss why some systems separate and others do not, the authors use the Flory–Huggins expression for the free energy of mixing of a three-component system of water (W), drug (D) and polymer (P), their Eq. 1:

Here $\Delta G_{\mathrm{m}\mathrm{i}\mathrm{x}}$ is the free energy of mixing, $R$ the gas constant, $T$ the temperature in kelvin, $n_i$ and ${\varphi }_i$ the amount (the paper says "mole … fractions") and the volume fraction of component $i$, and ${\chi }_{ij}$ the Flory–Huggins interaction parameter of the pair $ij$. In plain terms: the first three terms are the combinatorial entropy of mixing, which always favours one phase; the last three terms carry the enthalpy of mixing, set by the three pair parameters. A positive $\chi$ means the pair dislikes mixing; a strong drug–polymer hydrogen bond gives a low (favourable) ${\chi }_{DP}$. Because the water–drug term is multiplied by $n_W$, the more water the dispersion holds, the more an unfavourable water–drug interaction weighs. The authors state that one phase is expected when $\Delta G_{\mathrm{m}\mathrm{i}\mathrm{x}}$ is negative with a single minimum over composition, and, assuming that adding water favours the mixing entropy, that amorphous–amorphous separation needs an enthalpy of mixing positive enough to outweigh that entropy. The paper uses this equation only qualitatively; no $\chi$ value is measured or calculated.

Methods in detail

The study has a simple design: the same dispersions are examined dry, then after humid storage, and then dried again before analysis, so that any loss of mixing appears as changed IR bands or a second Tg. Water uptake and earlier sorption data supply the link to polymer hygroscopicity, and the Flory–Huggins equation supplies the interpretation.

Materials and the nine systems

Felodipine was a gift from AstraZeneca; quinidine, pimozide, indomethacin and PVP K29-32 came from Sigma-Aldrich; PVPVA (Kollidon VA 64) was a gift from BASF; HPMCAS was grade AQOAT AS-MF from ShinEtsu. Table 2 gives the drug properties used later in the argument: calculated log P of 2.24 (felodipine), 2.48 (quinidine), 5.34 (pimozide) and 3.58 (indomethacin), and Tg midpoints of 47, 61, 62 and 48 °C. The polymer Tgs are 166 °C (PVP), 107 °C (PVPVA) and 121 °C (HPMCAS). The four drugs offer different hydrogen-bond donors: an NH group in felodipine and pimozide, an OH group in quinidine (which has no carbonyl, so its interaction can be read on the polymer carbonyl) and a COOH group in indomethacin. Nine drug–polymer pairs were studied at 50 wt% polymer by both IR and DSC (Table 3): felodipine with PVPVA and HPMCAS, quinidine with all three polymers, and pimozide and indomethacin with PVPVA and HPMCAS. Felodipine–PVP appears only as one IR spectrum after storage at 84% RH (Figure 2a) and through a cited water uptake (Table 4); pimozide and indomethacin were not combined with PVP in this paper.

Films for IR and humid storage

Drug and polymer were dissolved together at several weight ratios in a common solvent: a 1:1 w/w mixture of dichloromethane and ethanol for the felodipine and quinidine systems and for indomethacin–HPMCAS, ethanol for indomethacin–PVPVA, chloroform for pimozide–PVPVA, and a 1:1 v/v mixture of methanol and chloroform for pimozide–HPMCAS. One or two drops were spin-coated on ZnS or KRS-5 windows (about 500/2500 rpm for 18 and 30 s, adjusted so that the absorbance in the region of interest lay between 0.4 and 1.0) and dried on a hot plate at 90 °C. Spectra were recorded in transmission from 500 to 4000 cm⁻¹ at 4 cm⁻¹ resolution with 128 scans, under a dry-air purge.

After the first spectrum, the films were placed in desiccators over saturated salt solutions: Mg(NO₃)₂ for 54% RH, NaCl for 75%, KCl for 84% and KNO₃ for 94%, at room temperature (22–25 °C). After at least one day at each RH, the films were dried with dry air (RH below 10%) until the OH region of the spectrum stopped changing, and measured again. The same films went through the RH series in turn. The figure captions show that most exposures were one day, but indomethacin–PVPVA spent 7 days at 94% RH (Figure 7) and felodipine–HPMCAS 2 days (Figure 8). Drying before each spectrum was deliberate: absorbed water would otherwise mask changes in the drug–polymer bands. The consequence is that every spectrum describes the dried film, not the film while it held water.

DSC before and after humid storage

Bulk powders were made by rotary evaporation from the same solvents, dried under vacuum for at least 12 h and ground in a dry glovebox. For each dispersion, two identical pans were prepared. The first was measured at once; the second was held at 25 °C and 94% RH for 24 h in a gravimetric analyzer, then dried with air below 3% RH until its weight stopped falling (about 48 h), sealed hermetically and measured with the same program. The program heated at 20 °C/min to 10 °C above the drug melting point or the polymer Tg (whichever was higher), held for 60 s, cooled to −40 °C and reheated, and the cycle was repeated three times. The paper does not say which heating scan gave the reported Tg values.

Water uptake

Water sorption at 25 °C and 85% RH was measured gravimetrically on 10–15 mg of four dispersions containing 25% polymer (felodipine–PVPVA, quinidine–PVP, quinidine–PVPVA and quinidine–HPMCAS), after 3 h of drying, and taken at the first plateau of the weight-gain curve. The values for felodipine–PVP and felodipine–HPMCAS come from a PhD thesis (ref 33). Note by the AI writer (not discussed in the paper): these uptakes refer to 75:25 drug:polymer powders at 85% RH, whereas the phase behaviour was tested on 50:50 films and powders at 54–94% RH, so they rank the polymers rather than give the water content of the tested samples.

Experimental analysis

All nine dispersions start as one phase

Immediately after preparation, every dry dispersion showed IR features of drug–polymer interaction that the pure components lack (Table 1). With quinidine, a shoulder at 1662 cm⁻¹ appeared beside the free PVP carbonyl at 1681 cm⁻¹; with PVPVA the corresponding band sat at 1665 cm⁻¹. Indomethacin shifted the vinylpyrrolidone carbonyl of PVPVA to 1640 cm⁻¹, pimozide to 1661 cm⁻¹. In felodipine–PVPVA the felodipine NH band moved from 3341 cm⁻¹ (bonded to another felodipine) to 3297 cm⁻¹ (bonded to PVPVA). DSC gave a single Tg for all nine pairs, between the Tgs of drug and polymer (Table A). The authors take the two results together to "support the conclusion" that all systems were molecularly mixed when dry, which for felodipine–HPMCAS confirms earlier work.

Table A. Glass transitions before and after 24 h at 25 °C/94% RH followed by drying (from Table 3; ΔCp after storage omitted; no replicates or uncertainties are given in the paper; card R3).

Five dispersions split after humid storage

For five systems the IR spectra changed with storage RH and DSC found two Tgs after 94% RH (Figures 2–6). The authors' reading is that storage "resulted in the formation of drug-rich and polymer-rich amorphous phases" ("The results suggest").

Felodipine–PVPVA (Figure 2). The NH band of felodipine bonded to another felodipine (3341 cm⁻¹) grew relative to the band bonded to PVPVA (3297 cm⁻¹) after 54% and 84% RH, which the authors read as drug–drug bonds forming "at the expense of" drug–polymer bonds. After 94% RH the single Tg of 77 °C became two, 64 and 77 °C. In the discussion the authors quantify the loss at 84% RH: a 17% reduction in the drug–polymer band relative to the drug–drug band, against 50% for felodipine–PVP at the same RH. Note by the AI writer (not discussed in the paper): only the post-84% RH spectrum of felodipine–PVP is shown, and the x-axis of Figure 2a is labelled "Normalized absorbance (cm⁻¹)" where wavenumber is meant.

Quinidine–PVP (Figure 3). The 1662 cm⁻¹ shoulder (PVP carbonyl bonded to the OH of quinidine) fell after 54% RH and was "virtually eliminated" after 84% RH, a 47% reduction relative to the free carbonyl. Its Tg of 70 °C became 63 and 81 °C.

Quinidine–PVPVA (Figure 4). The corresponding band at 1665 cm⁻¹ fell only slightly after 84% and 94% RH, an 11% reduction at 84% RH; the authors note ("Interestingly") that this loss is much smaller than with PVP. DSC still found two Tgs, 62 and 76 °C, against 69 °C before.

Pimozide–PVPVA (Figure 5). Two carbonyl features come from drug–polymer interaction: the weak shoulder at 1661 cm⁻¹ (vinylpyrrolidone carbonyl bonded to the drug) and the free pimozide carbonyl at 1709 cm⁻¹, which is visible only when drug–drug hydrogen bonds are broken. Both weakened after 54% and 84% RH. DSC "clearly indicated" two Tgs, 59 and 97 °C (83 °C before), and on further heating the trace shows an exotherm near 170 °C and an endotherm near 200 °C (read from Figure 5b, approximate), which the authors identify as recrystallization and melting of pimozide. The drug-rich Tg of 59 °C lies slightly below the 62 °C of amorphous pimozide itself (Table 2); the paper does not comment on this.

Pimozide–HPMCAS (Figure 6). Here the spectra are harder to read because bands overlap. In the dry films a weak shoulder at 1711 cm⁻¹ at 10% polymer grew at 30% and became a broad band at 1704 cm⁻¹ at 50% polymer (spectra not shown); this band is free pimozide carbonyl overlapping the drug–drug band at 1696 cm⁻¹. With increasing RH it split into a peak at 1709 cm⁻¹ and a shoulder at 1698 cm⁻¹. The shoulder signals more drug–drug bonding, but it stayed weaker than the 1709 cm⁻¹ peak, so drug–polymer bonding remained significant. The authors read this as drug-rich and polymer-rich regions that each still hold "significant amounts of the other component", a conclusion "better supported by the DSC results": two Tgs, 65 and 79 °C, close to each other and to the dry Tg of 78 °C.

For these five systems, similar trends "to differing extents" were seen at 30 and 70 wt% polymer, and the absence of crystalline drug was checked visually and in other spectral regions. At 10 and 90 wt% polymer, demixing "could not be completely verified nor disqualified" because of the sensitivity of the methods.

Four dispersions stayed one phase up to 94% RH

For felodipine–HPMCAS, indomethacin–PVPVA, indomethacin–HPMCAS and quinidine–HPMCAS, storage at RHs up to 94% left the relevant IR bands unchanged (Figures 7–10). In indomethacin–PVPVA, the band at 1640 cm⁻¹ (vinylpyrrolidone carbonyl bonded to indomethacin) and the band at 1683 cm⁻¹ kept the same relative intensity, even after 7 days at 94% RH. In felodipine–HPMCAS, the free (1701 cm⁻¹) and drug–drug bonded (1683 cm⁻¹) felodipine carbonyl bands kept their ratio. In indomethacin–HPMCAS, the cyclic-dimer band of indomethacin at 1712 cm⁻¹ was already weaker relative to the benzoyl band at 1685 cm⁻¹ than in pure amorphous indomethacin, a sign that the polymer had broken up drug–drug bonds, and this ratio did not change after storage. In quinidine–HPMCAS, the acid shoulder of HPMCAS at 1714 cm⁻¹ was absent in the dispersion (checked by second-derivative spectra), read as quinidine breaking polymer–polymer bonds, and it did not return after 94% RH.

DSC after 94% RH found a single Tg for each, at about the dry value (Table A), accompanied by an endothermic peak that the authors attribute to enthalpic recovery after enthalpy relaxation. Their conclusion is that for these systems "exposure to moisture during storage at RHs as high as 94% did not result in disruption of drug–polymer interactions".

Note by the AI writer (not discussed in the paper): two printed details in this part need care. Table 1 assigns the 1640 cm⁻¹ band of indomethacin–PVPVA to "PVP" and the 1683 cm⁻¹ band to a non-bonded acid carbonyl, while the text assigns 1683 cm⁻¹ to the free vinylpyrrolidone carbonyl overlapping the indomethacin benzoyl carbonyl; the text reading fits the spectra. In addition, the amorphous felodipine trace at the top of Figures 2b and 8b steps at about 18–30 °C (read from the plots), well below the 47 °C given for felodipine in Table 2; the other drugs' traces agree with their table values within a few degrees.

Water uptake ranks the polymers

Table B. Water sorbed at 25 °C/85% RH by 75:25 drug:polymer dispersions (from Table 4; card R18) and by the polymers alone at 25 °C (read from Figure 11, data of refs 9 and 34; card R19).

For both drugs the order is PVP > PVPVA > HPMCAS (Table B). The dispersions, three quarters drug by weight, sorb far less water per gram than the polymers alone, but the order is the same.

PVP versus PVPVA, and the role of the hydrogen bond

The two drugs tested with both PVP and PVPVA lost more of their drug–polymer band with PVP at 84% RH: 50% versus 17% for felodipine and 47% versus 11% for quinidine. The authors argue that the hydrogen bond itself is about as strong with either polymer, judged by similar red shifts, and that the smaller demixing with PVPVA "is most likely explained by the reduction in the amount of water sorbed". For quinidine, the Table 1 positions agree (calculation by the AI writer): the bonded carbonyl sits 19 cm⁻¹ below the free band with PVP (1681 → 1662 cm⁻¹) and 18 cm⁻¹ below it with PVPVA (1683 → 1665 cm⁻¹). For felodipine the shifts cannot be checked, because Table 1 lists no carbonyl band of felodipine bonded to PVPVA and no felodipine–PVP band at all.

Table C. Hydrogen-bond strength and outcome for the PVPVA dispersions. Free vinylpyrrolidone carbonyl of PVPVA: 1683 cm⁻¹ (Table 1; card R1, R20). From the tabulated positions the quinidine shift is 18 cm⁻¹, just below the stated range (card X5).

Among the PVPVA dispersions, indomethacin stands out (Table C). Its COOH group is a better hydrogen-bond donor than the NH of felodipine and pimozide or the OH of quinidine, and the vinylpyrrolidone carbonyl shifts by 42–44 cm⁻¹ against 19–22 cm⁻¹ for the others. Indomethacin–PVPVA is also the only PVPVA dispersion that did not split, matching the earlier finding that indomethacin–PVP did not split either. The authors call the stronger interaction a "supposition" that is "supported" by the red shift; stronger interactions "would result in lower values for χDP" and a more favourable enthalpy of mixing.

HPMCAS: weaker bonds but no splitting, except for pimozide

HPMCAS breaks the simple rule "strong bond, no splitting". Earlier work (Konno and Taylor 2006) found felodipine–HPMCAS interactions weaker than those with PVP or PVPVA, yet felodipine–HPMCAS did not split; the same holds for quinidine. So, the authors write, "consideration of only the drug−polymer interaction cannot explain the absence" of demixing in these systems. Pimozide is the exception: with an NH donor its interaction is "expected" to be similar to felodipine's, yet pimozide–HPMCAS showed slight demixing. Pimozide is also by far the most hydrophobic of the four drugs (calculated log P 5.34 against 2.24–3.58).

Discussion

The authors' explanation. The discussion pulls the results together with the ternary Flory–Huggins equation. All four drugs are hydrophobic (positive log P), so the water–drug parameter ${\chi }_{WD}$ is "expected to be positive", as had been calculated for indomethacin (Crowley and Zografi 2002). How much this unfavourable term weighs depends on ${\chi }_{WD}$ and on the amount of water sorbed, $n_W$, which in turn depends on the hygroscopicity of the matrix. A hygroscopic polymer therefore raises the positive contribution to the enthalpy of mixing; a strong drug–polymer bond lowers ${\chi }_{DP}$ and pushes the other way. The chain can be drawn as follows.

In this picture, HPMCAS dispersions "simply do not absorb sufficient water to create an unfavorable environment for the hydrophobic drug", which "may explain" why felodipine and quinidine demixed with PVP and PVPVA but not with HPMCAS. Pimozide's hydrophobicity "would lead to larger χWD values", favouring separation "even at low amounts of moisture", which is how the authors account for the single HPMCAS exception. The conclusions put the whole argument in two sentences: "It is suggested that the balance between the thermodynamic factors (enthalpy and entropy of mixing)" decides which systems are susceptible, and these factors depend on the amount of water sorbed, which depends on the hygroscopicity of drug and polymer and on the strength of drug–polymer interactions. In the abstract the same statement is introduced with "It is concluded".

A second explanation the authors keep open. Less water also means less plasticization. The authors write that "it is possible" that HPMCAS dispersions did not demix because of smaller plasticizing effects, that is, lower molecular mobility. As an example they cite Tg values after storage at 75% RH for dispersions with 25% polymer: about 35 °C for felodipine–HPMCAS and about 31 °C for felodipine–PVP (Konno and Taylor 2008). The data in this paper do not separate the two explanations, because every sample was dried before it was measured. The conclusions also combine them: water disrupted drug–polymer interactions "which, coupled with the ability of water to increase molecular mobility, led to phase separation".

Where the approach applies. The study covers four drugs, three polymers, mostly 50 wt% polymer and storage at room temperature up to 94% RH. Within that range the authors generalize in the abstract: systems with strong drug–polymer interactions and a less hygroscopic polymer "will be less susceptible to moisture-induced phase separation", while more hydrophobic drugs will be more susceptible even at low levels of sorbed moisture.

What it suggests for formulation. Note by the AI writer (not discussed in the paper): for a hydrophobic drug that forms only ordinary hydrogen bonds with the polymer, these results favour a low-hygroscopicity polymer such as HPMCAS when the product may meet high humidity, and they suggest that a strong donor–acceptor pair (here a COOH donor to the vinylpyrrolidone carbonyl) can protect even with a more hygroscopic polymer. A very hydrophobic drug stays at risk with any of the three polymers. The test used here, storing thin films at high RH, drying and comparing IR spectra and Tgs, is quick, but it reports the dried state.

Limitations

Limits the authors state. At 10 and 90 wt% polymer, demixing "could not be completely verified nor disqualified due to the sensitivity limitations of the methods used". For pimozide–HPMCAS, "specific locations and features of the peaks are difficult to discern due to relatively small changes as well as significant overlaps". The thermodynamic model is used "to help qualitatively understand" the behaviour, and the mobility explanation for HPMCAS is left as a possibility.

The AI writer's own judgement (1): what was measured and what was inferred. The phase behaviour and the water uptake were measured on different samples under different conditions: 50:50 films (54–94% RH) and 50:50 powders (94% RH) for phase behaviour, 75:25 powders at 85% RH for water uptake, two of whose six values come from a thesis. No uptake was measured for the pimozide or indomethacin systems. The hygroscopicity explanation is therefore an inference from a consistent ranking, not a measured water content of the samples that split. Likewise, no $\chi$ value is estimated: ${\chi }_{DP}$ is ranked by red shift and ${\chi }_{WD}$ by calculated log P. The Flory–Huggins reading is plausible and internally consistent, but qualitative.

The AI writer's own judgement (2): how firm the individual comparisons are. No replicates or uncertainties are reported for any Tg, ΔCp, water uptake or intensity reduction, and the method behind the four percentage reductions (peak heights or areas, baseline, which band pairs) is not described. The felodipine–PVP figure of 50% rests on a single spectrum after 84% RH. The IR films went through 54, 75, 84 and 94% RH in sequence, so a later spectrum carries the history of the earlier exposures, and holding times at 94% RH differ between systems (1, 2 and 7 days); the DSC powders saw only 94% RH. Separation "at RHs as low as 54%" therefore rests on IR alone. The DSC scan that gave each Tg is not stated, although samples were heated above the drug melting point and cycled three times, which could remix phases after the first scan. The ranking of HPMCAS as least affected also rests on the IR spectra: by DSC alone, the pimozide–HPMCAS Tg pair (65 and 79 °C) is as far apart as the felodipine–PVPVA (64 and 77 °C) and quinidine–PVPVA (62 and 76 °C) pairs.

The AI writer's own judgement (3): printing errors. Several printed details disagree with the text or with each other, none of which changes the conclusions: Table 1 assigns the 1661 cm⁻¹ shoulder of pimozide–PVPVA to a carbonyl "H-bonded to quinidine" (the text says the drug, i.e. pimozide); the indomethacin–PVPVA assignments in Table 1 differ from the text, as noted above; the x-axis of Figure 2a is labelled as absorbance; the Figure 3a caption places the quinidine bands at 1640–1660 cm⁻¹ while they are plotted at about 1620 and 1590 cm⁻¹; and the plotted amorphous felodipine trace does not match its tabulated Tg of 47 °C.

Conclusion

Rumondor and Taylor extended the earlier finding that humid air can unmix drug–PVP glasses to two less hygroscopic polymers. With the same drug, PVPVA dispersions unmixed less than PVP dispersions, and HPMCAS dispersions did not unmix at all except with pimozide, the most hydrophobic drug. Water uptake followed the same order, PVP > PVPVA > HPMCAS. Indomethacin, whose COOH group forms the strongest hydrogen bond with PVPVA, kept PVPVA mixed even after 7 days at 94% RH. The authors read these results through the ternary Flory–Huggins equation: the amount of water sorbed, the drug–polymer interaction and the drug's hydrophobicity together set whether the enthalpy of mixing turns unfavourable enough to split the glass. The pattern is clear across nine systems; the thermodynamic explanation is qualitative and was not tested with measured water contents or interaction parameters, and every sample was examined after drying.