Carrier or solid form? Probucol solid dispersions with PVP, PAA and PEO made by solvent evaporation and by compression moulding

Background and motivation

A solid dispersion places a poorly water-soluble drug in a hydrophilic carrier to make it dissolve faster. The term covers many different systems (Chiou and Reigelman 1971; Ford 1986). The introduction lists the routes by which a solid dispersion can speed up dissolution: a eutectic, a larger drug surface area after the drug precipitates finely within the carrier, a solid solution, better wetting through intimate contact with a hydrophilic carrier, precipitation of a metastable crystal form, or a lower degree of crystallinity. Which of these operates depends on the drug–carrier pair and on how the dispersion is made (Serajuddin 1999). C5·C6

Two manufacturing routes dominated at the time: co-melting the drug with the carrier, or dissolving both in a common solvent and removing it. The authors note that neither, as traditionally used, is ideal for scale-up, and point to processes from the polymer industry, such as melt extrusion and compression, transfer or injection moulding (Brydson 1995; Tadmoor and Gogos 1979). These "could be imported and tailored" for pharmaceutical production, for example by extruding melts directly into hard gelatin capsules or by compression moulding transdermal patches. C7

PVP and PEG were then among the most studied polymer carriers. The paper sets out three aims (Section 1). First, to characterise solid dispersions of probucol, a cholesterol-lowering drug described as extremely poorly water soluble, with three water-soluble polymers: PVP (polyvinyl pyrrolidone, grade K30), PEO (polyethylene oxide, the same monomer as PEG at a higher molecular weight) and PAA (polyacrylic acid). Second, to see what blending two of these polymers adds. Third, to test compression moulding, a co-melting process run under pressure in a heated mould, against the standard solvent evaporation method. A1·A2·A3

The paper is short and descriptive. Its data are a solid-state characterisation of every sample by Raman spectroscopy and differential scanning calorimetry (DSC), and one release value per sample after 29 h. The interest lies in a result the authors call "somewhat surprising": the amorphous dispersion was not the one that released best. C35

Background concepts

Polymorphs and the amorphous form of probucol. Probucol has at least two crystal forms (Gerber et al. 1993). Form I is the thermodynamically stable form, with an onset melting point of 125 °C; Form II is metastable and melts at 116 °C. The drug as received melted at 125 °C (Table 2 lists an onset of 126 °C), so it is Form I. The authors made Form II by rapid crystallisation from a concentrated ethanol solution, and amorphous probucol by melting the crystals and quench cooling. Amorphous probucol has a glass transition temperature (Tg) of about 26 °C, close to room temperature, and on its own it crystallised within hours to days, depending on how it was handled. M1·P6·P7·C10·R3

Reading the drug's form by Raman spectroscopy. Raman spectroscopy measures molecular vibrations by inelastic light scattering. The three forms of probucol give similar full spectra (Figure 1), because the molecule is the same, but differences in intermolecular packing shift some bands enough to tell the forms apart. The authors used the aromatic C–H stretching region (Figure 2): Form I shows a doublet at 3077 and 3082 cm⁻¹ and a further peak at 3102 cm⁻¹; Form II a doublet at 3083 and 3089 cm⁻¹; amorphous probucol a single broad band at 3083 cm⁻¹; all three share a broad band near 3150 cm⁻¹. None of the three polymers has aromatic groups, and their C–H bands all lie below 3050 cm⁻¹ (PVP 2975 and 2923, PAA 2936, PEO 2940, 2887 and 2840 cm⁻¹; Figure 3), so this window sees the drug alone. Aromatic drugs also scatter much more strongly than most polymers (Hendra 1996). R1·R2·C11·C12

Glass transition as a test of mixing. A homogeneous amorphous mixture of drug and polymer shows one Tg lying between those of the two components; two Tgs, or a Tg equal to one component's, point to separate phases (Bosma et al. 1988). The pure carriers have Tgs of 163 °C (PVP K30) and 132 °C (PAA; Table 2: 131 °C), against about 26 °C for amorphous probucol. PEO is different: it is semi-crystalline, its crystalline regions melt from 62 °C (Table 2: 68 °C), and the Tg of its amorphous regions, expected near −60 °C, lay outside the DSC range used. C14·M2·M3·M4

Hydrogen bonding seen in the PVP carbonyl band. PVP carries a carbonyl group that can accept a hydrogen bond. When a drug donates a hydrogen bond to it, the carbonyl stretching band moves to lower wavenumber and broadens; the authors read the band this way, citing earlier work by Taylor and Zografi (1998), and name the drug hydroxyl group as the donor. Hydrogen bonding to PVP has been linked to PVP's ability to inhibit crystallisation, together with its effect of raising the mixture's Tg (Yoshioka et al. 1995; Taylor and Zografi 1997). C15·C31

Melting point depression. When a crystal melts in the presence of a second component that mixes with its melt, it melts at a lower temperature than the pure crystal. The authors use this to read melting endotherms of probucol below 116 °C in the dispersions as Form II melting in contact with the polymer, rather than as a new polymorph. They also warn that DSC is a heating experiment: a dispersion can change during the scan, which is why they treat Raman, measured at room temperature with little handling, as the more reliable guide to the drug's form. C13·C17·C20

Gelling versus erosion of a hydrophilic carrier. In water a hydrophilic polymer tablet either swells into a coherent gel that dissolves slowly from its surface, or dissolves away (erodes) quickly. Linear PEO is a strongly gelling polymer whose swelling and erosion depend on molecular weight (Graham 1987); high-molecular-weight grades are used for delayed release in hydrophilic matrices. For poorly soluble drugs, release has been described as controlled mainly by the relative rates of polymer swelling and erosion (Kim 1998). Unpublished data from Khoo and Nygren cited here give the scale: PEO of MW 5 million swells up to 700% of its weight, keeps swelling after 20 h and is about 50% eroded after about 22 h; the MW 100 000 grade used in this paper swells only to about 30%, starts eroding noticeably after about 15 min, but keeps eroding over several hours. PVP K30, a linear polymer of low molecular weight (about 40 000), gels little and erodes fast. C38·C39·R17·C36

Compression moulding. A powder mixture is placed in a heated mould, held for a preheating period and then pressed, so that the components melt and fuse into a film, which is then cooled. Here the mould was a J. Wickert & Söhne machine at the Institute of Fibre and Polymer Technology in Mölndal; the powders were preheated for 2 min and pressed at 25 tons for 2 min, at 135 °C for the probucol–PAA and probucol–PEO mixtures and at 165 °C for the samples containing PVP, then cooled in water at about 15 °C. Both temperatures lie above the melting point of probucol Form I (125 °C). P3·M1

Methods in detail

The study is a factorial comparison at one drug loading: five compositions, each made by two routes, characterised in the solid state and then tested for release. The flow below summarises it.

Materials and compositions

Probucol came from Sigma; PVP K30 (linear, MW 40 000) from BASF; PAA (MW 450 000) and PEO (MW 100 000) from Aldrich. Ethanol (99.5%) served as the solvent and polysorbate 80 as the surfactant of the release medium. All dispersions contained 50 wt% probucol (Table 1): three binaries at 50:50 drug:polymer with PVP K30, PAA or PEO, and two ternaries at 50:25:25 with PAA plus PVP K30 or PEO plus PVP. The same compositions were made by both routes. M1–M6·P3·P4

Table A. The ten samples (from Table 1 and Sections 2.3.1–2.3.3). P3·P4·P5·R10·R11

Preparation routes

Solvent method. The physical mixtures were dissolved in ethanol, the ethanol was removed on a rotary evaporator at 45 °C, and the solids were dried for 24 h in a vacuum oven at 45 °C and stored desiccated at 4–8 °C. Two compositions needed exceptions. No common solvent could be found for probucol and PEO, so the binary was made by co-melting the two at 150 °C with stirring. For the PEO–PVP–probucol ternary, PEO was dissolved in a minimum of water and added to an ethanolic solution of PVP and probucol, and a precipitate formed on mixing; the authors flag this. P4·P5·R10·R11

Compression moulding. As described above: 2 min preheating and 2 min at 25 tons, at 135 °C (PAA and PEO binaries) or 165 °C (all PVP-containing samples), followed by cooling in water at about 15 °C and storage at 4–8 °C. P3

Solid-state analysis

FT-Raman spectra were recorded on a PerkinElmer System 2000 with a 1064 nm Nd:YAG laser at typically 800 mW at the sample, collecting back-scattered light; indene was the wavenumber standard (deviations below 1 cm⁻¹). Samples were dried for 24 h under vacuum at 45 °C, measured in rotating NMR tubes to limit laser heating, and every dispersion was analysed in duplicate (two batches). DSC used a Mettler Toledo 820 with about 5 mg in vented aluminium pans under nitrogen, at 10 °C/min for crystalline materials and 20 °C/min for amorphous samples and dispersions, generally in duplicate. P1·P2

Release testing

Tablets of 13 mm diameter were pressed from each dispersion at 10 tons for 1 min, each containing 100 mg probucol. The authors chose tablets so that the starting surface area would be similar and so that the behaviour of the polymer, for example gelling, could be observed. Release was measured in 900 ml of pH 6.8 phosphate buffer containing 3% polysorbate 80 and 10% ethanol, at 37.1 °C, with the JPXII paddle method at 100 rpm. Samples were taken at 1, 2, 4, 6, 24 and 29 h, filtered through 0.45 µm cellulose nitrate membranes and assayed by HPLC (C8 column, UV detection at 242 nm). All samples were run in duplicate; the repeats agreed within the mean ± 2% drug released. P8·P9·P10·C9

The medium is deliberately non-physiological. Surfactant and ethanol were needed because probucol is so insoluble, and only the 29 h value is reported because shorter times did not give detectable drug for every dispersion. The authors state that neither the medium nor the 29 h point is physiologically representative, but both allow the dispersions to be compared. Crystalline and amorphous probucol alone gave no detectable drug even at 29 h. C8·R13·R14

Experimental analysis

Result 1: the carrier decides whether probucol stays amorphous

Table B collects the solid-state results of Table 2. With PVP the drug was amorphous by Raman from both routes, and DSC showed a single glass transition and no melting peak. With PAA or PEO the drug was mainly Form II from both routes. In the ternaries, PVP mostly prevailed: PAA–PVP gave mainly amorphous probucol by both routes, with only a small endotherm near the drug's melting point, and PEO–PVP gave mainly amorphous probucol after moulding. The exception is the PEO–PVP sample from the water–ethanol route, which was mainly Form II. R4·R9

Table B. Solid-state summary of the dispersions (from Table 2). Reference values: Form I melting onset 126 °C, Form II 116 °C, amorphous probucol Tg 26 °C, PVP K30 Tg 163 °C, PAA Tg 131 °C, PEO melting 68 °C (Table 2; the text gives 125, 132 and 62 °C for Form I, PAA and PEO). R4·M1–M4·X3

The authors' reading of this pattern is that the drug's physical state "is very dependent upon the properties of the carrier, even for identical manufacturing processes", and, in the abstract, that the method "was not found to greatly influence" it. Their argument runs as follows. Probucol can be made amorphous by melting and cooling, so they "must assume" it was amorphous after melting with every polymer. PVP then inhibits its crystallisation, "presumably by interacting with the drug on a molecular level to some extent"; PAA and PEO do not, and "presumably" neither forms a single phase with probucol at room temperature. They set this against amorphous probucol on its own, which crystallised at room temperature within hours to days. C1·C2·C29·C30·C33·R3

The PEO–PVP exception is put down to the preparation: not all three components could be dissolved in the same solvent, "which may account for the observed crystallinity". A precipitate was seen when the aqueous PEO solution was added. C23·R11

Result 2: PVP interacts with probucol, less completely after moulding

In both PVP dispersions the carbonyl band of PVP broadened and moved to lower wavenumber compared with pure PVP K30 (Figure 4). The paper prints no band positions; read from Figure 4, the band maximum of pure PVP lies near 1673–1674 cm⁻¹, and the half-height point on the low-wavenumber side moves from about 1664 cm⁻¹ (PVP) to about 1657 (moulded) and about 1651 cm⁻¹ (solvent). The authors call this "consistent with the formation of a hydrogen bonding interaction between the drug hydroxyl group and carbonyl function of PVP", and take it, with the single Tg of 95 °C lying between 26 and 163 °C, as evidence of a one-phase amorphous system in which the two components mixed at the molecular level "to some extent". R5·R4.a·C15·C32·C14

The moulded PVP sample behaved differently in degree. It showed a single, broad glass transition at 129 °C (Figure 5 shows a gradual step spanning about 120–150 °C), well above the 95 °C of the solvent sample, and a smaller carbonyl shift. The authors read this as incomplete mixing during the melt, "perhaps due to the short moulding time allowed": the Tg would then belong to a PVP-rich phase, with an interphase where the two components interact. R4.b·R6·R9·C16

Result 3: with PAA and PEO the drug is Form II, with some amorphous material, and it melts lower

The Raman band of the solvent PAA dispersion follows the Form II doublet, with a broadening on the high-wavenumber side that the authors take as "suggesting the presence of some amorphous material" (Figure 6); the moulded PAA sample gave the same result. The moulded PEO dispersion shows the Form II doublet with no other polymorph and a similar broadening (Figure 7), and the co-melted PEO sample was mainly Form II with a small amorphous component. R7·R8·C18·C19

The DSC melting temperatures of probucol lie below the 116 °C of pure Form II in several samples: 110 °C in the solvent PAA dispersion, 101 °C in the moulded PEO dispersion and 106 °C in the solvent PEO–PVP ternary. The authors explain the PAA value by melting point depression in the presence of PAA ("can be explained by"), and the PEO value in the same way ("We therefore suggest that the DSC results may be due to melting point depression"), since Raman finds no other polymorph. PEO itself melted lower after moulding (58 °C) than after co-melting (65 °C), which "may suggest that some interaction can occur between the two components in the molten state, when pressure is applied". R4.c·R4.f·R4.i·C17·C20·C21

Two details of the DSC data matter for these readings. Table 2 does not say whether its values are onsets or peaks; Figure 5, which shows the moulded binaries, has endotherm minima at about 117 °C (PAA) and about 63 and 109 °C (PEO), a few degrees above the Table 2 values of 116, 58 and 101 °C (the AI writer reads the table values as probable onsets; the paper does not say). And the reference melting points of Form I and Form II were measured at 10 °C/min, the dispersions at 20 °C/min. Depressions of a few degrees should therefore be read with care. R6·X4·X5

Result 4: release at 29 h does not follow the drug's solid form

Figure 8 is the only release figure. Table C gives its values, read from the bars, together with the drug's solid form from Table 2.

Table C. Percent of probucol released after 29 h, mean of two tablets, read from Figure 8 (approximate), with the drug's form from Table 2. Medium: pH 6.8 phosphate buffer with 3% polysorbate 80 and 10% ethanol, 37.1 °C, paddle at 100 rpm. R12·R4·P9

The highest release, about 77%, came from the moulded PEO dispersion, in which the drug is crystalline Form II. The amorphous PVP dispersion released only about 23% (solvent) or 12% (moulded). The authors find it "somewhat surprising" that the amorphous probucol–PVP dispersion released less than the crystalline probucol–PEO system, since an amorphous form generally has a higher apparent solubility and often a faster dissolution rate (Hancock and Zograf 1997). They rank the carriers as: binary and ternary PEO blends highest, then PAA–probucol, PVP–probucol and lastly PAA–PVP–probucol. R12.a·R12.c·C35·C34·C24

Note by the AI writer (not discussed in the paper): this ranking holds for neither method on its own. Among the solvent samples, the PEO–PVP ternary (about 12%) is the lowest of all; among the moulded samples, PAA–PVP, PAA and PVP lie within about 2 percentage points (12–14%), within the replicate spread. The order matches the text only if the moulded PEO samples are set against the solvent samples of the other carriers. X7·X10

During the test, every PEO-containing tablet formed a gel. The PVP- and PAA-containing tablets instead eroded relatively rapidly and left a cloudy dispersion, "presumably composed of probucol particles". PEO was the only strongly gelling carrier, and the authors flag gel behaviour as "a factor to be considered when interpreting the dissolution results". R15·C27·C28

Result 5: moulding helped only where PEO was present, and blends gave no gain

Moulding raised release for the PEO binary (about 48 to 77%) and for the PEO–PVP ternary (about 12 to 50%), but lowered it for PVP (about 23 to 12%), PAA (about 39 to 13%) and PAA–PVP (about 16 to 14%). The authors describe moulding as having "worked significantly well for PEO-containing samples, but for none of the others", note that PEO is the only semi-crystalline carrier, and conclude that "melting methods appear to suit these systems better". R12·C25·C26

Blending gave no advantage over the better parent binary in either route: PAA–PVP released less than PAA from solvent and about as little as either binary after moulding, and PEO–PVP released less than the PEO binary by both routes. The abstract states that blending "was not found to have any advantageous effects in this study"; the summary calls the effect "little". R12.d·R12.e·C4

Note by the AI writer (not discussed in the paper): the PEO–PVP pair is the one composition whose solid form changed with the route, from Form II (solvent) to mainly amorphous (moulded), and it also shows the largest effect of the route on release (about 12 to 50%). It is also the sample whose solvent route produced a precipitate. Figure 8 marks its solvent bar with the asterisk that the caption explains as "prepared by co-melting", whereas the Methods describe it as a water–ethanol solvent preparation. R4.i·R4.j·R12.e·R11·X1

Discussion

How the authors connect the results

The authors' explanation (Sections 4.1–4.3) is a chain of proposals, none of them measured in this paper. It starts from the solid form and ends with what the carrier does in the dissolution medium.

PVP. "It can be speculated that the highly water soluble PVP erodes rapidly from the solid dispersion, leaving behind the hydrophobic drug, which most likely crystallises upon the loss of the crystallisation inhibiting PVP." The erosion was expected for a linear PVP of low molecular weight (about 40 000) that gels little. The drug's form after erosion was not analysed. C36·R15

PEO. Citing the view that the release of poorly soluble drugs follows the relative rates of polymer swelling and erosion (Kim 1998), the authors suggest that the longer contact of the PEO gel with the drug "may thus enhance the wetting of the drug", giving the medium access to the drug surface and preventing aggregation and loss of surface area. "In addition to wetting, some degree of interaction between the PEO and the drug may occur." Ozeki et al. (1997) had explained faster release of flurbiprofen from PEO dispersions by hydrogen bonding and by higher PEO mobility after melting; here the authors "found no evidence of hydrogen bond interactions between drug and polymer", but see the same improvement and "believe the melting of both drug and polymer to be important factors". C39·C40·C41·C42·C43·R16·R18

PAA. PAA held the same Form II as PEO, yet released poorly, so "some other reason must therefore exist to explain this difference of behaviour in PAA and PEO". The observation that the PAA tablets eroded rather than gelled fits the gel argument, but the paper does not test it. C37·R15

Compression moulding. For the moulding advantage with PEO (Section 4.3), "it is suspected" that the controlled high pressure, together with a high local temperature, produces the effect. The authors hypothesise that these give intimate contact and mixing of the molten components, and that on cooling the drug crystallises while staying in intimate contact with the polymer, "possibly in the form of a fine precipitate dispersed in the polymer" (Chiou and Reigelman 1971). The melting point depression of both PEO and probucol in the moulded sample "supports the suggestion" of close contact and interaction in the melt. The method "therefore appears to give an advantage over simple melt or solvent methods, at least for semi-crystalline polymer carriers like PEO". C45·C46·C47

The overall conclusion. In the summary the authors state that the final release properties "appear to depend mainly on the properties of the polymer carrier, rather than on the solid state modification of the drug". They also write that moulding "appeared to have a positive effect on the properties of the solid dispersions" and "provides a viable and practical alternative to existing methods for the large-scale production of solid dispersions". C48·C49

Where the conclusions apply

The data cover one drug at one loading (50 wt%), three carriers of single grades, two 50:25:25 blends, two routes per composition, and one release time point in a medium containing 3% polysorbate 80 and 10% ethanol. Within these limits, the solid-state result (amorphous with PVP, Form II with PAA or PEO) is consistent across routes for the binaries and is read by two independent methods. The release result shows large differences between carriers, well beyond the stated replicate spread, but only at 29 h. R4·R12·P9

Note by the AI writer (not discussed in the paper): two further points limit how far the carrier conclusion can be pushed. First, the drug's solid form is not varied at a fixed carrier: amorphous probucol occurs only in samples containing PVP, so "carrier rather than solid form" is an inference from confounded data, and the one sample pair whose form changed at fixed composition (PEO–PVP) also changed release about fourfold. Second, the moulding advantage with PEO cannot be assigned to pressure: moulding and co-melting of probucol–PEO also differ in temperature (135 against 150 °C), duration, stirring and cooling (water quench against unstated cooling). X13·X11

What it suggests for formulation

Note by the AI writer (not discussed in the paper): for a drug as insoluble as probucol, an amorphous dispersion whose carrier dissolves away quickly may leave the drug stranded, while a gelling carrier can give more release from crystals over a long test. The choice of process interacts with the carrier: moulding helped the semi-crystalline, gelling PEO and hurt PVP and PAA. For development, this argues for testing the carrier's behaviour in water and the release profile over time alongside the solid form, rather than treating an amorphous solid state as sufficient. R12·R15·R4

Limitations

Limitations stated by the authors. The dissolution medium, with surfactant and ethanol, and the 29 h time point are "not physiologically representative" and serve only to compare the dispersions. DSC results "should be interpreted with caution", because dispersions may change on heating and the polymer may depress the drug's melting point; Raman is considered the more reliable indicator. For the PEO–PVP ternary, not all components could be dissolved in one solvent and a precipitate formed. The PVP erosion mechanism is offered as speculation ("It can be speculated"), the moulding mechanism as a suspicion and hypothesis. C8·C13·R11·C36·C45

The AI writer's own judgement (1): the release data. Release is reported as one value per sample at 29 h, although samples were taken at 1, 2, 4, 6 and 24 h; the values are means of two tablets, shown without error bars, and must be read from a bar chart. The word "significantly" used for the moulding advantage is not backed by a statistical test. The large differences (for example about 77% against 12–13% for the moulded binaries) are robust to this; differences of a few percentage points (for example the moulded PVP, PAA and PAA–PVP samples, or PAA–PVP solvent against moulded) are not resolved. X10·R12·R14

The AI writer's own judgement (2): inconsistencies in the paper. Several labels and values disagree within the paper. The Figure 8 caption marks both PEO bars as co-melted, but the Methods describe co-melting only for the binary; Table 2 labels the co-melted PEO sample as "solvent". The text twice refers to Figure 4 for a DSC scan that is in Figure 5. Pure-component values differ between text and Table 2 (Form I 125 against 126 °C; PAA 132 against 131 °C; PEO 62 against 68 °C), and Table 2 does not say whether its values are onsets or peaks. The reference melting points were measured at a different heating rate from the dispersions. None of these changes the main conclusions, but they make the small melting point depressions uncertain. X1·X2·X6·X3·X4·X5

The AI writer's own judgement (3): generalisations in the summary. The abstract's statement that the method "was not found to greatly influence the state of the drug" holds for the binaries but not for PEO–PVP, and the summary's "positive effect" of moulding holds only for PEO-containing samples; for PVP and PAA moulding reduced release by about half and about two-thirds. The claim that moulding is a viable route for large-scale production was not tested here. No drug–PEO hydrogen-bond spectra are shown, and the drug's form after PVP erosion, the crystal size in moulded films and the wetting of the drug were not measured. X8·X9·X14·C49

Conclusion

This early study compared three water-soluble carriers and two blends for 50 wt% probucol, each made by solvent evaporation (or co-melting) and by compression moulding. Raman spectroscopy of the drug's aromatic C–H band, supported by DSC, showed that PVP kept probucol amorphous, with a carbonyl shift read as hydrogen bonding, while PAA and PEO left it mainly as crystalline Form II, largely regardless of the route. Release after 29 h did not follow the solid form: the crystalline PEO dispersions released the most, about 77% after moulding, and the amorphous PVP dispersion the least among the binaries. Moulding helped only where PEO was present, and blends gave no gain. The authors' explanations, PVP eroding and leaving the drug to crystallise, the PEO gel improving wetting, and a fine crystalline precipitate formed under heat and pressure, are proposals that the paper does not test. R1·R4·R5·R12·C36·C40·C45