Miscibility and drug-polymer interactions

How it was shown

Taylor and Zografi first fixed the reference state, because earlier dispersion spectra had been compared with crystalline drug and crystallinity alone changes vibrational spectra taylor1997 C7. They recorded γ, α and amorphous indomethacin by FT-Raman (Bruker RFS 100, 1064 nm, typically 500 mW, 200 scans at 4 cm⁻¹) and by transmission IR (Mattson Galaxy 5020, solids at 0.8–1% in KBr, 500 scans at 4 cm⁻¹) taylor1997 P7, P8. Amorphous drug was made by melting γ at 165 °C for 5 min and quench cooling in liquid nitrogen, α by precipitation from methanol with water taylor1997 P1, P2. A cyclic carboxylic acid dimer has a centre of symmetry, so by the mutual exclusion rule its asymmetric C=O stretch is seen only in the IR and the symmetric stretch only in the Raman, which the authors checked on liquid acetic acid (IR 1716, Raman 1669 cm⁻¹) taylor1997 C14, R5.j. Dispersions were made by dissolving 5 g of drug and PVP-K90 in 100 ml of anhydrous methanol at 65 °C, rotary evaporation at 50 °C and 24 h under vacuum, at 5, 10, 20, 30, 40, 50, 60 and 70 wt% PVP; physical mixes of amorphous drug and PVP were prepared at the same compositions, by a procedure that is not printed taylor1997 P3, P4, X7. Acetic acid with 10–90 wt% 1-methyl-2-pyrrolidone and indomethacin in 30–70 wt% methylpyrrolidone served as liquid models for the acid–lactam contact taylor1997 P5, M5.

What the data show

Amorphous indomethacin has an IR acid band at 1710 cm⁻¹ with no Raman counterpart, like the 1717 cm⁻¹ band of γ, which is known from its crystal structure to hold cyclic dimers; the authors conclude it is "considered to be composed mainly" of cyclic dimers, with a shoulder at 1735 cm⁻¹ read as free chain ends taylor1997 R5.a, R5.g, R5.h, R6, C19, C20. α-indomethacin has acid bands at 1681 (shoulder) and 1649 cm⁻¹ in the IR that appear at 1680 and 1649 cm⁻¹ in the Raman, which they read as chains, not dimers taylor1997 R5.e, R5.f, C16, C17. In the physical mixes the amorphous doublet near 1710 and 1684 cm⁻¹ persisted at every composition, with the PVP band (1679 cm⁻¹) growing on top taylor1997 R10, R5.l. In the dispersions the dimer peak lost intensity between 5 and 20 wt% PVP while the non-bonded carbonyl peak gained; from 20 wt% there was one broad peak at 1726 cm⁻¹, and a band at 1636 cm⁻¹ was present at 5–50 wt% PVP and merged into a shoulder at 60–70 wt% taylor1997 R11, R12. The shift of the PVP carbonyl from 1679 to 1636 cm⁻¹ is 43 cm⁻¹ (synthesis: 1679 − 1636) taylor1997 R5.l, R12. The PVP C–N stretch did not change in IR or Raman, which points to the carbonyl as acceptor taylor1997 R13, C25. In acetic acid with methylpyrrolidone the amide C=O first appeared at 1652 cm⁻¹, 36 cm⁻¹ below its pure-liquid position, and the free-acid shoulder at 1759 cm⁻¹ grew; indomethacin in methylpyrrolidone lost its 1710 cm⁻¹ band and showed one band at 1723 cm⁻¹ and a bonded amide band at 1649 cm⁻¹ taylor1997 R15, R16. With a monomer molecular weight of 111, a 1:1 mole ratio of PVP monomer to drug is reached at 24 wt% PVP, and the paper states 12 wt% for two hydrogen bonds per carbonyl, which the card recomputes as 13.6 wt% (synthesis: half a monomer per drug molecule gives 0.5/(0.5 + 3.17)) taylor1997 E1, R18, X2. The changes continue above both values taylor1997 R11, R12.

Figure

Left, the physical mixes keep the amorphous doublet at every PVP content; right, the dispersions lose the dimer band and gain the bands near 1726 and 1636 cm⁻¹. Percentages are wt% PVP; 0% is amorphous indomethacin and 100% pure PVP taylor1997 R10, R11, R12.

Why

The authors' reasoning, marked as their interpretation: indomethacin has a single donor, its acid OH, and PVP can only accept a proton, so a new bond must show in the acid C=O and in the PVP carbonyl or C–N bands taylor1997 C25. The more basic PVP amide "would explain" why the acid OH bonds to the PVP carbonyl in preference to forming the acid dimer taylor1997 C36. The spectra "support the fact" that the bond forms at the expense of dimers, and the physical mixes show none, so the bond needs molecular mixing taylor1997 C35, C26. The 1726 cm⁻¹ band is assigned to the non-bonded acid C=O of a PVP-bonded molecule and the 1636 cm⁻¹ band to hydrogen-bonded PVP carbonyl; the liquid models reproduce the changes and are used for the assignment taylor1997 C32, C33. Since spectral change continues above the stoichiometric composition, "this suggests" a low equilibrium constant and/or a stoichiometry below 1:1, so dimers, PVP-bonded drug and free PVP coexist, and these species are not taken to be separate phases, on the strength of a single Tg reported earlier taylor1997 C37, C38, C39. Non-ideal mixing or frozen-in concentration fluctuations are not excluded taylor1997 C40. The authors then propose that a γ nucleus, built of dimers, "would require" breaking the PVP–drug bond, so that the bond "offers an explanation" of why PVP inhibits crystallization at levels where its antiplasticizing effect is minimal taylor1997 C43, C45. What was not measured: crystallization, nucleation or stability in these dispersions taylor1997 X8, band intensities, areas or fits taylor1997 X5, and the phase state (the single Tg is cited) taylor1997 X9. The 1735 and 1726 cm⁻¹ bands are both called non-hydrogen-bonded acid carbonyl, and how one turns into the other between 5 and 20 wt% is not shown taylor1997 X4.

Conditions and limits

One acid drug with one donor, one PVP grade (PVP-K90, Tg 177 °C cited), methanol-cast dispersions dried under vacuum; IR solids were ground with KBr, the physical-mix preparation, sample temperature and water content are not printed taylor1997 M4, P3, X7. The comparison with methylpyrrolidone is qualitative and starts at 30 wt% methylpyrrolidone, whereas the dispersion changes begin at 5 wt% PVP taylor1997 X6.

Counterexamples and conflicts

None found in the library that contradicts the bond. The only discrepancy is internal: the two-bond threshold of 12 wt% PVP is a halved weight fraction, not a halved mole ratio, and should be about 13.6 wt% taylor1997 X2.

Transfer check

For an acid drug in PVP, record amorphous drug (melt-quenched), a solvent-evaporated dispersion and a physical mix of the same composition by IR, and look for three markers without any crystal reference: loss of the dimer band (1710 cm⁻¹ here), a free acid C=O band 16 cm⁻¹ higher (1726 cm⁻¹; synthesis: 1726 − 1710) and a PVP carbonyl band far below its free position (1636 against 1679 cm⁻¹) taylor1997 R5.h, R11, R12. Use Raman to test for dimers by the mutual exclusion rule taylor1997 C14. Do not expect all drug to be bonded at the stoichiometric composition taylor1997 C37. For a drug with several donors, intramolecular bonds or no donor, see F4 and F14.

History

Yoshioka and co-workers found that PVP inhibits indomethacin crystallization at low levels beyond what antiplasticization explains; this paper (1997) supplied the first spectroscopic evidence in the library of the molecular contact that could be responsible taylor1997 C5, C45. Kestur and Taylor (2010) took the bond as a proposed stabilization mechanism and tested it across polymers for felodipine (see F12) kestur2010a C3. Marsac and co-workers (2009) later measured a 45 cm⁻¹ PVP carbonyl shift for indomethacin, the largest of five compounds (see F2) marsac2009 R12.c. Rumondor and co-workers (2010) cited the larger carbonyl red shift of indomethacin against felodipine, quinidine and pimozide as evidence of stronger interactions with PVPVA rumondor2010c C28. Wegiel and co-workers (2013) read polyphenol–polymer OH and carbonyl shifts the same way (see F25) wegiel2013a C8, C36, and Que and co-workers (2020) took the 20–30 cm⁻¹ red shifts of drug–PVP hydrogen bonds as the comparator for the 2–6 cm⁻¹ shifts of halogen bonds (see F20) que2020 R17.q, C25.

Related: PEG dispersions F3, Crystallization tendency F8.

How it was shown

Marsac, Li and Taylor spin-coated dispersions of each compound with PVP K12 on ZnSe in a glovebox at RH < 10% from ethanol, dichloromethane–ethanol or acetone (acetonitrile–water for sucrose), heated the film to 110 °C for several minutes and measured it by FTIR (Bio-Rad FTS-6000, 128 scans, 4 cm⁻¹), at one hydrogen-bond donor group per carbonyl acceptor group marsac2009 P4. Dipole moments of single molecules were computed by DFT (B3LYP/6-311++G(2d,p)) marsac2009 P5. The interaction parameters come from the melting point depression of the crystalline compound in PVP K12 (offset melting) marsac2009 P6, R7; see F7 and F8.

What the data show

The carbonyl shift is 45 cm⁻¹ for indomethacin, 25 for felodipine, 25 for sucrose and 0 for ketoconazole, with computed dipole moments of 2.68, 5.30, 5.40 and 6.03 D; nifedipine, ibuprofen, glucose and vinyl pyrrolidone have dipole moments (9.58, 1.76, 4.22, 3.91 D) but no shift entry marsac2009 R12. Ketoconazole has no hydrogen-bond donor marsac2009 R14. The offset χ with PVP K12 are −0.82 (indomethacin), −0.08 (felodipine), 0.00 (nifedipine), −0.08 (ketoconazole) and 0.02 (sucrose) marsac2009 R7.a, R7.b, R7.c, R7.d, R7.e. Taylor and Zografi's indomethacin dispersions gave a hydrogen-bonded PVP carbonyl band at 1636 against 1679 cm⁻¹ for free PVP, a shift of 43 cm⁻¹ (synthesis: 1679 − 1636) taylor1997 R5.l, R12. Kestur and Taylor used the drug side of the contact: the felodipine NH band at 3341 cm⁻¹ moved to 3294 cm⁻¹ with 50% PVP and to 3298 cm⁻¹ with PVP/VA, but up to 3348 (HPMCAS) and 3354 cm⁻¹ (PVAc) kestur2010a R9. A later paper cites the felodipine–PVP shifts as NH about 45 cm⁻¹ and PVP carbonyl about 20 cm⁻¹, against 25 cm⁻¹ for the carbonyl here kestur2011 R20; marsac2009 R12.b. For nimesulide, whose NH is held intramolecularly, the PVP carbonyl moved from about 1684 to about 1680 cm⁻¹ with a shoulder near 1657 cm⁻¹ kestur2011 R14, X15. On the scale of the PVP carbonyl, dextran hydroxyls in DEX1–PVP12 films moved the band from 1680 to 1661 cm⁻¹, naringenin and quercetin at 30% polyphenol from 1682 to 1661 and 1658 cm⁻¹ (shifts of 21 and 24 cm⁻¹), and the halogens of clotrimazole and brotrimazole in PVPVA from 1684 to 1682 cm⁻¹ (6 cm⁻¹ for the iodine compound Me-DIBF), against the 20–30 cm⁻¹ that the halogen paper cites for PVP hydrogen-bonded to drugs vaneerdenbrugh2012a R2; wegiel2013a R3, R4; que2020 R11, R17.q.

Figure

Read the right-hand column: 45 cm⁻¹ for indomethacin, 25 for felodipine and sucrose, 0 for ketoconazole; the dash for nifedipine is a missing entry, not a zero marsac2009 R12, X9.

Why

The authors' interpretation: felodipine, nifedipine and sucrose hydrogen-bond to themselves in the amorphous state, but this is compensated by hydrogen bonds to PVP that are stronger because PVP carries the better acceptor, and the carbonyl shift is "evidence" of those bonds marsac2009 C39. Indomethacin's larger shift and more negative χ "can most likely be rationalized" by the higher acidity of its carboxyl donor, expected to bond more strongly than NH or OH groups marsac2009 C42. Ketoconazole, with no donor, "appears" to mix about athermally; the authors "speculate" on dipole–dipole forces and van der Waals contacts of a large molecule marsac2009 C40, C41. Donor acidity and bond strength were not measured marsac2009 C42, and the nifedipine–PVP hydrogen bonding rests on a cited paper, not on Table IV marsac2009 X9. A later moisture study of PVP dispersions used the drug pKa and the PVP carbonyl red shift as two measures of interaction strength, and linked the small phase separation of indoprofen to its being a somewhat weaker acid than the other carboxylic acids studied rumondor2011 C30, C33.

Conditions and limits

Five compounds with PVP K12, 110 °C-annealed spin-coated films at one donor per carbonyl; no replicates stated marsac2009 P4, R12. The felodipine–PVP shift is about 20 cm⁻¹ in a cited study and 25 cm⁻¹ here kestur2011 R20; marsac2009 R12.b. The interaction parameter does not rank the middle of the list: felodipine and ketoconazole are both −0.08 although one shifts the carbonyl and the other does not marsac2009 R7.a, R7.c, R12.b, R12.h. The solubility-route χ in the ethyl-pyrrolidone analogue puts nifedipine (−0.81) and sucrose (−0.53) close to indomethacin (−1.83 for that route), and the card warns against calling these "near zero" marsac2009 R9.b, R9.d, R9.e, X8.

Counterexamples and conflicts

(a) Ketoconazole mixes about athermally without any shift, so a null shift does not mean immiscible (see F4) marsac2009 R7.c, R12.h. (b) The dipole moment does not rank the shift: nifedipine has the largest computed moment but no shift entry, and indomethacin has the smallest dipole moment of the four compounds with a shift entry and the largest shift marsac2009 R12. (c) The felodipine–PVP shift is about 25 cm⁻¹ here and about 20 cm⁻¹ in a cited study, so the size of one entry is uncertain by a few cm⁻¹ marsac2009 R12.b; kestur2011 R20.

Transfer check

Measure the PVP carbonyl band (free about 1679 cm⁻¹ in the cited indomethacin work) in a spin-coated film at one donor per carbonyl; a shift of 20–45 cm⁻¹ puts a donor drug inside the range seen here, and no shift means no donor–carbonyl bond but not immiscibility. Also record the drug's own donor band (NH, OH, acid) against its amorphous spectrum, as for felodipine kestur2010a R9. If the drug has an NH locked intramolecularly, expect small shifts (see F14) kestur2011 X15.

History

The 43 cm⁻¹ carbonyl shift of indomethacin dispersions first appeared in 1997 taylor1997 R12. Marsac and co-workers (2009) compared five compounds on one scale and tied the ranking to donor acidity marsac2009 R12, C42. Kestur and Taylor (2010) used the drug-side NH shift to rank four polymers for felodipine kestur2010a R9. Rumondor and co-workers (2010, 2011) used the carbonyl shift as a measure of interaction strength rumondor2010c C28; rumondor2011 C30. Van Eerdenbrugh and Taylor (2012) fitted the free and bonded PVP carbonyl bands to a hydrogen-bonded fraction for dextran–PVP blends (see F22), Wegiel and co-workers (2013) used the drug-side OH and polymer carbonyl shifts to rank six polymers for two polyphenols (see F25), and Que and co-workers (2020) found shifts of only 2–6 cm⁻¹ for halogen-bonded drugs (see F20).

Related: PEG dispersions F3.

How it was shown

Rumondor and co-workers made dispersions of felodipine and nifedipine with PVP K29-32, ketoconazole with PVP, and felodipine with poly(acrylic acid) (Mv about 450,000) by rotary evaporation of a common solution (dichloromethane–ethanol, chloroform or ethanol), and chose the pairs by the kind of interaction: hydrogen bonds known (felodipine and nifedipine with PVP), none possible (ketoconazole with PVP) and a polymer that bonds to itself (PAA) rumondor2009a P1, M1, M2, M3, M4, M5, A1. DSC used ground powders in pinholed pans at 20 °C/min from 0 °C, to 180 °C for ketoconazole–PVP and in two regimens for felodipine–PAA (to 180 °C, or to 146 °C with a 30 s hold), each cycled three times; the cycle reported and the replicate number are not stated rumondor2009a P2, X10. IR used films spin-coated on ZnS or KRS-5 and dried about 1 min at 90 °C (4 cm⁻¹, 128 scans), with physical-mixture spectra at 10–90% polymer calculated from the pure amorphous spectra, and PCA over 700–4000 cm⁻¹ after unit normalization (SIMCA-P+) rumondor2009a P3, P4, P5. PXRD used powders held about 10 °C above the drug melting point for at least 15 min, quenched in ice and salt, measured in transmission, analyzed as a PDF fit (dispersion PDF fitted with a scaled sum of the pure PDFs at 30 and 70% polymer) and by a pure curve resolution method on the whole composition series rumondor2009a P6, P7, P8, P9. The criteria: a correct fitted composition means separate phases; a wrong one a unique mixed structure; an extra component or a halo shift of about 1° 2θ or more in the pure curves is "indirect proof" of mixing rumondor2009a C10, C11.

What the data show

Felodipine–PAA DSC (Table I): felodipine 48 °C and PAA 132 °C; at 10% PAA two Tgs, 48/133 and 46/130 °C in the two regimens; at 30% PAA 48/132 and 48/133 °C, against Couchman–Karasz values of 59 and 79 °C; at 50% PAA two Tgs (48/121) in one regimen and one (102 °C) in the other; at 70 and 90% PAA single Tgs of 121 and 114 °C in one regimen and 86 and 84 °C in the other, against 112 and 126 °C rumondor2009a R1. IR scores: calculated felodipine–PVP mixtures lay on one line (t[2] ≈ −0.5 to −0.15) while the dispersions at 10, 30, 50 and 70% PVP had t[2] ≈ 0.8, 0.6, 1.25 and 1.2, and nifedipine–PVP dispersions at 25, 50 and 70% PVP ≈ 1.6, 2.1 and 1.3 against −0.75 to −0.2; the second loading is largest in the NH and carbonyl regions rumondor2009a R6, R7, R8, R9. Felodipine–PAA dispersions at 30, 50 and 70% PAA lay close to the mixture line (t[2] ≈ −0.3 to 0.0), with PC1 loadings correlating at ρ > 0.99, while the 90% PAA dispersion (t[2] ≈ 3.2) and the 10% PAA dispersion were outliers rumondor2009a R13, R14. PDF fits (polymer content found for 30/70% actual): felodipine–PVP 37/58, nifedipine–PVP 53/85, felodipine–PAA 29/64, ketoconazole–PVP 15/69 rumondor2009a R16. Pure curves: the felodipine second halo moved from ≈23.1° to ≈21.3° 2θ in felodipine–PVP (nifedipine–PVP "the same", not shown) and did not move in felodipine–PAA rumondor2009a R19, R20, R22. All PXRD samples were X-ray amorphous rumondor2009a R15. An independent example of the same ambiguity: DSC gave a single broad Tg (midpoint 87 °C) for itraconazole–HPMC dispersions that fluorescence detected as phase separated purohit2017a R3, C26. Three later papers of the chapter show the same gap between a single Tg and structure-sensitive methods: itraconazole–HPMC powders gave one broad Tg at every loading while four film methods found drug-rich domains (see F18) purohit2015a R7, R1, R11; telaprevir and PVPVA differ in Tg by only about 5 °C, which the authors say DSC cannot readily resolve (see F19) li2016c C26, M1, M4; and DEX6–PVP90, miscible by DSC, showed discrete domains of about 0.3–0.6 µm in AFM films (see F22) vaneerdenbrugh2012c R16.

Figure

(a) Dispersions (blue, % PVP) lie above the line of calculated physical mixtures (red); the vertical distance from the red line is the mixing signal. (b) The PC1 loading matches the felodipine-minus-PVP difference spectrum; the PC2 loading is largest in the boxed NH region rumondor2009a R6, R7.

At 70% polymer the felodipine–PAA dispersion and calculated curves nearly coincide, while the felodipine–PVP curves differ in peak heights and beyond ≈11 Å rumondor2009a R17, R18, X6.

Why

The authors' interpretation: a physical mixture is the composition-weighted sum of its components, while molecular mixing adds new contacts that shift IR bands (a second PC) and a new local structure that changes the PDF and the pure curves rumondor2009a C7, C9, C10, C11. The PC2 features are "presumably" drug–polymer hydrogen bonds known from earlier work rumondor2009a C20, C39. For felodipine–PAA they "speculate" that PAA's own multiple hydrogen bonds compete with the drug, leaving a drug-rich and a polymer-rich phase each holding a little of the other; at 10 and 90% PAA some mixing is "speculated" rumondor2009a C45, C46. They also give reasons why DSC alone can mislead: domains below 15–30 nm may give one Tg, heating during the scan can change miscibility, and the regimen-dependent Tgs "could be an indication" of temperature-dependent mixing rumondor2009a C6, R25, C15, C42. The energies of drug–polymer and PAA self-association were not measured rumondor2009a C39, C45.

Conditions and limits

Samples differ between methods: DSC used ground powders, IR films dried at 90 °C and PXRD melt-quenched powders, so agreement or disagreement reflects sample history as well as method sensitivity rumondor2009a X4. "The complete range" for felodipine–PVP and nifedipine–PVP rests on IR at 10–90% PVP (felodipine) and 25–70% PVP (nifedipine) and on PDF fits at 30 and 70%; DSC for the PVP systems is cited, not measured, and the nifedipine PXRD results are not shown rumondor2009a X5, R5, R20. PCA settings and the rule for the number of components are not printed; no threshold is given for good PDF agreement, since 7 points counts as poor (felodipine–PVP, 30 → 37) and 6 points as good (felodipine–PAA, 70 → 64) rumondor2009a X12, X14.

Counterexamples and conflicts

(a) DSC shows two Tgs only at 10–30% PAA, and the Discussion cites two Tgs at 90% PAA that Table I places at 10% PAA rumondor2009a X2, X3. (b) The 10 and 90% PAA outliers also sit off the composition axis of PC1, so a labelling or composition error cannot be ruled out rumondor2009a X13. (c) For ketoconazole–PVP the methods disagree (see F4) rumondor2009a X1. (d) The Discussion's firmest statement ("clearly established") covers the compositions studied, whereas the abstract speaks of the complete range rumondor2009a C38, X5.

Transfer check

Do not rely on a single Tg for a new drug–polymer pair: prepare the sample the way the product will be made, and add at least one structure-sensitive method on the same preparation. For IR, build calculated physical-mixture spectra from the pure amorphous drug and polymer spectra and test whether the dispersions need a second PC; for PXRD, fit the dispersion PDF with the pure PDFs at 30 and 70% polymer and inspect the PDF shapes too, because the composition fit alone separated the cases by one percentage point rumondor2009a P4, P5, P8, X14.

History

The calculated-spectrum procedure and the drug–PVP hydrogen bonding were taken from earlier library papers on felodipine and nifedipine in PVP (builds_on links in the card) rumondor2009a R23, P4. This paper (2009) combined DSC, IR with PCA and PXRD with PDF and pure-curve analysis on one set of dispersions and calls the three complementary rumondor2009a C47, C48; no earlier paper of this chapter does so (synthesis). Later library papers cite it for IR as a test of drug–polymer mixing and for the observation that the PXRD halo of a molecular dispersion is not the sum of the pure halos (evidence index, citation links). Purohit and Taylor (2015) later paired DSC with fluorescence probes and AFM-IR on itraconazole–HPMC and found the separation that DSC had not shown (see F18), and Van Eerdenbrugh and Taylor (2012) paired DSC and IR on 20 polymer blends (see F22).

How it was shown

In the melting route, crystalline ketoconazole and PVP K12 (45–90 µm, dried over nitrogen for at least a week) were mixed as powders, one to three samples per composition, and heated at 1 °C/min in a modulated DSC; the offset of the melting endotherm was taken, and χ followed from the Flory–Huggins relation with the volume of ethyl-pyrrolidone as lattice site marsac2009 P1, P6. In the solubility route, excess ketoconazole was stirred in 1-ethyl-2-pyrrolidone at 25 °C for at least 24 h and assayed by UV–Vis marsac2009 P2. Rumondor and co-workers made ketoconazole–PVP K29-32 dispersions from chloroform and measured DSC (to 180 °C at 20 °C/min), IR (30 and 70% PVP), PXRD of melt-quenched powders (PDF fit and pure curves) rumondor2009a P1, P2, P3, P6, P8, P9. The drug was chosen because it cannot hydrogen-bond to PVP rumondor2009a C22, M3.

What the data show

Offset melting: ≈150.0 °C at no polymer, ≈149.5, 149.1, 148.75, 148.3, 148.2, 147.9, 147.65 and ≈147.5 °C at Φ ≈ 0.03 to 0.23, only slightly below the entropy-only line (≈147.8 °C at 0.25) marsac2009 R3. χ −0.08 (melting) and −0.12 (solubility route), γ 0.68 in the analogue, predicted solubility in PVP K29/32 0.010 weight fraction and γ 5.92 there marsac2009 R7.c, R9.c. The PVP carbonyl shift is 0 and the computed dipole moment 6.03 D marsac2009 R12.h. DSC: one Tg for every composition from ≈47 °C (neat drug) to ≈167 °C (PVP), with measured values ≈44, 54, 84, 109 and 135 °C at 10, 30, 50, 70 and 90% PVP against a Couchman–Karasz line of ≈51, 64, 80, 103 and 140 °C, so ≈7–10 °C below it at 10–30% PVP rumondor2009a R3, R4, X9. IR: carbonyl peaks at the pure-component positions, the drug carbonyl peak slightly weaker than in the calculated mixture; PCA was dropped because the calculated mixtures alone needed two components rumondor2009a R10, R11. PDF-fitted polymer content 15 and 69 for 30 and 70% actual; pure curves with a slight PVP halo shift and a broader ketoconazole halo rumondor2009a R16, R21.

Figure

The offset melting point falls only slightly, just beyond the entropy-only line; the fitted line misses the curvature of the points marsac2009 R3, X15.

Why

The authors' interpretation: ketoconazole cannot donate a hydrogen bond, so they "speculate" that it interacts with PVP through dipole–dipole forces, supported only by its computed dipole moment and large size marsac2009 C40, C41; rumondor2009a C23. For the Tg result, Rumondor and co-workers offer two readings: heating may drive the system to one phase, or DSC may not detect a second phase ("it is possible") rumondor2009a C42. They state that the PDF/PXRD data show structural change at 70% PVP but not at 30% and "suspect" partial miscibility, one phase "only at higher PVP", and that the system "clearly needs further investigation" rumondor2009a C40, C43. No interaction energy was measured rumondor2009a C39.

Conditions and limits

One drug, PVP only, two preparation routes (physical powders for melting, rotary-evaporated films and quenched powders for the rest) marsac2009 P1; rumondor2009a P1, P3, P6. The Couchman–Karasz agreement is called "reasonable" by the authors but is approximate; the 10% PVP value (≈44 °C) lies below the neat drug's ≈47 °C rumondor2009a C17, X9.

Counterexamples and conflicts

(a) The text and Table II assign the agreeing PDF fit to opposite compositions: Table II gives 15 at 30% and 69 at 70% PVP, so by the paper's own criterion mixing would appear at 30% and not at 70%, while the Discussion states change at 70% and not at 30% rumondor2009a R16, C40, X1. (b) Cited Tg evidence classes the pair as miscible (one Tg at 50% polymer and above) marsac2009 R1.c; rumondor2009a R5. (c) In PEG, ketoconazole splits the readings of interaction in a similar way: an activity coefficient of 0.09 in PEG 400 but a positive χ (0.02) and no solubility in PEG 3350 marsac2008b R14.f; zhu2011 R5.f.

Transfer check

For a drug without donors, do not read a null IR or carbonyl shift as immiscibility, and do not read one Tg as miscibility. Combine the offset melting curve (small depression and χ near zero are expected for an athermal pair), IR on a 30 and 70% polymer film, and a structure-sensitive method on the same preparation, and expect the answer to differ with composition marsac2009 R3, R7; rumondor2009a R10, R16.

History

Marsac and co-workers (2009) first placed ketoconazole–PVP at near-zero χ with a null carbonyl shift and proposed dipole–dipole forces marsac2009 C40, C41. Rumondor and co-workers (2009) tested the same pair by DSC, IR and PXRD and left the phase state open rumondor2009a C40, C43, X1. Ketoconazole in PEG later repeated the pattern of conflicting readings marsac2008b R14.f; zhu2011 R5.f. Moseson and Taylor 2023 (pp. 13–14) cite Ivanisevic (2010), outside the library, who classified ketoconazole/PVP as miscible at 30% and immiscible or phase-separated at 70% drug loading; the 70% sample crystallized after about 2 months and the 30% sample did not within about 1 year.

Related: PEG dispersions F3.

How it was shown

Purohit and Taylor dissolved itraconazole (ITZ) and HPMC grade 606 at 50:50, 40:60, 30:70, 20:80 and 10:90 w/w in 50:50 v/v methanol–dichloromethane at 20 mg/mL total solids; a 50:50 solution in 1:1 methanol–tetrahydrofuran served as a solvent control purohit2015a P1. For fluorescence, 100 µl was spin-coated on quartz at 3000 rpm for 30 s and vacuum-dried for 24 h, with pyrene or prodan at 0.01% w/v; emission spectra were taken on a spectrofluorophotometer (excitation wavelength not stated) and images with a fluorescence microscope (excitation 330–380 nm, emission from 420 nm) purohit2015a P2, P3, P4. For AFM-nanoIR, probe-free films on ZnSe were imaged in contact mode and spectra were taken from 1600 to 1800 cm⁻¹ (256 co-added scans, 4 cm⁻¹) at marked points on domains and on the continuous phase; nanoTA ramped a heated probe at 2 °C/s on one domain and two matrix spots of a 40:60 film purohit2015a P9, P10. Powders for DSC came from rotary evaporation at a 45 °C bath and were scanned by modulated DSC at 2 °C/min to 170 °C (±1 °C every 60 s), held 2 min, quenched to 0 °C and reheated purohit2015a P7, P8. Films were heated for 2 min at 165 °C for the fluorescence tests; prodan was not measured after heating because it is unstable at high temperature purohit2015a P5, P3.

What the data show

Both probes read a more polar surrounding in pure HPMC than in amorphous ITZ (pyrene I/III 1.54 against 1.39; prodan 452.9 against 439.0 nm). In the blend films before heating the readouts stayed near the ITZ values: I/III 1.39 (50:50), 1.40 (40:60), 1.41 (30:70), 1.42 (20:80) and 1.42 (10:90); prodan 439.3, 438.7, 439.3, 440.7 and 443.7 nm, with stated standard deviations below 0.02 and below 2 nm purohit2015a R1, X13. A broad pyrene band near 470–480 nm (text 470, caption 475, plotted maximum about 480 nm) appeared in ITZ and in every blend but not in HPMC at the same probe level, which the authors take as an ITZ–pyrene exciplex that shows the probe located in drug-rich domains purohit2015a R3, C11, C12, X5. Films were translucent to opaque as cast purohit2015a R5. A 10:90 film imaged with pyrene showed bright domains of about 1 µm on a darker continuous phase, and the authors report similar images at all ratios purohit2015a R6, C20. AFM topography showed round raised domains at every ratio, with diameters read from the images of about 0.4–1.0 µm at 10:90–30:70 and 0.5–1.4 µm at 40:60; the authors state 500–1000 nm, and the one chemical map (30:70) shows ITZ-rich regions of about 1.5–2.5 µm purohit2015a R10, R14, C37, X8. The ITZ carbonyl band at about 1700 cm⁻¹, absent from HPMC, was strong on the domains and weak or absent in the matrix: about 0.3–0.5 of the domain band at 10:90, about 0.2 at 20:80, 0.1–0.4 at 30:70, none above noise at 40:60 and about 0.05 at 50:50 purohit2015a R9, R11. On the 40:60 film nanoTA gave 71 °C on a domain (amorphous ITZ Tg 60 °C) and 108 °C on the matrix (HPMC Tg 140 °C) purohit2015a R13, M1, M2.

DSC of the rotary-evaporated powders gave one step at every composition. First-run onset, midpoint and offset were 70/77/81 °C (50:50), 78/87/95 (40:60), 84/100/108 (30:70), 94/116/119 (20:80) and 112/131/135 °C (10:90); the offset minus onset is 11, 17, 24, 25 and 23 °C (synthesis: 81 − 70, 95 − 78, 108 − 84, 119 − 94, 135 − 112) purohit2015a R7. After 2 min at 170 °C and a quench the second-run midpoints were 64, 70, 80, 90 and 115 °C, that is 13, 17, 20, 26 and 16 °C lower (synthesis: 77 − 64, 87 − 70, 100 − 80, 116 − 90, 131 − 115) and the steps were narrower, 9, 10, 13, 16 and 6 °C (synthesis: offset minus onset of the second run, for example 71 − 62) purohit2015a R7. The 50:50 first scan also has a hump near 105 °C and a dip near 165 °C that the text does not discuss purohit2015a R8, X11. After heating the films at 165 °C they were transparent, I/III was 1.50, 1.50, 1.53, 1.50 and 1.52 for 50:50 to 10:90, the exciplex band was gone from the 10:90 spectrum, the fluorescence image was more uniform, topography no longer showed round domains and each of eight nanoIR spectra per film carried the carbonyl band; its height varied by about 0.3–1 of the highest at 50:50 purohit2015a R1, R4, R5, R6, R12, X9. The 50:50 film from methanol–tetrahydrofuran gave I/III 1.39, the same as the methanol–dichloromethane film purohit2015a R2.

Figure

Round raised domains stand in a continuous matrix at each loading; blue spectra from domains carry a strong 1700 cm⁻¹ band, red spectra from the matrix a weak band or none. Note the different height scales and the 2.5 µm scan of the 50:50 film purohit2015a R10, R11.

Before heating every blend sits near the ITZ values; after heating the pyrene ratio of every blend rises toward the HPMC value purohit2015a R1.

Why

The authors' interpretation: the probes are located "preferentially" in an ITZ-rich phase in all films ("suggests"), and the exciplex band shows probes in drug-rich domains because in one miscible phase the pyrene–ITZ distances would "most likely" exceed about 10 Å purohit2015a C8, C11, C12. The domains are ITZ-rich by the carbonyl band and by softening near the ITZ Tg ("indicating"; "indicates") purohit2015a C29, C34. For the origin of the separation they offer two routes and say that choosing between them is "very hard" purohit2015a C48. Kinetic: dichloromethane evaporates first, ITZ is freely soluble in it and poorly soluble in methanol and "might reasonably be expected" to exceed its miscibility limit in the methanol that remains; both phases are then glassy at room temperature (Tg 60 and 140 °C), so a phase-separated system "would be expected to persist" purohit2015a C44, C45, C46. Thermodynamic: ITZ has no hydrogen-bond donor and HPMC has few free hydroxyls, so any miscibility is "likely" driven by entropy and may be limited at room temperature and rise with temperature, as in an upper critical solution temperature diagram purohit2015a C41, C42, C47. On heating, mobility returns and entropy favours mixing near the drug melting point, and cooling would "very likely" trap a miscible glass purohit2015a C46, C47. The single DSC Tg is explained as the powder mixing during the first scan ("most likely"), a reading supported only by the narrower second-run steps; the midpoint drop of 13–26 °C on reheating is reported but not explained purohit2015a C39, C26, X14. What was not measured: the phase state of the DSC powders before the scan, a χ or phase diagram, the solvent composition during drying, and any crystallization or dissolution consequence purohit2015a C47, C45, C51, X1. (synthesis: the exciplex argument excludes pyrene crowding, since HPMC with the same probe level shows no band, but not an ITZ–pyrene exciplex inside a molecular mixture purohit2015a X3.) A competing explanation from outside the chapter is water: itraconazole–HPMC was thermodynamically miscible as a dry binary and separated when made from a water-containing solvent, and the film structure changed with the alcohol of the solvent mugheirbi2017 C39; mugheirbi2018 R1, C50. The card of this paper records no humidity or solvent water content.

Conditions and limits

One drug (no hydrogen-bond donor), one polymer grade, five loadings, one solvent pair plus one control, spin-coated films of unrecorded thickness purohit2015a P1, P2, A1, A3. DSC was run on powders and the film methods on films, so "DSC misses the separation" compares differently made samples purohit2015a X1. The AFM and nanoTA results come from single films and few spots, and heating time and cooling of the AFM films, spin conditions on ZnSe and the number of locations are not stated purohit2015a X12. Domain size is not quantified purohit2015a X8. No stability, crystallization or dissolution data were collected purohit2015a C51. Spray drying and hot melt extrusion were not prepared, so the processing statement ("might"; "would be more likely") is an extrapolation purohit2015a C49, X10.

Counterexamples and conflicts

(a) Domain sizes of 500–1000 nm in the text are exceeded by readings of up to about 1.4 µm in Fig. 6d and about 2.5 µm in the chemical map purohit2015a C37, X8. (b) "Almost equal intensity" of the carbonyl band after heating fits 10:90 to 30:70 better than 50:50, where heights still vary about three-fold purohit2015a C31, X9. (c) The claimed agreement with an earlier thermal-analysis study compares solvent-cast films with melt-extruded material and a 40% threshold with separation from 10% purohit2015a C43, X10. (d) The kinetic and the thermodynamic reading are not separated, and the dry-binary result of the later water study points to a preparation effect rather than limited miscibility purohit2015a C48; mugheirbi2017 C39. (e) The methods of the film series could not say whether the separation is complete: the matrix at 10:90 and 30:70 still carries a weak carbonyl band purohit2015a R11.

Transfer check

For a drug that has no hydrogen-bond donor or interacts weakly with the polymer: (1) do not accept one Tg as proof of one phase; (2) make the film or particle by the route that will be used, with the humidity and solvent water recorded, and test it with two methods that read local composition (a polarity probe against pure-component references, and AFM-IR or imaging); (3) run DSC twice, with a hold above Tg, and note whether the transition narrows or moves, as a cheap sign that the as-made material was not uniform; (4) if the pair separates, compare loadings and solvent pairs before concluding that the drug and the polymer cannot form one phase. A result with no donor on the drug, 10–50% drug and a mixed solvent whose volatile component dissolves the drug would put the system inside the evidence of this paper purohit2015a P1, A1, M1.

History

Rumondor and co-workers (2009) had said that a phase-separated dispersion should crystallize faster purohit2015a C2. Van Eerdenbrugh and co-workers (2012) showed AFM-IR on a drug–polymer and a polymer–polymer blend (see F22) purohit2015a C4. Purohit and Taylor (2015) combined fluorescence probes, AFM-nanoIR and nanoTA on itraconazole–HPMC and found the separation at every loading. Li and Taylor (2016) applied the AFM tools to telaprevir with three polymers (see F19). Purohit and co-workers (2017) showed that the pair separated whichever route made it (rotary evaporation, film casting, spray drying), and Mugheirbi and co-workers (2017, 2018) found that water and the alcohol in the solvent changed the outcome purohit2017a R3, R7, C26; mugheirbi2017 C39; mugheirbi2018 R1.

Related: Solid-state methods F19, Solid-state methods F20, ASD process F1, ASD process F2.

How it was shown

Li and Taylor dissolved telaprevir (TPV) and HPMC (Methocel E5), HPMCAS (HF grade) or PVPVA (Kollidon VA 64) at 5% w/v total solids in 1:1 methanol–dichloromethane, spin-coated 30–50 µL on 1 cm ZnS flats (6 s at 50 rpm, then 50 s at 3100 rpm) and dried the films overnight in a vacuum oven at ambient temperature; ratios were 1:9 to 5:5 with all three polymers, plus 6:4 (HPMC, HPMCAS) and 7:3 and 8:2 (HPMCAS) li2016c P1. Film thickness, storage time and humidity were not reported li2016c X9. Height images were taken in contact mode (tapping for 30–50% PVPVA) and domains were measured with the software ruler li2016c P2. Local AFM-IR spectra (1200–1800 cm⁻¹, 4 cm⁻¹ steps, 256 co-averages, four to five points per phase) were compared as band ratios because AFM-IR intensity follows film thickness; the telaprevir marker is the 1525 cm⁻¹ band, absent from the three polymers li2016c P4, C11, C12, R2. Images at the drug band, ratio images and contact-resonance frequency maps were taken on 60:40 HPMC, 80:20 HPMCAS and 30:70 PVPVA; nanoTA ramped a heated probe at 5 °C/s (five or six ramps per phase); Lorentz contact resonance (LCR) imaged stiffness at frequencies chosen per film; pyrene (0.01%) films on quartz were imaged by fluorescence li2016c P5, P7, P8, P9. DSC of the pure components gave Tg 102.7 ± 0.4 (TPV), 140 ± 5 (HPMC), 122 ± 1 (HPMCAS) and 108 ± 1 °C (PVPVA); no blend was run by DSC li2016c M1, M2, M3, M4, X8.

What the data show

Topography: telaprevir–HPMC films carried discrete domains at 10% drug (below 50 nm wide), growing with loading and becoming irregular between 40 and 50% (about 30 nm high at 50%) li2016c R4, R7. Telaprevir–PVPVA films carried domains from 10% drug, crowding closer and growing from about 280 to 380 nm in diameter and from about 20 to 100 nm in height up to 50% li2016c R6. Telaprevir–HPMCAS films below 40% drug looked like the pure polymer film (irregular height variation, read by the authors as uneven film formation); at 40% very small circular domains of about 120 nm and 50 nm height appeared, with gaps of only 21 nm (40%) and 25 nm (50%), and at 50:50 the domains were 122 nm wide and 13 nm high li2016c R5, R7, C5.

Local spectra: in 60:40 telaprevir–HPMC (four points per phase) the HPMC band at 1450 cm⁻¹ was comparable to the 1525 cm⁻¹ band in the matrix and considerably weaker in the domains; in 80:20 telaprevir–HPMCAS the 1525 cm⁻¹ bands were similar in both phases while the HPMCAS band at 1236 cm⁻¹ was clearly stronger in the matrix; in 50:50 telaprevir–PVPVA the 1244 cm⁻¹ PVPVA band exceeded the drug band in the matrix and fell below it in the domains li2016c R8, R9, R10, X11. HPMCAS was analysed at 80:20 because at 40–50% the domains sit 21–25 nm apart, below the typical AFM-IR resolution of about 100 nm; at 50:50 the spectra of the two phases overlaid exactly and IR imaging failed li2016c C15, R7, R24, R11. Bands of both components appeared in both phases, which the authors attribute either to partial miscibility or to spectral mixing through or beside the domain li2016c C14. In the dispersed-phase spectra the drug band grew relative to the polymer bands as loading rose; since the compositions of coexisting phases in a miscibility gap should stay fixed, the authors judge spectral mixing that falls as the domains grow a "more likely explanation" than a kinetically trapped composition, a choice not tested li2016c R12, C20, C21, X10. Ratio images (polymer band over drug band) were higher in the continuous phase in all three films, whereas single-band images at both the drug and the polymer bands were strongest on the raised domains, a thickness effect li2016c R13, R14, R15, C22, C23.

nanoTA softening, dispersed against continuous phase (Table 2): 130 ± 2 and 156 ± 5 °C (60:40 HPMC), 124 ± 14 and 143 ± 3 °C (80:20 HPMCAS) and 131 ± 5 and 146 ± 4 °C (30:70 PVPVA); pure components 102 ± 5 (TPV), 209 ± 11 (HPMC), 151 ± 2 (HPMCAS) and 132 ± 3 °C (PVPVA) li2016c R16. The text swaps the two HPMCAS phases relative to the table li2016c X1. The authors read softening far below the telaprevir melting point (240 °C) as proof that the drug in the domains is amorphous, so the separation is not crystallization li2016c C29, M1. They read the HPMC-rich matrix softening below pure HPMC as plasticization by telaprevir ("most likely") and the PVPVA-rich matrix softening above pure PVPVA as possible drug–polymer interaction ("may be caused") li2016c C30, C32. With the PVPVA Tg within about 5 °C of that of the drug (synthesis: 108 − 102.7 = 5.3), nanoTA gave little contrast li2016c C48, M1, M4. LCR placed the first resonance of TPV at about 130 kHz, HPMC at about 127.5 and HPMCAS at about 128 kHz; images at 128 kHz marked the drug-rich domains and at 123 or 124 kHz the matrix li2016c R18, R19, R20. The smallest feature the authors discriminate is about 98 nm, and the 50:50 HPMCAS film gave "good contrast" by LCR (data not shown); the marking circles hide the image in this manuscript version li2016c C49, R19, R20, X7. Pyrene fluorescence was even in a pure HPMCAS film and brighter in the dispersed phase of the three high-loading films li2016c R22, C40. The microstructure of telaprevir–HPMCAS films at 20–50% drug varied from day to day with uncontrolled humidity li2016c R23, C18.

Figure

Rows are polymers and columns loadings; discrete domains are present in the top rows (10% drug) with HPMC and PVPVA, and with HPMCAS only faint irregular features appear until 40% and 50%. The height scales differ between panels li2016c R4, R5, R6.

Dispersed phases soften between the telaprevir value and the polymer value in the cellulose films; the two PVPVA phases overlap li2016c R16.

Why

The authors' interpretation: raised domains that grow with loading are "most likely" telaprevir-rich in HPMC and "probably" in PVPVA; for HPMCAS topography alone cannot settle either point li2016c C6, C7, C8, C9. Chemical contrast then shows the domains drug-rich in the three high-loading films ("indicating"; "most likely"; "preliminary analyses … suggest") li2016c C13, C16, C19. For HPMCAS the smaller miscibility gap is read from the later onset of separation and from the near-identical phase spectra at 50:50, where both phases are rich in telaprevir li2016c C17, C46. Moisture, which promotes separation, is offered as the reason for the day-to-day variation of the HPMCAS films ("seemed to be somewhat dependent") li2016c C18. The paper measures no drug–polymer interaction and does not say why telaprevir remains in one phase with HPMCAS to a higher loading than with HPMC or PVPVA. Outside the chapter, TPV–PVPVA dispersions were also expected to separate "probably through water taken up during preparation and storage" and the poor release of PVPVA ASDs was attributed to possible solid-state phase separation mosqueragiraldo2018b C54, C55.

Conditions and limits

One drug and three polymers whose Tg lies 5–37 °C above that of telaprevir (synthesis: 108 − 102.7 = 5.3; 140 − 102.7 = 37.3) li2016c M1–M4; thin spin-coated films from methanol–dichloromethane; ambient humidity not controlled; chemical evidence of two different phases exists only at 80:20 for HPMCAS, so the 30–40% onset is a topography-based estimate li2016c X3, X9. Bulk dispersions (spray-dried, extruded) were not examined, and the authors expect spray-dried particles to need embedding and microtoming li2016c C53. No dissolution, stability or crystallization data tie domain size or composition to performance li2016c C41, C42.

Counterexamples and conflicts

(a) The onset for HPMC and PVPVA is given as "above 10%" in the abstract although domains are reported at 10% li2016c C52, X4. (b) The HPMCAS onset is quoted as above 40%, above 30%, about 30–40% and above 30–40% in different places li2016c X2. (c) The text swaps the HPMCAS phases relative to Table 2, misstates the nanoTA offset from DSC (20–60 °C against 24–69 °C in the tables) and, for PVPVA, reads a lower contact-resonance frequency as stiffer against its own rule that frequency rises with stiffness li2016c X1, X5, X6. (d) The statement that similar Tgs preclude DSC rests on no DSC of any blend li2016c C44, X8. (e) The HPMCAS 128 kHz LCR assignment sits at the pure HPMCAS resonance, so it is relative contrast, not peak identity li2016c X14.

Transfer check

For a drug whose Tg is close to that of the candidate polymers, DSC cannot be relied on, so make spin-coated films at 10, 30 and 50% drug from the intended solvent at recorded humidity and look first for domains by AFM; use ratio images, not single-band images, to say which phase is drug-rich; use stiffness imaging where the domains are below about 100 nm li2016c C22, C46, C49. Put the polymer with the highest onset loading (HPMCAS here) first in a screen and compare it with HPMC and PVPVA at the same loading, remembering that HPMCAS structure varied with humidity li2016c C52, C18. Do not carry the onset loadings to spray-dried or extruded material without testing it.

History

Rumondor and co-workers (2009) had argued that a phase-separated dispersion crystallizes faster, and Van Eerdenbrugh and co-workers (2012) had imaged drug–polymer separation by AFM-IR li2016c C1, C43. Li and Taylor (2016) added thermal and mechanical contrast and compared three polymers for one drug. Later library papers used the AFM-LCR method for water-induced separation of films and for telaprevir–PVPVA dispersions, used AFM-IR and LCR as an orthogonal check for miconazole–PVPVA, took over the nanoTA protocol for lopinavir–HPMC and used nanoTA for local miscibility (evidence index, citation links). Hiew and co-workers (2022) cited the polymer comparison for PVPVA against HPMCAS hiew2022a R31.

Related: Solid-state methods F12, Solid-state methods F13, Solid-state methods F14, ASD process F7.

How it was shown

Que and co-workers chose compounds that carry Cl, Br or I but no hydrogen-bond donor: clotrimazole and loratadine as received, brotrimazole (the Br analogue of clotrimazole) and Me-DIBF (four iodines per molecule) synthesized in-house que2020 M1–M4, P1, P2. Dispersions were made by rotary evaporation from ethanol (tetrahydrofuran for Me-DIBF) at 60 °C (40 °C for tetrahydrofuran), vacuum-dried for one day, at 10, 20, 25, 30, 40 and 50 wt% drug for clotrimazole and loratadine and 10–40 wt% for brotrimazole and Me-DIBF; no test of the amorphous single-phase state is reported que2020 P4, X10. XPS used 100 mg of dispersion pressed into an 8 mm tablet (1500 psi, 60 s), monochromatic Al Kα, five 0.5 mm spots, energies referenced to the C–C peak at 284.8 eV, a depth of about 10 nm que2020 P5. NMR titrations added 11 to 1219 mM 1-ethyl-2-pyrrolidone (EP, the monomer analogue of the vinylpyrrolidone unit) to a 14.5 mM solution of the compound at 25 °C in deuterated cyclohexane, and in deuterated toluene, fitting the shift of a proton next to the halogen-bearing carbon (¹H) and of that carbon (¹³C) to a 1:1 model; three titrations each; a carbonyl-free control used 1-ethylpyrrolidine que2020 P6, C4, C5, X6, X7:

K_F = \frac{[\mathrm{AD}]}{[\mathrm{A}]\,[\mathrm{D}]} que2020 E1

with A the free acceptor (EP), D the free halogenated compound and AD their complex. IR used films spin-coated from 100 mg/mL ethanol solutions onto KRS-5 at 10–90 wt% (128 scans, 4 cm⁻¹) que2020 P7. Each low-energy halogen atom was assumed to bond one vinylpyrrolidone (VP) carbonyl, which with the VP content of the polymer gives the molar ratio of free to bonded carbonyls que2020 E2.

What the data show

XPS binding energies (Table 1, eV, first value of each doublet): clotrimazole Cl 2p neat 200.4, in PVPVA 198.1 and 196.7, in PVP 197.9 and 196.8; loratadine neat 200.6, in PVPVA 198.0 and 196.8; brotrimazole Br 3d neat 70.3, in PVPVA 67.1; Me-DIBF I 3d neat 620.7, in PVPVA 618.1 que2020 R7. The two new Cl doublets are not assigned separately que2020 X12. Neat brotrimazole already carries a low-energy Br component of 25% of the Br, attributed to two conformations seen as 18.3% disorder in its crystal, and this is not subtracted from the dispersions que2020 R8, R5, C10, X16. The low-energy share by loading (Table S2, %) was:

que2020 R9, X23. The fall holds end to end in all five systems, but not step by step in three (clotrimazole–PVPVA 53.6 → 56.7 between 25 and 30 wt%, clotrimazole–PVP 40.9 → 51.6 between 30 and 40 wt%, loratadine 69.7 → 82.1 between 10 and 20 wt%) que2020 X13. The derived ratio of free to bonded VP carbonyls read from the figure is about 22, 21, 26, 36 and 16 at 10 wt% (clotrimazole–PVPVA, brotrimazole, loratadine, clotrimazole–PVP, Me-DIBF) and about 8, 5, 7, 8 and 7 at 40 wt%, about 5 and 7 at 50 wt% for the two clotrimazole systems; the authors speak of "about 1 halogen bonded carbonyl per 8 free carbonyls", but the plateau spans 5–8, and two plotted points do not follow from the table que2020 R16, C31, X14, X15. K_F with EP in cyclohexane: clotrimazole 0.63 ± 0.05 (¹H) and 0.97 ± 0.04 (¹³C), loratadine 0.65 ± 0.05 and 1.11 ± 0.07, brotrimazole 0.56 ± 0.06 and 0.81 ± 0.04 M⁻¹; in toluene 0.32/0.35, 0.35/0.39 and 0.30/0.21 M⁻¹ que2020 R12, R13. The ¹³C values are 1.45–1.71 times the ¹H values (synthesis: 0.97/0.63 = 1.54; 1.11/0.65 = 1.71; 0.81/0.56 = 1.45), which the authors call "similar" que2020 C13, X8. IR: the VP band moved from about 1684 to 1682 cm⁻¹ with clotrimazole and brotrimazole between 10 and 90 wt% and by 6 cm⁻¹ with Me-DIBF; the VA band at 1736 cm⁻¹ stayed; no replicate spectra are reported, and 2 cm⁻¹ is half the 4 cm⁻¹ resolution que2020 R11, X19. In PVP, the O 1s peak of the carbonyl moved from about 530.8 to 531.0 eV with 50 wt% clotrimazole que2020 R10. With 654 mM 1-ethylpyrrolidine, which has no carbonyl, the proton shifts were only minor against clear shifts with EP; the SI shows two spectra, no additive-free reference and no shift values que2020 R15, C16, X9.

The same polymer carbonyl moves much further with hydrogen-bond donors, as the earlier findings show: 45 cm⁻¹ with indomethacin, 25 with felodipine or sucrose, 21 and 24 with naringenin and quercetin and 19 with dextran hydroxyls (synthesis: 1680 − 1661), against 2–6 here (see F2) marsac2009 R12; wegiel2013a R4; vaneerdenbrugh2012a R2; que2020 R11. The authors compare with a binding constant of 0.7 M⁻¹ for weak hydrogen bonds to fluorine and with IR red shifts of 20–30 cm⁻¹ when PVP bonds drugs que2020 R17.p, R17.q, C25.

Figure

Each panel has the dispersion on top and the neat compound below; red is the doublet at the neat position, blue the new low-energy doublets seen only in the dispersions que2020 R7, R8.

A lower value means more carbonyls are bonded; the ratio falls from 16–36 at 10 wt% and levels at about 5–8. It is derived from the surface-weighted XPS shares and the bulk composition que2020 R16, X20.

Why

The authors' interpretation: a carbon-bound halogen has an electron-poor region, the σ-hole, opposite the C–X bond; a lower binding energy means the halogen became more electron-rich, so the halogens "likely received electrons from carbonyl groups" of the polymer que2020 C9. A shift of 2–4 eV is comparable to shifts caused by hydrogen bonding, which "lends support" que2020 C23. The acceptor is the VP carbonyl, not the vinyl acetate ester, because only the VP IR band moves, the PVP carbonyl oxygen loses electron density, and a carbonyl-free titrant gives only minor shifts; the authors find this "not surprising" que2020 C18, C19, C20, C21. The fall of the bonded share with loading is "rationalized" as drug molecules far apart meeting mainly polymer at low loading and meeting each other at high loading, and the plateau "may be due, at least in part, to steric constraints"; Me-DIBF with four iodines "likely" saturates quickly que2020 C11, C12, C31, C32, C34. Iodine "appeared" stronger and Cl and Br "appeared similar", which "can perhaps be rationalized" by Br···C contacts of 3.373 and 3.425 Å in the brotrimazole crystal, inside the 3.55 Å van der Waals sum, so that brotrimazole self-interacts more que2020 C26, C27, C29, C30, R5. Not measured: the strength in the solid dispersion itself, a physical mixture or charging control for XPS, self-association in the amorphous state, drug–polymer contact by solid-state NMR or DFT, and any stability or release effect que2020 C1, C27, X11, X12, C2.

Conditions and limits

Four compounds without hydrogen-bond donors in PVPVA and PVP only; no test that the dispersions were amorphous and single-phase (up to 50 wt% for XPS, 90 wt% for IR) que2020 X10. XPS probes about 10 nm of a pressed tablet, the IR films and the XPS dispersions were made differently and cover different loadings, and the K_F values come from a monomer analogue in cyclohexane or toluene que2020 P5, X18, C4. Fluorine, drugs that carry both a halogen and a donor, other polymers and water were not tested que2020 A1, A2.

Counterexamples and conflicts

(a) Brotrimazole has the larger low-energy share than clotrimazole at every common loading (87.9 against 73.0% at 10 wt%, 55.9 against 32.8% at 40 wt%), whereas the authors say that Br did not raise the extent of bonding; the 25% low-energy Br of the neat compound is not subtracted que2020 R9, C33, X16. (b) The iodine ranking rests on the IR peak position alone, which with four iodines per Me-DIBF molecule could also reflect a larger bonded fraction que2020 C30, X17. (c) Hydrogen bonding is favoured over halogen bonding when both are possible in crystals, so the contact would matter most for drugs without donors (cited) que2020 R17.f. (d) The abstract and conclusions say "demonstrated" and "were found" where the discussion says "we propose" que2020 C1.

Transfer check

For a halogenated drug without donors in PVP or PVPVA, expect a weak specific contact, far weaker than the drug–PVP hydrogen bonds in F2. To check it: take the XPS halogen core level of the neat drug (amorphous) and of a physical mixture as well as of the dispersion, with a charging control; repeat the IR of the VP and VA carbonyls with replicates; titrate a monomer analogue in a non-aromatic solvent; and test whether the share drops with loading que2020 X11, X12, X19. For a drug that also carries a donor, hydrogen bonds should dominate que2020 R17.f.

History

Taylor and Zografi (1997) and Rumondor and co-workers (2009) had made the PVP carbonyl shift of hydrogen-bonding drugs the reference for drug–polymer contacts que2020 R17.j, R17.q. Que and co-workers (2020) reported the first halogen bond between a drug and a polymer in the library que2020 C3. In 2021 the same group used the iodine analogue Me-TIPP to separate halogen from hydrogen bonding in PVPVA and found the halogen bond weaker que2021 C7, C15, and in 2023 Deac and co-workers linked a lower limit of congruency to stronger specific interactions, including halogen bonds deac2023b C2.

Related: F2, F1.

How it was shown

Wegiel and co-workers spin-coated polyphenol–polymer films on KRS-5 and followed the two bands that the literature assigns to non-ionized dimethylamino groups of E100 (2820 and 2770 cm⁻¹) at 0, 10, 30 and 50% quercetin; the polyphenols are weak acids and E100 is the only basic polymer of their set wegiel2013a P4, R6, C16. Hiew and co-workers made lumefantrine (tertiary amine pKa 8.73) dispersions by rotary evaporation with four enteric polymers and PVPVA, pressed 25% drug-loading compacts, and measured the N 1s peak of the surface by XPS (n = 3), split into unprotonated and protonated components; they also subtracted the amorphous drug spectrum from the ATR-FTIR spectra of the dispersions to look for carboxylate bands, and re-measured the surface after 10 or 30 min in pH 6.8 buffer hiew2022a M1, P13, P16, R19, R20. Van Eerdenbrugh and Taylor calculated pKa values (PSSA −1.5, PAA 4.5, E100 amine 8.5; basic pKa 2.0 for anthranilic acid, 7.8 for lidocaine, 9.8 for dibucaine) and took ATR-FTIR spectra of melts at equimolar functional groups vaneerdenbrugh2011a P4, P7, R1, R2, R8.

What the data show

In quercetin–E100 both amine bands were sharp in pure E100 and at 10% quercetin, smaller at 30% and gone at 50%, while a broad OH band near 3400 cm⁻¹ grew; for naringenin–E100 "similar" spectra are described but not shown, and no polyphenol-side marker (such as a phenolate band) is reported wegiel2013a R6, C17, X10. The carbonyl of E100 did not shift (0 cm⁻¹ with both polyphenols) wegiel2013a R4. Outside the chapter the same two bands were lost with resveratrol (absent at 50% resveratrol) and with curcumin (eradicated at 60% curcumin, with a rich dark red E100 dispersion against orange ones with the other polymers) wegiel2013b R7, C15; wegiel2014 R7, R10, C16, C20. Protonated lumefantrine at 25% drug loading before dissolution: HPMCAS 19 (6), HPMCP 46 (4), CAP 55 (3), Eudragit L 100 49 (6) and PVPVA 0%; after dissolution 3 (5), 21 (7), 54 (4), 12 (4) and 0% hiew2022a R20. The subtraction spectra show extra bands at 1594–1560 and 1410–1343 cm⁻¹, the asymmetric and symmetric carboxylate stretches, in all four enteric dispersions hiew2022a R19. An earlier XPS study cited by the authors had found no protonation of lumefantrine with HPMCAS at 20 and 40% drug loading and 50–75% with HPMCP hiew2022a R30. In the screen, anthranilic acid stayed amorphous with PSSA (average amorphicity index about 92) far better than with PAA (about 60) and ATR-FTIR was reported to show an acid–base reaction with PSSA and not with PAA; lidocaine and dibucaine showed proton transfer with PSSA and not with PAA, and bifonazole showed no acid–base reaction with the acidic polymers vaneerdenbrugh2011a R3.e, R8, R15, R16, R17, C22, C38, C44. The authors give pKa as the reason: the acidic pKa of PAA (4.5) and the basic pKa of flufenamic acid (−2.1) "apparently prevent" such a reaction in the other pairs vaneerdenbrugh2011a C23, R9. Outside the chapter, cationic drugs formed a complex with HPMCAS-MF in pH 6.0 buffer at a 1:1 drug-to-carboxyl ratio that took ≥70% of the polymer out of solution for the cationic drugs abbreviated LOR, CNZ and CLO, against 5–17% for unionized or anionic drugs bapat2024 C19.

Figure

The two sharp E100 bands near 2820 and 2770 cm⁻¹ shrink as quercetin is added and are gone at 50% wegiel2013a R6.

Zero with PVPVA; 19–55% with the enteric polymers, falling after dissolution except for CAP hiew2022a R20.

Why

The authors' interpretation: only E100, with its basic dimethylamino group, can form ionic interactions with the weakly acidic polyphenols, and ionic interactions are "typically stronger" than hydrogen bonds (an assumption, not measured) wegiel2013a C16. The lack of E100 hydrogen bonding is explained by ionization of the dimethylamino group wegiel2013a C33. For lumefantrine the extra carboxylate bands "provide evidence of an acid–base interaction", and the protonation seen with all four enteric polymers differs from the earlier HPMCAS result hiew2022a C25, C26, C48. The screen authors take the pKa gap as the deciding factor: PSSA (pKa −1.5) with the basic amine or ring nitrogen of anthranilic acid, lidocaine and dibucaine, and not PAA (pKa 4.5) vaneerdenbrugh2011a C22, C23, C38, C44. Not measured: the strength of any ionic contact, the ionization state of the polyphenols in the dispersion, and the ionization at the hydrated surface (the lumefantrine values are from dry compacts) wegiel2013a X10; hiew2022a C47, X13.

Conditions and limits

All spectra are from dry films or compacts; the E100 evidence is shown for quercetin only, and the PSSA evidence rests on spectra that the paper does not show and on melts, not films wegiel2013a X10; vaneerdenbrugh2011a X13. XPS protonation was measured at 25% drug loading only and after different dissolution times for the PVPVA and enteric compacts hiew2022a X14.

Counterexamples and conflicts

(a) An ionic contact did not guarantee stability: the 50% drug-loading HPMCAS dispersion, with 1.61 mol lumefantrine per mol of polymer acid groups, crystallized within a week hiew2022a R6, R26, C63. (b) The causal role of the ionic contact in the stability of the E100 dispersions was not separated from other properties of E100 wegiel2013a X10. (c) The earlier XPS study found no protonation with HPMCAS where this paper found 19% hiew2022a R30, R20. (d) Bifonazole, a base, was best inhibited by the acidic polymer PAA without any acid–base reaction by ATR-FTIR (spectra not shown), so inhibition does not need a proton transfer (see F13) vaneerdenbrugh2011a R16; kestur2011 R7.

Transfer check

For a drug with a basic or acidic centre and an ionizable polymer: compare the pKa of the drug group with that of the polymer group; mix at the intended ratio by the intended route; look for a spectral marker of ionization in the dry dispersion (a lost amine band, carboxylate bands, an N 1s component) and, where the polymer must dissolve, compute the moles of drug per mole of polymer acid group, since an excess of drug over acid groups coincided with early crystallization wegiel2013a R6; hiew2022a R19, R20, R26. Expect inhibition of crystallization and slower polymer dissolution (see F26).

History

Van Eerdenbrugh and Taylor (2011) first invoked acid–base reaction for the exceptions of their screen. Wegiel and co-workers took the E100 amine bands as the marker for resveratrol (2013), curcumin (2014) and, in the paper of this finding, naringenin and quercetin. Hiew and co-workers (2022) measured protonation of a weak base by XPS and carboxylate IR bands, extending the picture to enteric polymers hiew2022a C48. Bapat and co-workers (2024) later tied complexation of cationic drugs with HPMCAS to the release of the polymer (evidence index, citation links; bapat2024 C29).

Related: F13, F23.

How it was shown

Marsac, Shamblin and Taylor treated the amorphous drug as a solvent on a Flory–Huggins lattice whose site is one drug molecule (nifedipine 288.62 cm³/mol, from molecular weight 346.34 g/mol and measured density 1.20 g/cm³), took volume fractions from weight fractions and true densities, and computed the entropy of mixing from Eq. 1 with χ = 0 for nifedipine with methyl pyrrolidone and with PVP K12 (molecular weight 2,500), K29/32 and K90 (1,250,000) marsac2006b E1, E7, C9, P5, P6, R1, M1, M3, M5. The follow-up paper used the same lattice expression (Eq. 10) with χ from its two experimental routes (see F7 and F8), assumed χ independent of composition, and drew ΔG/RT against polymer fraction for five PVP pairs and sucrose–dextran, at an unstated PVP grade and temperature marsac2009 E10, P8, X12.

What the data show

For nifedipine with methyl pyrrolidone the entropy curve has a far deeper minimum (near Φ ≈ 0.44) than for any PVP grade (minima near Φ ≈ 0.6); K12 lies slightly below K29/32 and K90, which coincide at every composition (read from Fig. 1, which has no ordinate scale) marsac2006b R1, X7. Predicted crystalline solubilities show the same saturation with chain length: indomethacin falls from 0.396 weight fraction in the ethyl-pyrrolidone analogue to 0.144 in PVP K12 and 0.134 in K29/32 and K90, which the authors explain by the term 1/m_drug − 1/m_polymer approaching 1/m_drug as the chain grows marsac2009 R9.e, C24. In the follow-up curves the indomethacin minimum is about 2.4 times (melting-route χ) and about 5 times (solubility-route χ) as deep as the athermal minimum, whereas for felodipine and ketoconazole the three curves nearly coincide, and the minima lie at polymer fractions of about 0.5–0.6 marsac2009 R10. The authors give typical molecular-weight ranges of 200–600 for drugs and 10,000–1,000,000 for pharmaceutical polymers marsac2009 R13. Elsewhere in the library, Flory–Huggins theory with data for comparable PVP systems predicted a significant lowering of small-molecule activity only when the polymer volume fraction exceeded 0.5, and the activity curves of nifedipine and felodipine in PVP were "very similar" and small over the range studied konno2006 R20, C35; marsac2006a R19. For two polymers the picture differs: the calculated combinatorial entropy per lattice site (25 °C, no enthalpy term) has its minimum at about −27 × 10⁻²³ J for the 783 Da maltodextrin with PVP K 12 and about −22 to −21.5 for PVP K 25, K 30 and K 90, at −9.1 (K 12) to −2.8 (K 90) for a 6 kDa dextran, and at about zero for the 450 kDa dextran with PVP K 90 vaneerdenbrugh2012a R15, E5.

Figure

Panel (a) is nifedipine with methyl pyrrolidone, (b) to (d) PVP K12, K29/32 and K90; compare the shapes only, because the vertical axis carries no numbers marsac2006b R1, X7.

Why

The authors' explanation: the small entropy of mixing with a polymer arises from the reduced configurational entropy of the polymer, caused by the connectivity of its repeat units ("This is because") marsac2006b C11. They generalize that the entropy contribution is "predicted to be relatively constant" for typical drug and polymer sizes, and that "this observation leads to the inference" that the magnitude of the enthalpic interactions, the balance of adhesive against cohesive contacts, essentially decides miscibility marsac2006b C12, C13. A pair can tolerate a certain maximum unfavourable enthalpy of mixing, and predicting miscibility therefore needs the magnitude of χ marsac2009 C26. Dextran with ibuprofen or indomethacin and Eudragit E100 with itraconazole are offered, with the hedge "it can be speculated", as pairs where entropy cannot offset unfavourable contacts marsac2009 C43, C44. Van Eerdenbrugh and Taylor (2012) use the same lattice expression with a division by the chain length of each polymer and state that the combinatorial term always favours mixing, so that a miscibility gap needs a positive enthalpy ("it is anticipated") vaneerdenbrugh2012a C32, E5.

Conditions and limits

Entropy curves exist for nifedipine only; the claim for molecular weights up to 1,500,000 and drugs up to 600 is an extension marsac2006b X8, C12. PVP K12 (molecular weight 2,500) lies below the stated polymer range of 10,000 and above marsac2006b X7, R13. The critical χ for miscibility depends on how the lattice is defined and is system specific marsac2006b C18. Dispersions that take up water are outside these binary curves: with a third component the entropy still favours mixing, and demixing is expected only when the enthalpy of mixing is positive enough, which is set by three χ values rumondor2010c C27. The two-polymer curves assume that χ does not depend on molecular weight, use the PVP monomer as lattice site, and two of the 20 plotted curves do not follow from the stated inputs vaneerdenbrugh2012a C33, P6, X1.

Counterexamples and conflicts

None found in the library for binary drug–polymer systems; the claim that grade matters little is shown by simulation, not tested against a measured entropy or against a series of grades of one polymer marsac2006b X8, R1. For pairs of two polymers the claim does not hold: the dextran–PVP calculations vary with both chain lengths and the DSC gap widens with both (see F22) vaneerdenbrugh2012a R13, R15.

Transfer check

For a new pair, do not expect a lower-molecular-weight grade of the same polymer to make an immiscible system miscible; compare the interaction (χ from F7, IR from F2) first and, if the pair does not mix, switch polymer chemistry marsac2006b C14. Compute the Flory–Huggins curve for the drug and polymer volumes of the pair, using the lattice-site convention of the source you compare with. This holds for a small-molecule drug with a polymer; when both components are polymers, lowering the molecular weight of either did bring miscibility in the dextran–PVP blends (see F22) vaneerdenbrugh2012a R13.

History

The library's first Flory–Huggins entropy simulation for a drug–polymer pair is this 2006 paper marsac2006b C6, C9. Konno and Taylor (2006) had already used Flory–Huggins theory with data for comparable systems to argue that little activity lowering is expected at the polymer levels they used konno2006 C35. Marsac and co-workers (2009) repeated the argument with measured χ and concluded that differences between compounds come from χ marsac2009 C31. Van Eerdenbrugh and Taylor (2012) extended the entropy argument to pairs of two polymers and found that miscibility fell as both chain lengths rose (see F22) vaneerdenbrugh2012a C38, C39. The title was narrowed by the papers of 2011–2022: the statement that the entropy is nearly the same for any grade holds for a drug of molecular weight 346 with PVP K 29/32 and K 90 marsac2006b R1, not for two polymers vaneerdenbrugh2012a R15.

Related: PEG dispersions F2.

How it was shown

Table I gives molecular weight, measured helium-pycnometry density, molecular volume and the Fedors group-contribution solubility parameter of nifedipine (346.34 g/mol, 1.20 g/cm³, 288.62 cm³/mol, 22.9), felodipine (384.26, 1.28, 300.20, 25.0) and PVP K12 (2,500, 1.11, 2,252.25, 27.4) marsac2006b M1, M2, M3, P4, P5. Eq. 2 turns the squared difference of solubility parameters, times the lattice-site volume over RT, into χ, which cannot be negative marsac2006b E2. Free-energy-of-mixing curves of nifedipine–PVP K29/32 were drawn for χ = −4.2, −3.8, 0.5, 1.0 and 2.0 marsac2006b R2, A4. Miscibility of both drugs with PVP at all concentrations, and extensive nifedipine–PVP hydrogen bonding, are taken from a companion study marsac2006b R13, R14. The melting-route χ is described in F7.

What the data show

In the simulated curves, χ = 2.0 is positive everywhere (maximum near Φ ≈ 0.37), χ = 0.5 and 1.0 are shallow and slightly negative, and χ = −3.8 and −4.2 are deep with one minimum near Φ ≈ 0.54 (read from Fig. 2) marsac2006b R2. The simulations put the χ needed for miscibility over the entire range at about 0.6 or less marsac2006b R3. The solubility-parameter route gave 2.0 for nifedipine–PVP and 0.5 for felodipine–PVP, and the card recomputed 2.36 and 0.70 at 298.15 K, or 1.58 and 0.50 at each drug's melting temperature, without finding the printed pair at any single temperature marsac2006b R4, X1. Melting point depression gave −3.8 and −4.2 marsac2006b R9. Only nifedipine–PVP K29/32 was simulated, with felodipine's χ drawn on the nifedipine system, and the temperature of the simulation is not stated marsac2006b X9. A third route in the library, the fit of the ternary Flory–Huggins model to moisture sorption of dispersions, gave χ for drug–PVP ranging from −9.6 to 26.9 (felodipine) and from −20.4 to 22.0 (indomethacin), which its authors call ambiguous rumondor2010a R14, C30.

Figure

Curve (e), χ = 2.0, stays above the baseline (immiscible); (c) and (d) are shallow; (a) and (b), χ = −4.2 and −3.8, have one deep minimum (miscible). The vertical axis has no numbers marsac2006b R2.

Why

The authors' interpretation: Eq. 2 assumes the unlike interaction equals the geometric mean of the like ones, which holds for van der Waals forces but not for directional ones such as the nifedipine–PVP hydrogen bonds, so the discrepancies "most likely arise" from using solubility parameters for systems with specific interactions, and such estimates "are likely to be inaccurate" for the many drugs and polymers that form them marsac2006b C24, C25. The positive value predicts endothermic mixing and immiscibility, "in conflict" with the reported miscibility marsac2006b C23. The hydrogen bonding itself is cited, not measured here marsac2006b R14.

Conditions and limits

Two dihydropyridine drugs and one polymer family, PVP; χ from melting data refers to temperatures close to the drug melting point (about 120–172 °C) and is applied without a temperature correction marsac2006b C32, X11. The positive solubility-parameter value is a statement about sign; the numbers 2.0 and 0.5 are unverified marsac2006b X1.

Counterexamples and conflicts

None found in the library in favour of solubility parameters for hydrogen-bonding pairs. Solubility-parameter differences had been advocated for pharmaceutical miscibility before this paper marsac2006b C22. The ternary-sorption route is a second method that failed, for a different reason: because the χ₂₃ term is small, small changes in the water–drug or water–polymer parameters produce large changes in it rumondor2010a C29, C30.

Transfer check

Before using a solubility-parameter difference to rank polymers, check whether the drug or the polymer has donor or acceptor groups, and whether IR shows a carbonyl or NH shift (see F2); if so, expect the sign to be wrong for some pairs. Test the prediction on one pair with a measured miscibility reference, and prefer a measured χ from F7 marsac2006b C25.

History

Marsac and co-workers (2006) state that melting point depression had been used for polymer blends and polymer–solvent systems but not for drug–polymer systems marsac2006b C26, and they obtained χ from it for two drugs and compared it with the solubility-parameter route marsac2006b C36; no earlier paper of this chapter does so (synthesis). In 2009 the same group used two experimental routes instead marsac2009 C45. Rumondor and co-workers (2010) found the sorption route unable to give unambiguous χ rumondor2010a C30.

Related: PEG dispersions F3.

How it was shown

In 2006 crystalline drug and polymer particles (45–75 µm) were dried over phosphorus pentoxide for at least a week and blended geometrically (nifedipine with 5–25 wt% PVP K12, felodipine with 5–25 wt% in finer steps, two to three samples per composition), and heated at 1 °C/min in a modulated DSC under nitrogen; melting was taken as the extrapolated onset marsac2006b P1. A first attempt with PVP K29/30 at 10 °C/min showed no depression; slower scanning, controlled particle size and the lower-Tg PVP K12 revealed it marsac2006b P2, R11. χ followed from the slope of the rearranged Flory–Huggins relation plotted against Φ_polymer², over 0–0.25 marsac2006b P7, E3, E10, R8. In 2009 the particles were 45–90 µm, the offset of the melting endotherm was taken (the paper argues that, in theory, it gives the melting point at the final composition), and nine pairs were tested: five with PVP K12, three with dextran (sucrose, ibuprofen, indomethacin) and one, itraconazole, with Eudragit E100 marsac2009 P1, P6, C16, R1. A premixed felodipine–PVP K12 dispersion (50 wt% felodipine, made from ethanol) was used as the polymer phase to test mixing during the run marsac2009 P3, A4.

What the data show

2006 onsets (read from Fig. 3b) at Φ_polymer of 0, ≈0.057, ≈0.114, ≈0.169, ≈0.224, ≈0.277: nifedipine ≈171, 170, 168, 165.5, 160.5, 156.5 °C and ≈148 and ≈144.5 °C at ≈0.331 and ≈0.383; felodipine ≈141, 138, 134, 131 and then scattered between ≈120 and ≈126 °C from Φ ≈ 0.224 to ≈0.331 marsac2006b R6. With χ = −3.8, Eq. 3 followed the nifedipine onsets to within about 1 °C (≈157 against ≈156.5 °C at Φ ≈ 0.277); χ = 0 gave ≈168.4 °C and χ = 2.0 a rise to ≈175.8 °C at Φ = 0.30 marsac2006b R10. χ values are −3.8 and −4.2 marsac2006b R9; the nifedipine volume fractions seem to use felodipine's density, so the second digit of −3.8 is uncertain marsac2006b X2. 2009 offsets for indomethacin at Φ ≈ 0, 0.06, 0.12, 0.175, 0.23: ≈161.5, 160, 159.5, 157.5, 155 °C, against onsets ≈160, 157.5, 152, 146, 135.5 °C; the athermal line gives ≈158.6 °C at Φ = 0.25 marsac2009 R2. The offset χ with PVP K12 are −0.82, −0.08, 0.00, −0.08 and 0.02 for indomethacin, felodipine, nifedipine, ketoconazole and sucrose, and 0.20 (sucrose–dextran) and the lower limits 0.38, 0.72 and 0.11 (ibuprofen–dextran, indomethacin–dextran, itraconazole–Eudragit E100) marsac2009 R7. The felodipine measurement against the premixed dispersion gave a depression "slightly lower" than against pure PVP K12 (data not shown), which the authors read as slow mixing causing underestimated interaction; as worded, a smaller depression with the premix would point the other way marsac2009 R8, C21, X6. At a polymer level of 3 wt% there was no depression at all: the offset melting point of felodipine was 140.7, 142.9, 141.4 and 142.6 °C with PVP, PVP/VA, HPMCAS and PVAc against 141.5 °C alone kestur2010a R1. For the pair indomethacin–PVPVA a different group obtained −1.97 and quotes four literature values from −0.64 to −4.5 moseson2018 R4, R7.

Figure

Left, traces; right, onset (squares) and offset (diamonds) melting points with fitted lines and the entropy-only line (dashed). Diamonds below the dashed line mean favourable mixing (negative χ); the onset falls much further than the offset marsac2009 R2, R6.

(a) The straight line whose slope is χ = −3.8; (b) measured nifedipine onsets against Eq. 3 with χ = 2.0, 0 and −3.8 marsac2006b R8, R10.

Why

The authors' interpretation: melting point depression is "a manifestation of the reduced thermodynamic activity of the drug in the presence of the polymer": the dispersed drug reaches the chemical potential of the crystal at a lower temperature marsac2006b C39. The method needs melting before decomposition and a polymer that is liquid-like at the drug melting point, so it suits polymers with Tg well below the drug Tm marsac2009 C6. The absence of depression with PVP K29/30 is "most likely" due to lack of physical interaction resulting from its Tg of 168 °C, below which it is not liquid-like at the drug melting points of 171 and 142 °C marsac2006b C29. The loss of linearity at higher polymer content "most likely reflects" the composition dependence of χ and slower mixing as Tm approaches the polymer Tg marsac2006b C31. The equation is linear only at low polymer fractions, so χ is a composite value over that range marsac2009 C7. Crystal and polymer meet only at particle interfaces and mixing after melting may still be kinetically limited marsac2009 C8. Differences between studies of one pair were attributed to sample preparation, DSC method and onset against offset moseson2018 C21.

Conditions and limits

χ from this route refers to temperatures near the drug melting point (about 120–172 °C in 2006) and to low polymer fractions marsac2006b C32; marsac2009 C7. The polymer must be a low-Tg grade: K12 worked and K29/30 did not at the rates used marsac2006b C29, R11. The 2009 positive values for non-depressing pairs are lower limits, since an immiscible polymer leaves the melting point unchanged, and how a finite value was obtained is not described marsac2009 C19, X11. The felodipine fit range, the pure-drug melting point entered in Eq. 2 and the volume of the lattice site are not printed marsac2009 X15, X4, P6. Sucrose degrades on melting marsac2009 C14. Dextran, with a Tg around 140 °C, "may" give an underestimated depression marsac2009 C20.

Counterexamples and conflicts

(a) The same sample read three ways gives χ of −6.16, −1.94 and −0.82, a 7.5-fold spread for one pair (synthesis: 6.16/0.82) marsac2009 R6. (b) The scale of χ is not shared between groups: indomethacin–PVPVA gave −1.97 here and −0.64 to −4.5 elsewhere moseson2018 R4, R7. (c) The direction of the felodipine premix result is unverified marsac2009 X6. (d) At 3 wt% polymer no depression is seen, so the method gives no χ for the low polymer levels of the growth studies in F10 (synthesis: from the unchanged offset melting point) kestur2010a R1.

Transfer check

Use a polymer whose Tg lies well below the melting point of the drug (PVP K12 rather than K29/30), small sieved particles, a slow scan (1 °C/min in these papers, or a series extrapolated to 0 °C/min), and report onset, midpoint and offset; include the entropy-only (χ = 0) line and judge a pair by whether the offset falls below it marsac2006b P1, R11; marsac2009 R2, R6; moseson2018 R5. Compare χ only within one protocol moseson2018 C21.

History

Marsac and co-workers (2006) first applied the lattice melting-depression equation to drug–polymer pairs and obtained −3.8 and −4.2 marsac2006b C26, R9. Marsac and co-workers (2009) extended it to nine pairs, introduced the offset and the miscible/immiscible split marsac2009 C12, C16. Zhu and co-workers (2011) used the same relation for drugs in PEG 3350 at 50–80% drug zhu2011 E1, P5. Moseson and co-workers (2018) extrapolated the offset to zero scan rate and used the resulting χ to predict processing temperatures moseson2018 R5, C22.

Related: PEG dispersions F3.

How it was shown

For the solubility route, excess crystalline compound was stirred in EP at 25 °C for at least 24 h in the dark, centrifuged, filtered and assayed by UV–Vis (felodipine, nifedipine, ketoconazole, indomethacin) or HPLC with refractive-index detection (sucrose) marsac2009 P2. The ideal solubility came from Eq. 1 with literature heats of fusion, melting points and heat capacities, and γ in EP is the ideal divided by the measured solubility; Eq. 8 then corrects γ for the smaller entropy of mixing with PVP, under the assumptions of ideal mixing entropy with EP, identical drug–EP and drug–PVP contacts, and equal γ at both solubility limits marsac2009 P7, E1, E8, C11. The melting-route χ is described in F7. In 2006 the same group had transferred the melting-route χ to methyl pyrrolidone with ideal entropy and compared the predicted with the measured solubility at 25 °C marsac2006b P9, P3, C46, C47.

What the data show

Solubility in EP (weight fraction, mole fraction, γ, χ): felodipine 0.268, 0.099, 0.89, −0.08; nifedipine 0.215, 0.083, 0.26, −0.81; ketoconazole 0.079, 0.018, 0.68, −0.12; sucrose 0.074, 0.026, 0.37, −0.53; indomethacin 0.396, 0.173, 0.15, −1.83 marsac2009 R9. Melting-route χ: −0.08, 0.00, −0.08, 0.02 and −0.82 marsac2009 R7.a, R7.b, R7.c, R7.d, R7.e. The absolute differences are 0.00, 0.81, 0.04, 0.55 and 1.01 (synthesis: nifedipine |−0.81 − 0.00|, sucrose |−0.53 − 0.02|, indomethacin |−1.83 + 0.82|), and the orders are indomethacin < felodipine = ketoconazole < nifedipine < sucrose (melting) against indomethacin < nifedipine < sucrose < ketoconazole < felodipine (solubility) marsac2009 X7. In 2006 the melting-route χ predicted 0.15 (nifedipine) and 0.48 g/g (felodipine) in methyl pyrrolidone against 0.21 (± 0.02) and 0.45 (± 0.07) g/g measured, "good agreement" for felodipine and about 30% low for nifedipine, whereas χ = 0.5 would have given 0.004 g/g, over an order of magnitude below the measurement marsac2006b R17.c, R17.d, C48, C49, X10. The ternary-sorption route gave unusable ranges and the same indomethacin–PVPVA pair gave −0.64 to −4.5 in four other studies rumondor2010a R14, C30; moseson2018 R7.

Figure

Left columns: measured weight and mole fraction, γ and χ in EP; right columns: predicted weight fraction and γ in PVP K12, K29/32 and K90 marsac2009 R9.

Why

The authors' interpretation: the solubility route gives mixing thermodynamics at room temperature and one composition, the solubility limit, whereas melting gives them near the melting point and over low polymer fractions; slow mixing in the melting experiment may bias χ marsac2009 C27, C28. The EP route assumes that interactions with EP match those with a polymerized version of it, and describes an equilibrium liquid, not a glass marsac2009 C9, C33. Because γ depends on temperature and composition, solubility in a polymer is "best estimated" from γ in an analogue liquid marsac2009 C32.

Conditions and limits

Five compounds, one PVP grade for the melting route, EP at 25 °C; inputs (heat capacities, ideal solubilities, lattice volume, number of replicates) are not printed, so neither χ set can be recomputed from the paper marsac2009 X4, P7, P2. The table T_M values are 2–3 °C below the plotted pure offsets marsac2009 X4.

Counterexamples and conflicts

(a) The authors' statement that four compounds give "slightly negative or close to zero" χ in both routes does not hold for nifedipine (−0.81) and sucrose (−0.53) in EP marsac2009 C38, X8. (b) The two routes disagree on which of felodipine and ketoconazole is more favourable marsac2009 X7. (c) In PEG the magnitudes of readings of drug–polymer interaction also differ between papers for the same phase diagram, though the signs agree for most drugs baird2011 R9.a; zhu2011 R5.c.

Transfer check

Treat χ from either route as a ranking tool within one method, not as an absolute input for a design calculation, and do not mix values across routes or papers marsac2009 X7, X16. Where possible use both routes on the same compounds: agreement in sign is the usable part, agreement in size is not.

History

Marsac and co-workers (2006) validated the melting-route χ against a measured solubility in methyl pyrrolidone marsac2006b C48. In 2009 they added the EP route and compared the two marsac2009 C45. The PEG chapter later found the same pattern of agreement in sign and disagreement in size for readings of drug–PEG interaction baird2011 R9; zhu2011 R5.

Related: PEG dispersions F3.

How it was shown

The driving force was computed with the Hoffman equation for pure amorphous nifedipine from 0 to 180 °C and, with the depressed melting temperature of the dispersion, for 30 and 40% PVP marsac2006b P8, E4, R12 (formula page: Hoffman equation). Solubility of the crystal in a polymer was obtained by solving the solid–liquid equilibrium equation with the Flory–Huggins activity coefficient and the melting-route χ, for PVP K12, PVP K29/32 and methyl pyrrolidone (with ideal mixing entropy); the heat-capacity terms and the temperature are not printed marsac2006b P9, E5, E6, X11. In 2009 the activity coefficient measured in EP was scaled to PVP K12, K29/32 and K90 with Eq. 8, and the predicted solubility in PVP K29/32 was compared with drug loading to give the fraction of drug "available for crystallization" marsac2009 P7, P9, E8. The nucleation argument rests on a companion study that found pure nifedipine and felodipine nucleate much faster than dispersions with even small amounts of polymer marsac2006b R16.

What the data show

Free energy of crystallization of nifedipine (read from Fig. 5): about −9.45, −9.0 and −5.45 kJ/mol at 0, 20 and 100 °C for the pure drug; −9.1, −8.4 and −4.1 for 30% PVP; −8.6, −7.95 and −2.7 for 40% PVP, reaching zero at about 171, 148 and 128 °C; at 20 °C the pure drug and the 40% PVP dispersion differ by about 1.05 kJ/mol (synthesis: 9.0 − 7.95) marsac2006b R12. The 40% PVP curve rests on a melting temperature that is not shown in the paper's melting data marsac2006b X5. Predicted solubility (g drug per g polymer): nifedipine 0.06 in PVP K29/32 and 0.07 in K12, felodipine 0.25 and 0.31; with χ = 0.5, nifedipine would be 0.001 in both; in methyl pyrrolidone the prediction is 0.15 and 0.48 and the measurement 0.21 and 0.45 g/g marsac2006b R17. In 2009, predicted weight-fraction solubilities in PVP K12, K29/32 and K90 are 0.072, 0.065, 0.065 (felodipine), 0.054, 0.049, 0.049 (nifedipine), 0.011, 0.010, 0.010 (ketoconazole), 0.016, 0.015, 0.015 (sucrose) and 0.144, 0.134, 0.134 (indomethacin) marsac2009 R9. The fraction of drug above solubility in PVP K29/32 reaches zero only at about 0.86 (indomethacin), 0.93 (felodipine), 0.95 (nifedipine) and 0.99 (sucrose, ketoconazole) polymer, and at 0.50 polymer is about 0.82, 0.93, 0.95, 0.99 and 0.99 (read from Fig. 6) marsac2009 R11. Other library papers use these numbers: felodipine solubility of about 0.25 g/g dry PVP is cited as the value that moisture would lower by at least an order of magnitude, and "all the dispersions studied are supersaturated" with respect to the estimated solubility in PVP rumondor2009b R26; marsac2008a C47. At 3% polymer the felodipine melting point is unchanged, so the authors there see no macroscopic effect on the driving force kestur2010a C36, R1.

Figure

Pure drug (a), 30% PVP (b) and 40% PVP (c): the curves differ by only 1–2 kJ/mol at room temperature marsac2006b R12.

From left, indomethacin, felodipine, nifedipine, sucrose, ketoconazole; each line reaches zero at the predicted solubility marsac2009 R11.

Why

The authors' reasoning, marked as their interpretation: a miscible drug has a lower chemical potential than the pure amorphous drug, so the driving force must fall, and melting point depression is its visible sign marsac2006b C38, C39. The estimates "do not forecast a large reduction", which they think is "most likely underestimated" because equilibrium depression is hard to reach at high polymer content marsac2006b C41, C40. Since pure drugs nucleate much faster than dispersions, the reduced driving force "seems unlikely" to explain the slower crystallization, "suggesting" kinetic and/or interfacial factors marsac2006b C42. Solubility in PVP is lower than in methyl pyrrolidone because a polymer adds less mixing entropy, so estimates from a small analogue "need to be corrected" marsac2006b C50, C51. High-melting, high-ΔHfus drugs without strong polymer interactions "will display very low solubility" and will "most likely be harder to stabilize" marsac2006b C53, C54. Most dispersions therefore contain drug above its solubility, and any apparent stability is "most likely achieved through modification of the kinetics"; a small dissolved fraction "may help explain" why crystallization is not always complete marsac2009 C34, C35. Miscibility is not solubility: crystals in a dispersion do not necessarily mean the two liquids are immiscible marsac2006b C43.

Conditions and limits

Two dihydropyridine drugs (2006) and five compounds (2009) with PVP; χ from melting near 120–172 °C is used without a temperature correction, and the temperature and heat-capacity terms of Table II are not printed marsac2006b C32, X11. No solubility in a polymer was measured; the analogue check is about 30% low for nifedipine, and the 2009 predictions rest on the EP assumptions (ideal entropy, identical contacts, equal γ at both limits) marsac2006b X10; marsac2009 C11, X17. The two-drug comparison of melting properties is confounded by the different χ of the two drugs marsac2006b X13. No crystallization, nucleation or mobility was measured in either paper marsac2006b X12; marsac2009 X17.

Counterexamples and conflicts

(a) Being below the estimated solubility did not keep a dispersion amorphous: 20/80 benzocaine and ibuprofen dispersions in PEG lay below their estimated solubilities in supercooled PEG and crystallized within hours, because crystalline PEG nucleates them zhu2012 X13; zhu2011 R5.b, R5.d. (b) The driving-force reading of neat-drug differences is only partial in the library: the driving force pointed the right way for nifedipine against felodipine but the wrong way for indomethacin against felodipine and for the vemurafenib polymorphs marsac2006a C15; alonzo2010 C42; lu2016 C52. (c) For ibuprofen and fenofibrate in PEG the melting-point depression should lower the driving force, yet crystallization began earlier than for the neat drug zhu2012 C34.

Transfer check

Before choosing a drug load, compute or measure the solubility of the crystalline drug in the polymer at the storage temperature, with the activity coefficient from a liquid analogue (and scale it for the polymer entropy) or from the melting route, and compare it with the load; expect a load above solubility unless the polymer content is very high or the drug has a low melting point and an interaction (indomethacin was the most soluble of the five) marsac2009 R9, C32; marsac2006b C54. A supersaturated dispersion then depends on slow molecular motion, so test it with storage well below Tg and low moisture, as the thermodynamic estimate does not predict crystallization marsac2006b C56.

History

Marsac and co-workers (2006) proposed the solubility and driving-force argument and concluded that stabilization is predominantly kinetic marsac2006b C56. Marsac and co-workers (2008) used the same estimates to say that all the dispersions in their moisture study were supersaturated marsac2008a C47. In 2009 the analogue-solvent route extended the estimate to five compounds and PVP grades marsac2009 C46, C34. The PEG chapter later applied the same logic with a van't Hoff extrapolation of the liquidus zhu2011 R5.

Related: Crystallization tendency F6, Crystallization tendency F17, PEG dispersions F4.

How it was shown

Van Eerdenbrugh and Taylor held the chemistry fixed and varied only chain length: four dextran grades (DEX1, maltodextrin 783 Da; DEX6, 6 kDa; DEX40, 40 kDa; DEX450, 400–500 kDa) and five PVP grades (K 12, K 17, K 25, K 30, K 90; 2–3, 7–11, 28–34, 44–54 and 1000–1500 kDa) at nine dextran weight fractions from 0.1 to 0.9 vaneerdenbrugh2012a M1–M9, A1, A2, A3. Aqueous solutions of 10% (w/v) total solids were freeze-dried in a 36.5 h cycle; the powders in pin-holed pans were heated at 10 °C/min to 250 °C (200 °C for DEX1 blends), cooled to 40 °C and reheated, and the midpoint Tg was read on the second heating (the first for DEX1 blends); one Tg event was classed miscible and two phase separated, the lower being assigned to the PVP-rich phase vaneerdenbrugh2012a P1, P2, P3, C17. FTIR used films spin-coated on KRS-5 (8000 rpm) from the same solutions and dried overnight under dry nitrogen at 0.2, 0.5 and 0.8 dextran for all pairs; the PVP carbonyl band (1680 cm⁻¹ free, 1661 cm⁻¹ hydrogen-bonded to dextran hydroxyls; PVP has no donor) was fitted as a mix of the pure-PVP band and the band of the maximally bonded reference, taken as DEX1–PVP12 at 0.9, with an absorptivity ratio of 1.3 vaneerdenbrugh2012a P4, P5, R2, C10, C14, C25:

f_{HB} = \frac{\frac{a_F}{a_{HB}}\,A_{HB}}{A_F + \frac{a_F}{a_{HB}}\,A_{HB}} vaneerdenbrugh2012a E1

with A_F and A_HB the fitted areas of the free and the maximally bonded band. For the thermodynamic argument the combinatorial entropy of mixing was calculated per lattice site at 25 °C with equal densities of 1.2 g/ml, the PVP monomer as lattice site and the mean of each molecular-weight range, and no enthalpic term vaneerdenbrugh2012a P6, E5:

-T\Delta S_{mix} = kT\left[\frac{\phi_1}{N_1}\ln\phi_1 + \frac{\phi_2}{N_2}\ln\phi_2\right] vaneerdenbrugh2012a E5

with φ the volume fractions and N the number of lattice sites per chain; the enthalpy of mixing per site, \Delta H_{mix} = kT\chi\phi_1\phi_2, was not evaluated because χ was not determined vaneerdenbrugh2012a E4, C33.

What the data show

Number of phase-separated compositions out of nine by DSC (Table III; DEX1 pairs all miscible):

vaneerdenbrugh2012a R13. In the PVP25 series the numbers of two-Tg compositions are 0, 4, 7 and 8 for DEX1, DEX6, DEX40 and DEX450; the gap sits at high dextran fractions for DEX6 (two Tgs at 0.5–0.8) and spreads over most of the range for the longer dextrans vaneerdenbrugh2012a R1, R4, R7, R10, C16, C24. Pure-polymer Tg rises with dextran length: about 136, 205, 224 and 229 °C for DEX1, DEX6, DEX40 and DEX450, against about 172 °C for PVP25 vaneerdenbrugh2012a R1, R4, R7, R10. Hydrogen-bonded fraction of PVP carbonyls at 0.8 dextran:

vaneerdenbrugh2012a R14. For DEX1–PVP12 the fraction rose from about 0.28 at 0.1 dextran to 0.62 at 0.3, 0.80 at 0.5, 0.96 at 0.7 and 1.00 at 0.8 and 0.9; the authors take the plateau as the maximum possible bonding, not as bonding of every carbonyl vaneerdenbrugh2012a R14.a, C14, C15. Where two Tgs appear, the dextran-rich Tg stays near that of the pure dextran (about 205–211 °C against 205 °C for DEX6–PVP25; 216–223 °C against 224 °C for DEX40–PVP25) while the PVP-rich Tg moves with composition, which the authors read as PVP-rich phases that take up more dextran than dextran-rich phases take up PVP; phase compositions were not measured, and in the one DEX40 series shown the PVP-rich Tg at 0.7–0.8 dextran is about 203–204 °C, some 30 °C above pure PVP25 (synthesis: 204 − 172 = 32) vaneerdenbrugh2012a R4, R7, C18, X5. The dextran OH band of DEX1–PVP12 moved from about 3356–3358 cm⁻¹ at 0.9–1.0 dextran to about 3370–3383 cm⁻¹ at 0.1–0.7, not smoothly and at 4 cm⁻¹ resolution vaneerdenbrugh2012a R16, X6. Entropy minima per lattice site (10⁻²³ J) for PVP12, PVP17, PVP25, PVP30 and PVP90: DEX1 about −27, −23, −22, −22, −21.5; DEX6 −9.1, −4.4, −3.3, −3.1, −2.8; DEX40 −3.1*, −2.2, −0.9, −0.7, −0.4; DEX450 −2.8*, −1.9, −0.6, −0.35 and about 0; the two starred PVP12 curves cannot be reproduced from the stated inputs (recomputed −7.07 and −6.76) vaneerdenbrugh2012a R15, E5, X1. With the stated inputs the curve depth would not order DEX40–PVP12 and DEX450–PVP12 against DEX6–PVP17 and DEX6–PVP25 by their observed miscibility vaneerdenbrugh2012a X11.

Earlier solid-state work cited in the paper agrees at the high-molecular-weight end: dextran of 520 kDa with PVP K 90 was immiscible by DSC, dextran of 64–76 kDa with PVP K 90 by XRPD, and dextrans of 8.5–11.5 and 4–6 kDa with PVP of 29 and 10 kDa were dominated by phase separation, whereas freeze-dried maltodextrins of 860 and 1050 Da with PVP K 90 were miscible vaneerdenbrugh2012a R17. A later AFM-IR study of 50/50 films of the same pairs found two phases in all five DSC-immiscible blends and homogeneous films in three of four DSC-miscible ones, with DEX6–PVP90, miscible by DSC, showing domains of about 0.3–0.6 µm vaneerdenbrugh2012c R1, R9, R12, R14, R16, R22.

Figure

Black cells are miscible, grey phase separated and hatched cells inferred; the grey block grows from DEX1 (left) to DEX450 (right) and, within each dextran, towards the longer PVPs vaneerdenbrugh2012a R13.

The DEX1–PVP12 reference curve is drawn in every panel; the points fall further below it from DEX1 (a) to DEX450 (d) and from PVP12 to PVP90 vaneerdenbrugh2012a R14.

Why

The authors' interpretation, in their order: mixing chains at random always gains entropy, so a miscibility gap needs a positive mixing enthalpy ("it is anticipated") vaneerdenbrugh2012a C32; the enthalpy is positive because dextran–PVP contacts replace dextran–dextran ones, and dextran–dextran hydrogen bonds "may stem from" being stronger or more numerous, with the blue shift of the dextran OH band in DEX1–PVP12 offered as support vaneerdenbrugh2012a C36, C37. With χ taken as independent of molecular weight ("if we assume") only the combinatorial term differs between pairs, and it shrinks as the chains grow (Eq. 4), so that "the higher the combined polymer molecular weights, the lower the miscibility" vaneerdenbrugh2012a C33, C34, C35, C38. The conclusions say it was "surmised" that the mixing entropy becomes less favourable and cannot offset the enthalpy; the abstract says the entropy "was observed" to decrease although it was calculated vaneerdenbrugh2012a C39, X7. The unequal phases are explained by PVP gaining hydrogen bonds on mixing while dextran exchanges dextran–dextran hydrogen bonds for dextran–PVP ones, which "probably" lowers its overall bonding vaneerdenbrugh2012a C18. The plateau of bonding below 1.00 is attributed to steric hindrance for part of the carbonyls ("likely") vaneerdenbrugh2012a C15. (synthesis: with a small-molecule drug the entropy term of a PVP of any grade above K 12 is nearly the same F5, so the chain-length effect seen here belongs to pairs in which both components are long.) Not measured: χ, the mixing enthalpy, the strength of dextran–dextran against dextran–PVP bonds, and the composition of the coexisting phases vaneerdenbrugh2012a C33, C36, X7.

Conditions and limits

Polymer–polymer pairs, no drug; freeze-dried from water; one chemistry vaneerdenbrugh2012a P1, P2, A1. DSC samples were heated to 200–250 °C and read on reheating, whereas the FTIR films were never heated, so the two methods saw differently prepared samples, and heating through Tg may itself change a temperature-dependent miscibility vaneerdenbrugh2012a X8, C3. No replicates or error bars are reported, and the DSC series of the 16 pairs without PVP25 and the spectra of the other pairs are in a supplement that was not available vaneerdenbrugh2012a X9. The f_HB values rest on references set by one system and on a free-PVP reference whose grade is not stated vaneerdenbrugh2012a X12. Humidity, where water competes for hydrogen bonds, and drug–polymer pairs were not studied vaneerdenbrugh2012a C40.

Counterexamples and conflicts

(a) The trend holds for the size of the gap but not cell by cell: within DEX6 and DEX40 the PVP90 gap is narrower than the PVP30 gap, and within DEX450 PVP17 has the widest gap after PVP90 vaneerdenbrugh2012a R13, X10. (b) Table III marks DEX450–PVP25 at 0.1 dextran as miscible by measurement while Fig. 5a shows two Tgs there, and it marks a DEX1–PVP25 point as inferred that Fig. 2a plots vaneerdenbrugh2012a X2, X3. (c) The plotted entropy curves for DEX40–PVP12 and DEX450–PVP12 do not follow from the stated inputs, and with the stated values depth alone does not order pairs across dextran grades vaneerdenbrugh2012a X1, X11. (d) The statement that PVP-rich Tgs differ little from pure PVP for the long PVPs does not match the one DEX40 series shown vaneerdenbrugh2012a C20, X5. (e) A pair that DSC classes as miscible can look phase separated in films (DEX6–PVP90) vaneerdenbrugh2012c R16.

Transfer check

For a blend of two polymers in one dispersion: (1) lower the molecular weight of one or both polymers before searching for a different chemistry, since a 783 Da dextran mixed with every PVP grade and DEX6 mixed with PVP12 at all compositions vaneerdenbrugh2012a R13; (2) map composition, not one ratio, because the gap sat at high dextran fractions for DEX6 and across most of the range for DEX40 vaneerdenbrugh2012a C16, C20; (3) read DSC together with a structure-sensitive method such as the PVP carbonyl fit, on samples prepared the same way vaneerdenbrugh2012a X8. Do not carry the pattern to a drug–polymer pair, where the entropy term of a small drug stays favourable (see F5).

History

Taylor and Zografi (1998) had found freeze-dried PVP K 90 miscible with maltodextrins of 860 and 1050 Da, Shamblin and co-workers (1998), Newman and co-workers (2008) and Padilla and co-workers (2011) found high-molecular-weight dextran–PVP blends immiscible, and the authors concluded that solid-state studies had used relatively high molecular weights vaneerdenbrugh2012a R17, C4. Van Eerdenbrugh and Taylor (2012) mapped one chemical pair from miscible to immiscible by varying both molecular weights and proposed the pairs as model systems for testing methods of judging miscibility vaneerdenbrugh2012a C40, C8. The AFM-IR study of the same pairs followed in 2012, and Purohit and Taylor (2015) cited it for the observation that the lower-viscosity phase forms discrete domains (evidence index, citation links; see F18).

Related: Solid-state methods F15, F5, F3.

How it was shown

Kestur and Taylor dissolved felodipine and each polymer (10% w/w polymer) in 1:1 dichloroethane–ethanol (as printed), rotary-evaporated and dried the dispersion for 48 h, then diluted it with crystalline felodipine in a cryogenic mill to 3% w/w polymer and in the same way to 0.5, 1.5 and 4.5%, because direct evaporation below 5% w/w would have let felodipine partly crystallize kestur2010a P1, C5, X8. The polymers were PVP K-12 (Mw 2000–3000), PVP/VA (Kollidon VA64, Mw 45 000–47 000, 40% acetate side chains, basis not stated), HPMCAS (AQOAT AS-MF, Mw 18 000) and PVAc (Mw 100 000) kestur2010a M2, M3, M4, M5, X11. Films of 3–5 mg were melted at 145–150 °C between cover glasses (30–50 µm thick), cooled rapidly to about 50 K below Tg, stored over P2O5 at 40 °C to form nuclei, and the nuclei were grown on a calibrated hot stage at 70, 80, 90, 95, 100, 105 and 110 °C; the spherulite diameter grew linearly with time and its slope was the growth rate, in triplicate kestur2010a P2, P3, P4, R4. Raman microscopy and the melting point identified the crystals as the metastable form II, although the starting material was form I kestur2010a P5, R3.

What the data show

Felodipine alone grew at about 3.5 × 10⁻⁹ m/s at 70 °C, about 6.9 × 10⁻⁸ at 90 °C and about 1.8 × 10⁻⁷ at 105 °C (read from Fig. 4); the text calls the fall from 105 to 70 °C 3 orders of magnitude, while the figure shows about 51-fold kestur2010a R5.a, X1. With 3% polymer, the inhibition ratios (rate without over rate with polymer, read from Fig. 5) fell from 70 to 110 °C: PVP from about 10.0 to 2.9, PVP/VA from 6.5 to 1.1, HPMCAS from 4.4 to 1.1, PVAc about 1.2 throughout kestur2010a R6. At 90 °C the rates at 0.5, 1.5, 3 and 4.5% w/w were about 4.2 × 10⁻⁸, 1.8 × 10⁻⁸, 1.0 × 10⁻⁸ and 4.7 × 10⁻⁹ m/s for PVP; 4.4 × 10⁻⁸, 2.5 × 10⁻⁸, 1.5 × 10⁻⁸ and 6.4 × 10⁻⁹ for PVP/VA; 5.2 × 10⁻⁸, 4.3 × 10⁻⁸, 2.9 × 10⁻⁸ and 1.7 × 10⁻⁸ for HPMCAS; and 5.9 × 10⁻⁸, 5.3 × 10⁻⁸, 5.0 × 10⁻⁸ and 3.8 × 10⁻⁸ for PVAc kestur2010a R8. The authors say HPMCAS needs 3% for any measurable decrease, although the plotted point at 1.5% already lies below the neat rate, and that more PVAc "appears" to have little effect, though 4.5% PVAc is about 1.8-fold below it kestur2010a C16, C15, X4. With polymer the maximum of the growth curve appears shifted to 110 °C, except for PVAc kestur2010a R5, C11. In a companion study with PVP K-29/32, 1.5 and 4.5 wt% reduced felodipine growth 2- and 6-fold at 110 °C and 3- and 35-fold at 70 °C kestur2010b R13. For the same polymers in other drugs: with 5% w/w polymers, flutamide growth fell by more than one order of magnitude with PVP, about 4-fold with HPMCAS and about 2.5-fold with HPMC trasi2012b R5; for acetaminophen at 10% w/w PAA was most effective on a weight basis, PVP and PVPVA substantial, HPMCAS, E100 and HPMC clearly less trasi2012a C15.

Figure

Log growth rate against temperature; from the top, felodipine alone, PVAc, HPMCAS, PVP/VA, PVP. The vertical gap to the top curve is the inhibition and is widest at 70 °C; the curves meet near 110 °C except PVP kestur2010a R5.

Growth rate without divided by growth rate with polymer; every polymer except PVAc falls linearly toward 110 °C kestur2010a R6.

Why

The authors read the growth results as showing that bulk properties do not explain the order (see F11) and that the strength or extent of drug–polymer hydrogen bonding does (see F12); they link the temperature trend to hydrogen bonds that "would be expected" to weaken as thermal energy rises, so that the polymers lose their effect near 110 °C, and call this a "speculation" supported by the data kestur2010a C43, C44. They note that Ishida and co-workers had seen a polymer effect that grew as the temperature fell kestur2010a R19, C12. Polymers at 3% are described as acting on something other than Tg or Tm kestur2010a C41.

Conditions and limits

One drug, dry films, nucleation at 40 °C and growth of form II only; nucleation with the polymers was not quantified, although the conclusions speak of crystallization inhibitors kestur2010a P3, X10. Several text statements differ from the figures: the fall of about 51-fold against "3 orders", PVP still inhibiting about 3-fold at 110 °C, and Fig. 3 slopes about twice the Fig. 4 rates, so the radius or diameter convention must be checked before absolute rates are compared with other papers kestur2010a X1, X2, X3. Error bars and the significance test for PVAc are not defined kestur2010a X9, R7. The solvent of the growth samples is printed inconsistently (dichloroethane or dichloromethane) kestur2010a X8.

Counterexamples and conflicts

(a) Polymers can accelerate growth: TPGS raised celecoxib growth by about two orders of magnitude at 70 °C and PHB raised that of higher-Tg compounds, with ratios up to about 10 mosqueragiraldo2014 C9; sato2016 R11. (b) The best polymer is not the same for every drug: PAA did not differ significantly from neat flutamide, while it was best for acetaminophen and bifonazole (see F13) trasi2012b R3, C7; trasi2012a C15; kestur2011 R7. (c) The best growth inhibitor can promote nucleation: PAA gave nearly three times more nuclei of acetaminophen, and HPMCAS, the weakest growth inhibitor, was the best nucleation inhibitor trasi2012a C23, C46.

Transfer check

Measure the neat growth rate and the rate with 1.5 and 3–5% polymer at two temperatures, one of them close to the intended storage temperature (here 70 °C, which is 27 °C above the felodipine Tg of 43.3 °C; synthesis: 70 − 43.3), because at 110 °C most of the differences between polymers disappear kestur2010a R6, M1. Confirm the polymorph by Raman or melting point, test nucleation separately, and expect no effect from a polymer such as PVAc that lacks a strong acceptor kestur2010a R3, R7.

History

Ishida and co-workers (outside the library) had reported the temperature dependence kestur2010a R19. Kestur and Taylor (2010) made the first systematic comparison of polymer chemistry at low levels in the library, and a companion paper varied PVP molecular weight, concentration and temperature kestur2010a C28; kestur2010b R13. In 2011 the method was extended to bifonazole and nimesulide kestur2011 P3. Trasi and co-workers (2014) compared the factor across drugs of very different growth rates, in the Crystallization tendency chapter trasi2014a C10.

Related: Crystallization tendency F19, Crystallization tendency F2.

How it was shown

Kestur and Taylor measured the onset Tg of the pure polymers and of the 3% dispersions by DSC at 10 °C/min after erasing thermal history (TA Q2000, Tzero pans with a pinhole), and the offset melting point of crystalline felodipine and of its 3% cryomilled physical mixtures at 1 °C/min kestur2010a P8, P9. The 2011 study did the same for bifonazole and nimesulide with 5 wt% polymer, in triplicate for growth, with a single Tg for every dispersion kestur2011 P5, R5, R6. Following Wu and co-workers, who had collapsed growth rates of polyoxyethylene with additives by normalizing to Tm − T and T − Tg, an unchanged Tm was read as an unchanged driving force and an unchanged Tg as unchanged mobility kestur2010a R16, C31, E2.

What the data show

Felodipine: onset Tg 43.3 °C alone, 44.1 (PVP), 43.7 (PVP/VA), 42.7 (HPMCAS) and 43.5 °C (PVAc); offset Tm 141.5 °C alone, 140.7, 142.9, 141.4 and 142.6 °C; pure-polymer Tg 102.9 (PVP K-12), 102.6 (PVP/VA), 115.5 (HPMCAS) and 38.7 °C (PVAc); Mw 2000–3000, 45 000–47 000, 18 000 and 100 000 kestur2010a R1, M1, M2, M3, M4, M5. The order of Mw is PVP < HPMCAS < PVP/VA < PVAc, but the order of inhibition is PVP > PVP/VA > HPMCAS > PVAc kestur2010a C35. Bifonazole (Tg 15.2 °C) with 5 wt% HPMCAS, PVP/VA, PVP K-29/32 and PAA: dispersion Tg 16.9, 16.4, 16.5 and 17.4 °C; nimesulide (Tg 19.5 °C) with PVP K-12, PVP/VA and PVAc: 20.8, 20.6 and 19.9 °C kestur2011 R6, M1, M2. Pure-polymer Tg and Mw: PAA 103.0 °C, Mw 1800; PVP K-29/32 163.9 °C, 45 000–55 000; PVP K-12 102.9 °C, 2000–3000; PVAc 38.7 °C, 100 000 kestur2011 M3, M4, M7, M8. At higher polymer content the Tg does change: PVP raised the Tg of felodipine by 10 °C at 25 wt% while HPMC and HPMCAS did not change it over 0–25 wt%, yet the polymers were equally effective on nucleation konno2006 R6, C25; in the PVP study the Tg was "largely unaffected" by PVP and concentration kestur2010b C34. In the introduction of the indomethacin work, inhibition at low PVP levels beyond antiplasticization was already cited taylor1997 C5. At the much higher polymer contents of later screens the result is the same: HPMC (Tg about 140 °C) gave lower amorphicity indices than E100 (around 55 °C) for all five carboxylic acids of the 15-drug screen, and the onset Tg of the polyphenol dispersions (69–128 °C) did not follow their stability (see F24, F23) vaneerdenbrugh2011a R3, C47; wegiel2013a R9, X14.

Figure

Dispersion Tg 42.7–44.1 °C against 43.3 °C for felodipine; Tm 140.7–142.9 °C against 141.5 °C; pure-polymer Tg from 38.7 to 115.5 °C kestur2010a R1.

Why

The authors' reasoning: an unchanged Tg means unchanged mobility and an unchanged Tm an unchanged chemical potential of the crystallizing component, so "increases in Tg cannot account" for the slower growth and there "does not appear" to be a macroscopic effect on ΔG kestur2010a C31, C32, C36. Pure-polymer Tg is "clearly not the case" and for molecular weight "no correlation was found" kestur2010a C34, C35. They place the result among a "mounting body of evidence" that Tg does not predict crystallization tendency in multicomponent amorphous systems kestur2010a C33, R20. In 2011 they reach the same conclusion for bifonazole and nimesulide kestur2011 C12, C13. The exact mechanism of retardation is then "not readily apparent" from bulk properties kestur2010a C41.

Conditions and limits

Three drugs, polymer levels of 3–5 wt%, growth above Tg; at the high polymer contents of ordinary dispersions (20 wt% and more) the blend Tg does change, and other studies show antiplasticization by PVP konno2006 C25. The melting argument uses form I in physical mixtures while form II grows, and no form II melting point with polymer is reported kestur2010a X7. Tm of the dispersions was measured in physical mixtures, not in dispersions kestur2010a P9. The later screens (10–90% and 30–80% polymer) lie outside the low-polymer regime, so they show that Tg does not rank polymers at high content either, not that antiplasticization is absent there vaneerdenbrugh2011a C47; wegiel2013a C5.

Counterexamples and conflicts

(a) For nimesulide the inhibition order PVP ≈ PVP/VA > PVAc does follow polymer Tg (102.9, 102.6, 38.7 °C), and for both bifonazole and nimesulide the rank of dispersion Tg matches the inhibition rank, although all differences from the drug are 0.4–2.2 °C against a stated ± 1 °C; the "no correlation" rests on the bifonazole PAA–PVP pair and on the small Tg change, not on rank kestur2011 X8, R6. (b) In the library's other polymer work, a polymer that raises Tg can also inhibit through mobility: PHB was thought "most likely" to change the growth rate through the mobility of the solute, with Tg varying with polymer content sato2016 C19.

Transfer check

For a new drug, measure the Tg of the drug with the intended polymer at the intended level and the melting point of a physical mixture; if neither moves (±1 °C), do not select the polymer by Tg, molecular weight or melting point, and test growth directly (see F10). If the polymer is used at 20 wt% or more, the Tg change becomes a factor again konno2006 C25.

History

Taylor and Zografi (1997) cited the finding that inhibition by PVP at low levels is more than antiplasticization explains taylor1997 C5. Konno and Taylor (2006) showed polymers equally effective on nucleation despite different Tg effects konno2006 C25. Kestur and Taylor (2010) tested Tg, Tm, polymer Tg and Mw as explanations of growth inhibition and found none sufficient kestur2010a C32, C34, C35; the 2011 paper confirmed it for two more drugs kestur2011 C12, C13. Van Eerdenbrugh and Taylor (2011) and Wegiel and co-workers (2013) saw the same absence of a Tg ranking in a dry screen and in humid storage at high polymer content vaneerdenbrugh2011a C47, X11; wegiel2013a C5, X14.

Related: PEG dispersions F6, Crystallization tendency F19.

How it was shown

Felodipine with 0, 20, 30, 40 or 50% w/w polymer was spin-coated from ethanol–dichloromethane on a substrate rotating at 500–2000 rpm, dried at 90 °C in a nitrogen glove box and measured by transmission IR (BioRad FTS 6000, 128 scans); the authors used 20–50% polymer because IR needed that much to show differences kestur2010a P6, P7, C6. The NH band at 3341 cm⁻¹ of amorphous felodipine is the NH bonded to a carbonyl of a neighbouring molecule, a free NH is expected near 3420 cm⁻¹, and the carbonyl bands at 1701 and 1682 cm⁻¹ are non-bonded and bonded carbonyl; a lower NH wavenumber means a stronger hydrogen bond kestur2010a R15, C18.

What the data show

NH at 50% polymer: 3341 cm⁻¹ for amorphous felodipine, 3354 (PVAc), 3348 (HPMCAS), 3298 (PVP/VA) and 3294 (PVP) kestur2010a R9. With PVAc the NH rose to 3345 cm⁻¹ at 20% and 3354 at 50%; with HPMCAS to 3343 and 3348 cm⁻¹; in both the free carbonyl peak grew relative to the bonded one kestur2010a R11, R12. With PVP/VA the NH stayed at 3341 at 20% but broadened toward lower wavenumber, and a shoulder at 3290 cm⁻¹ grew with polymer content; with PVP the NH shifted down and broadened more with concentration; a band near 3225–3230 cm⁻¹ also grew in both pyrrolidone polymers and is not discussed kestur2010a R13, R14. The carbonyl region could not be read for PVP and PVP/VA because the polymer carbonyls overlap it kestur2010a C22. The same drug and PVP gave NH and carbonyl shifts of about 45 and 20 cm⁻¹, cited in a later paper kestur2011 R20. In other drugs, growth inhibition and drug–polymer hydrogen bonding also correlated: for acetaminophen hydrogen bonding "seems to correlate well" with the decrease in growth, for flutamide hydrogen bonding has "a very clear effect" on growth but not on nucleation, and for nilutamide PVP moved the drug NH band from about 3313 to about 3130 cm⁻¹ and inhibited far more than PAA, which left the bands nearly unchanged trasi2012a C34; trasi2012b C45; trasi2014b R12, R11, R3. The temperature trend supports the picture without proving it: the inhibition ratio falls on warming for every polymer except PVAc kestur2010a R6.

Figure

(a) The NH band shifts down and broadens with PVP content, and a second band near 3225 cm⁻¹ grows; (b) the carbonyl region is dominated by the PVP band kestur2010a R14, C22.

Lower wavenumber, stronger hydrogen bond: PVP and PVP/VA below the 3341 cm⁻¹ of felodipine, HPMCAS and PVAc above it kestur2010a R9.

Why

The authors' interpretation: PVP and PVP/VA form hydrogen bonds with felodipine stronger than those between felodipine molecules, while HPMCAS and PVAc form weaker ones, formed at the expense of drug–drug bonds ("providing evidence") kestur2010a C19, C20, C21, C23. PVP/VA, in which 40% of the polymer is acetate side chains that "appear to interact only weakly", inhibits less than the same weight of PVP, and PVAc has essentially no effect kestur2010a C24, C38. For growth, a drug molecule must first demix from the polymer at the growth front; if mixing is favourable, demixing is an activated step that breaks drug–polymer interactions, and stronger interactions make it less favourable ("This speculation") kestur2010a C42. At high temperature thermal energy weakens the bonds, "especially weak ones", so the polymers lose their effect; the greater effect at low temperature is "consistent with our supposition" kestur2010a C43, C44. In blends of polymers, at much higher additive levels, strongly hydrogen-bonding diluents slowed spherulite growth most kestur2010a C39, R17, R18. Not measured: drug–polymer demixing or polymer enrichment at the growth front; hydrogen bonding at the growth temperatures and polymer levels; nucleation kestur2010a X6, X5, X10.

Conditions and limits

Four polymers, one drug with an NH donor and ester carbonyls; the hydrogen-bond order is a rank correlation, and no quantitative relation between band shift and growth rate was tested kestur2010a C38, X5. The order of HPMCAS and PVAc rests on a 6 cm⁻¹ difference at 50% (3348 against 3354 cm⁻¹), and the IR resolution is not printed kestur2010a R9.b, R9.c, P7. Polymers were compared at equal weight, not at equal numbers of acceptor groups, and PVP/VA's acetate share is not defined as mass or mole kestur2010a X11.

Counterexamples and conflicts

(a) HPMCAS shifts the NH up like PVAc, yet it slows growth 4-fold at 70 °C while PVAc does not, so interactions weaker than the drug's own can still inhibit, or HPMCAS acts by another route kestur2010a R9.b, R9.c, R6. (b) A drug whose donor is bound intramolecularly (nimesulide) gives only small shifts and a flatter ranking (see F14) kestur2011 C27. (c) PAA, a strong donor polymer, shifted the nilutamide bands little and inhibited marginally, while for bifonazole it was the best inhibitor (see F13) trasi2014b R3, R11; kestur2011 R7.

Transfer check

For a drug with an NH or OH donor, record the donor band of the drug in films with several polymers at 20–50% w/w and rank the polymers by the shift, then check the ranking in a growth test at 3–5% polymer and a temperature near storage; do not extrapolate the IR order to a drug whose donor may be intramolecularly bonded without testing it (F14) kestur2010a C6, X5.

History

Taylor and Zografi (1997) proposed that a drug–PVP hydrogen bond could explain crystallization inhibition taylor1997 C45. Kestur and Taylor (2010) tested it across four polymers in a growth assay and found the rank correlation; Trasi and co-workers (2012, 2014) found similar correlations for acetaminophen, flutamide and nilutamide kestur2010a C38; trasi2012a C34; trasi2012b C45; trasi2014b R3. Kestur and co-workers (2011) then asked whether the link holds for drugs of different hydrogen-bond chemistry kestur2011 C4. Wegiel and co-workers (2013) carried the approach to two polyphenols and six polymers; their IR ranking agreed with stability at the extremes only (see F25) wegiel2013a C25, X4.

Related: Crystallization tendency F19, PEG dispersions F10.

How it was shown

Kestur, Van Eerdenbrugh and Taylor cryomilled crystalline drug with 5% w/w polymer, melted 3–5 mg between cover glasses, quenched to 50 degrees below Tg, stored the films over P2O5 at ambient temperature so that nuclei formed, and measured the spherulite diameter against time on a Linkam hot stage (40–100 °C) or in ovens and coolers when growth was slow; bifonazole was measured at 18 °C and 30–100 °C, nimesulide at 18 °C and 30–80 °C, in triplicate; Raman microscopy showed the same polymorph with and without polymer kestur2011 P1, P2, P3, P4, R3. The bifonazole polymers were PVP K-29/32, HPMCAS, PVP/VA (data not shown) and PAA; the nimesulide polymers PVP K-12, PVP/VA and PVAc kestur2011 A2, A3, M3, M4, M5, M6, M7, M8, X11.

What the data show

Bifonazole (read from Fig. 3): no polymer ≈1.7 × 10⁻¹² m/s at 18 °C, ≈1.0 × 10⁻⁸ at 50 °C and ≈4.7 × 10⁻⁶ at 100 °C; HPMCAS ≈4.3 × 10⁻¹³, 5.0 × 10⁻⁹ and 2.7 × 10⁻⁶; PVP ≈2.9 × 10⁻¹³, 3.4 × 10⁻⁹ and 2.2 × 10⁻⁶; PAA ≈1.1 × 10⁻¹³, 1.4 × 10⁻⁹ and 9.2 × 10⁻⁷; the ratios without over with polymer are ≈4.0 (HPMCAS), ≈5.9 (PVP) and ≈15 (PAA) at 18 °C and ≈1.7, ≈2.1 and ≈5.1 at 100 °C kestur2011 R7. Nimesulide (Fig. 4): no polymer ≈3.7 × 10⁻¹¹ m/s at 18 °C, ≈1.6 × 10⁻⁷ at 50 °C and ≈8.5 × 10⁻⁶ at 80 °C; PVAc ≈1.1 × 10⁻¹¹, 4.5 × 10⁻⁸ and 4.2 × 10⁻⁶; PVP/VA ≈2.3 × 10⁻¹², 3.4 × 10⁻⁸ and 2.7 × 10⁻⁶; PVP ≈1.8 × 10⁻¹², 2.1 × 10⁻⁸ and 1.8 × 10⁻⁶; the ratios at 18 °C are ≈21 (PVP), ≈16 (PVP/VA) and ≈3.4 (PVAc), at 80 °C ≈4.7, ≈3.1 and ≈2.0 kestur2011 R9. The text gives only 2–10-fold for nimesulide, which understates the plot at 18 °C, a temperature just below the nimesulide onset Tg of 19.5 °C kestur2011 C10, X7, X13. By comparison the felodipine ratios at 70 °C were 10.0 (PVP), 6.5 (PVP/VA), 4.4 (HPMCAS) and 1.2 (PVAc) kestur2010a R6. Polymer effects in other drugs also differ in order: PAA was not significantly different from neat flutamide, and marginally inhibiting for nilutamide, but the best growth inhibitor on a weight basis for acetaminophen trasi2012b R3, C7; trasi2014b R3; trasi2012a C15.

Figure

Log scale, each decade a 10-fold change; black squares no polymer, red triangles HPMCAS, blue diamonds PVP, open circles PAA, lowest at every temperature kestur2011 R7.

Black squares no polymer, open red triangles PVAc, blue triangles PVP/VA, open circles PVP; the polymer curves separate below 50 °C and nearly merge at 60–80 °C kestur2011 R9.

Why

The authors' interpretation: bifonazole has an imidazole nitrogen acceptor, and in its crystal the molecules are linked only by weak C–H···N contacts; a polymer with a donor stronger than that C–H, such as the COOH of PAA, "should be capable" of forming more favourable bonds with the imidazole nitrogen, disrupting drug–drug contacts; the abstract says only that this "may be related" kestur2011 R18, C15. The imidazole nitrogen "would be expected" to be a better acceptor than the PAA carbonyl, so it should break PAA self-association kestur2011 C16, R22. PVP, with acceptors only, "can at most" form very weak bonds with the acidic C–H of bifonazole; HPMCAS has donors but few COOH groups (1 against 13.9 mmol/g for PAA), competes with many acceptors and may be limited sterically kestur2011 C17, C18. For nimesulide see F14. The authors conclude that polymer effectiveness "appears to depend" on drug chemistry and on the availability of hydrogen-bonding groups kestur2011 C30.

Conditions and limits

Two drugs at one polymer level (5 wt%), growth from the dry melt only; nucleation, higher loadings, moisture and dissolution were not tested kestur2011 P1, C5. Different PVP grades were used for the two drugs (K-29/32 and K-12), the PVP/VA data for bifonazole are not shown, and no spectroscopic evidence of bifonazole–polymer interaction was collected kestur2011 X11, X14, X9.

Counterexamples and conflicts

(a) The PAA advantage is not general: it did nothing for flutamide, little for nilutamide and failed with acidic drugs in the screen of F15 trasi2012b R3; trasi2014b R3; vaneerdenbrugh2010b R3; it was also by far the least effective polymer for the weak-acid polyphenols naringenin and quercetin, which needed 70–80% PAA to give an amorphous dispersion wegiel2013a R8, C48. (b) For nimesulide the order follows the polymer Tg, which the card flags against the claim of no Tg correlation (see F11) kestur2011 X8. (c) The text says 2–10-fold for nimesulide, the plot shows up to about 21-fold kestur2011 X7.

Transfer check

Classify the drug by its free hydrogen-bond groups from its structure and crystal packing: an acceptor-only drug like bifonazole with no strong self-association is a candidate for a donor polymer, a donor drug for a strong-acceptor polymer such as PVP. Test two chemically different polymers at 5 wt% by melt growth at two temperatures including one near Tg, and decide only after checking nucleation and compatibility (PAA is acidic and may be unsuitable with basic or acid-labile drugs; PAA complexed with bases in solution in F15) kestur2011 C29, C15; vaneerdenbrugh2010b C25.

History

Kestur and Taylor (2010) had shown that hydrogen-bond strength ranked polymers for felodipine; Van Eerdenbrugh and Taylor (2010) had shown that acidic polymers suited amide and basic drugs kestur2011 C3, C2. The 2011 paper tested the hypothesis that the best polymer depends on drug chemistry on two drugs of opposite hydrogen-bond character kestur2011 C4. Trasi and co-workers (2012, 2014) later found PAA ineffective for flutamide and nilutamide in the Crystallization tendency chapter trasi2012b C7; trasi2014b R3. Wegiel and co-workers (2013) found PAA the least effective polymer for two weak-acid polyphenols, in line with the rule that acidic polymers inhibit acidic compounds poorly (see F25) wegiel2013a C48.

Related: Crystallization tendency F19.