Solid state stability

How it was shown

Baird and co-workers (2010) quenched melts of five molecules between cover glasses and stored them at 0% RH, 14–20 °C below Tg, following crystallization by polarized light microscopy baird2010 P7, A6, C45, X16 (Crystallization tendency F17). Van Eerdenbrugh and Taylor (2010) spin coated eight drugs of known class with seven polymers and averaged an amorphous index over 7 days of dry storage vaneerdenbrugh2010b P1, P2, R2, E4 (Crystallization tendency F17). Marsac and co-workers compared nifedipine and felodipine in PVP K29/32 by nucleation counts in spin-coated films, dry and at 0–75% RH marsac2006a P4, P5, A3 (Physical stability F2); marsac2008a P1, P2, P3 (Physical stability F2). Zhu and co-workers followed 50 melts each of neat ibuprofen and fenofibrate at 25 °C for about 500 h, measured 20/80 PEG 3350 comelts by XRD and Raman, and placed a piece of crystalline PEG into the supercooled neat drug zhu2012 P2, P3, P5, P6, P7 (PEG dispersions F8, PEG dispersions F9). Trasi and co-workers stored the contents of six marketed tacrolimus capsules open at 40 °C/75% RH trasi2017 P7, A1 (Physical stability F33), and stored 60:40 and 40:60 lumefantrine dispersions with five polymers under the same conditions trasi2020 P1, P5, P6 (Physical stability F34).

What the data show

In the storage test indoprofen (class I) showed one spherulite after 5 min and a fully birefringent field after 1 day, while clotrimazole and felodipine (class III) showed only edge crystallization after 3 months baird2010 R23, R25 (Crystallization tendency F17). Class I drugs averaged an amorphous index of 47–59% and class II drugs 73–97% vaneerdenbrugh2010b R4, C15 (Crystallization tendency F17). Nifedipine and felodipine have nearly the same Tg, yet nifedipine nucleated faster in every one of the 30 combinations of PVP content and RH, and at equal Tg − T about 5–20 times faster marsac2006a R4.c, R16, C19 (Physical stability F2); marsac2008a R8, C23 (Physical stability F2). In PEG the picture changed: neat benzocaine and haloperidol crystallized within seconds, neat ibuprofen and fenofibrate had started in 10 and 17 of 50 samples after 3 weeks, yet in the 20/80 comelts ibuprofen began after about 100 min and fenofibrate after 120–180 min, and a crystal grew at the PEG interface within minutes and nowhere in the bulk liquid zhu2012 R1, R2, R3, R23.a (PEG dispersions F8); zhu2012 R15, R19 (PEG dispersions F9). Of six marketed tacrolimus capsules of about 1:1 drug:HPMC only one generic crystallized, from 7.5% at 1 day to 88% at 3 weeks, while the pure amorphous drug crystallized only after 2 weeks trasi2017 R6, R7, R12 (Physical stability F33); trasi2017 R16, C28 (Crystallization tendency F17).

Figure

The class I drugs sit lowest and the class II drugs highest, the same order as their neat-drug classes vaneerdenbrugh2010b R4 (Crystallization tendency F17).

Why

Baird and co-workers attribute the faster crystallization of class II and quench-glassy class I glasses below Tg to nuclei formed during cooling ("It is believed"); the nuclei were not detected baird2010 C57, C58 (Crystallization tendency F17). Zhu and co-workers explain where the link breaks: both the drug's own tendency and its ability to interact with the carrier matter, and a crystalline carrier can add heterogeneous nucleation; the extractor's reservation is that PEG crystallized first, so the drug was also concentrated in the remaining non-crystalline fraction zhu2012 C38, C39, C4, X13 (Crystallization tendency F17, PEG dispersions F9). For the tacrolimus generic the authors write that the result "strongly suggests" a processing cause, because the product crystallized faster than the drug alone trasi2017 C28 (Physical stability F33). Lumefantrine forms a salt with the carboxylic acid groups of the enteric polymers, and the authors find that stability "correlates" with protonation, while PVPVA "most likely" failed through hygroscopicity and the drug's high log P trasi2020 C36, C37, C31 (Physical stability F34). (synthesis: the class measures nucleation and growth of the neat drug under one cooling and heating protocol; a dispersion changes the nucleation (a polymer lowers it, a crystalline carrier can raise it), the growth and the route, so the class can order drugs only while those changes are of similar size for all of them.)

Conditions and limits

The classes come from neat-drug DSC runs and the storage ranking from one or two compounds per class, one temperature each, judged by eye at 0% RH only baird2010 X16, P7, A6 (Crystallization tendency F17). The spin-coating comparison rests on eight drugs and films stored dry for 7 days vaneerdenbrugh2010b P1, P2 (Crystallization tendency F17). The PVP result is one drug pair marsac2006a X7 (Physical stability F2). The PEG results are two class III drugs, one PEG grade and one composition zhu2012 P7, X6 (PEG dispersions F9), and the tacrolimus work is one lot per brand stored open outside the package trasi2017 A1, P7 (Physical stability F33).

Counterexamples and conflicts

The class did not predict the outcome for pretomanid (class I), which stayed amorphous with HPMCAS for several months nguyen2023a R4, R27, C62 (Crystallization tendency F17), nor for delamanid, PEG or tacrolimus vaneerdenbrugh2010a C48; duong2022a C44; zhu2012 C41; trasi2017 C28 (Crystallization tendency F17). Lumefantrine is a class III drug trasi2020 R5, R6, C7 (Crystallization tendency F9), yet its PVPVA dispersions crystallized after 1, 2, 4 and 8 weeks at 50, 25, 10 and 5% drug loading hiew2022a R6 (Miscibility and drug-polymer interactions F23). The class also carried only partly into supersaturated aqueous solution: felodipine, slow from the melt, exceeded its amorphous solubility and underwent liquid–liquid phase separation in water, but several slow crystallizers from the melt, such as efavirenz, papaverine, warfarin and glyburide, crystallized readily from solution raina2014a R20, C31; raina2015b C13, C28; ilevbare2013c R8, C24; hsieh2014 R18, C30 (Crystallization tendency F18). Class III is therefore no guarantee in a polymer that lacks a specific contact with the drug, which is the authors' reading of the PVPVA failure hiew2022a C61, C62 (Miscibility and drug-polymer interactions F23), and class I is no verdict where a polymer with a strong contact is present.

Transfer check

Run the heat–cool–heat cycle on the neat crystalline drug at the published rates and report the class together with the rates baird2010 P1; trasi2012a C7 (Crystallization tendency F9). Treat class I as a warning that strong inhibition will be needed, then test the actual dispersion at its drug loading, polymer, humidity and temperature, because the class alone did not predict pretomanid, delamanid, PEG or tacrolimus outcomes vaneerdenbrugh2010a C48; nguyen2023a C62; duong2022a C44; zhu2012 C41; trasi2017 C28 (Crystallization tendency F17). If the carrier is crystalline, place a piece of it in the supercooled neat drug and look for interface crystals before relying on the neat-drug class zhu2012 R15, R19 (PEG dispersions F9); for a finished product stress-test each batch, not only the drug–polymer dispersion trasi2017 A1, C28 (Physical stability F33).

History

2006: Marsac, Konno and Taylor found, in dry PVP films, that nifedipine nucleates faster than felodipine of the same Tg marsac2006a R4.c, R16 (Physical stability F2). 2010: Baird and co-workers published the three classes and the storage test that followed them, and Van Eerdenbrugh and Taylor found the same order in the class I drugs of the polymer screen baird2010 R23, R25; vaneerdenbrugh2010b R4 (Crystallization tendency F17). 2012: Zhu and co-workers showed that crystalline PEG breaks the link for two class III drugs zhu2012 C4, C41, R23.a (Crystallization tendency F17, PEG dispersions F8). 2017: Trasi and co-workers found one marketed tacrolimus product that crystallized faster than the pure amorphous drug and traced it to the manufacturing route trasi2017 R16, C28 (Physical stability F33). 2020–2022: Trasi and co-workers classed lumefantrine as class III and showed that polymer choice moved its stability from one week to 3 months, and Hiew and co-workers repeated the PVPVA result across drug loadings trasi2020 R5, R3, C6 (Crystallization tendency F9, Physical stability F34); hiew2022a R6 (Miscibility and drug-polymer interactions F23). 2023: Nguyen and co-workers kept class I pretomanid amorphous with HPMCAS nguyen2023a C62 (Crystallization tendency F17).

How it was shown

For neat drugs Baird and co-workers (2010) calculated T_{rg} = T_g/T_m, glass-stability parameters and fragility for the 51 molecules of the DSC screen, and in 2012 measured the melt viscosity of 34 of them and fitted the VTF equation to 28 curves baird2010 E4, P8, P11, P3, P10; baird2012 P1, P5, P6 (Crystallization tendency F12). Trasi and co-workers (2014) compared growth rates of nine drugs with the melting entropy trasi2014a R11 (Crystallization tendency F6). For dispersions the Konno, Marsac, Kestur and Rumondor studies compared the onset Tg, the Tg-scaled temperature, enthalpy relaxation and an Adam–Gibbs–Vogel mobility with nucleation and growth in films konno2006 P3, P6, R6, R9 (Physical stability F5); konno2008 P9, R13 (Physical stability F5); rumondor2010 P8, E1, R7 (Physical stability F5), and measured the Tg and melting point of 3–5 wt% dispersions by DSC kestur2010a P8, P9 (Miscibility and drug-polymer interactions F11). Marsac and co-workers estimated the driving force and the activity of the drug in PVP by the Hoffman and Flory–Huggins equations marsac2006b P8, E4, R12 (Miscibility and drug-polymer interactions F9); marsac2006a E10, R18 (Physical stability F4). Rumondor and co-workers tabulated log P, Tm, heat of fusion and the Hoffman driving force of six drugs against humid separation rumondor2009c M1, M6, P10 (Physical stability F13).

What the data show

Tg/Tm did not separate the classes, the glass-stability parameters picked the wrong class II molecules, fragility m ran from 18 to 731 without a class pattern and the VTF strength parameter D was 4.3–8.7 in all classes baird2010 R14, C37, C27, C28, R18, C31; baird2012 R12, C39 (Crystallization tendency F12). Viscosity was different: the six class I-A melts lay below 0.01 Pa·s at Tm and all sixteen class III melts above it, but within classes it did not rank compounds, tolazamide reaching about 105 °C of undercooling before crystallizing and celecoxib about 45 °C baird2012 R1, C22, R4, R2, C11, C16 (Crystallization tendency F5). Molecular descriptors fared no better: in two principal component analyses rapid crystallizers were on average smaller, more rigid molecules with larger heats of fusion, but the separation was partial, outliers occurred in every class and no single descriptor sufficed baird2010 R21, C63; vaneerdenbrugh2010a R19, C45 (Crystallization tendency F13). The driving force and the melting entropy explained part of the spread: growth normalized for viscosity and driving force fell with melting entropy with R^2 = 0.61 for nine drugs, and vemurafenib α had the larger driving force at every temperature yet β dominated at 120–190 °C trasi2014a R11 (Crystallization tendency F6); lu2016 R22, C52 (Crystallization tendency F6).

In dispersions the same holds. At 3 wt% polymer the Tg of felodipine was 42.7–44.1 °C against 43.3 °C alone and its melting point 140.7–142.9 against 141.5 °C, whereas growth fell up to about 10-fold, and PAA (Tg 103.0 °C) outperformed PVP K-29/32 (Tg 163.9 °C) for bifonazole kestur2010a R1 (Miscibility and drug-polymer interactions F11); kestur2010a R6 (Miscibility and drug-polymer interactions F10); kestur2011 M3, M7, C12 (Miscibility and drug-polymer interactions F11). Activities of nifedipine and felodipine in PVP were estimated at about 0.96 at a polymer volume fraction of 0.1 and 0.84 at 0.2, and 3 wt% PVP K-12 moved the felodipine melting point from 141.5 to 140.7 °C, within its standard deviation marsac2006a R19, C47 (Physical stability F4); kestur2010b R1.e, X11 (Physical stability F4). Nifedipine and felodipine, with the same Tg, nucleated 2–20 times apart, dry and at equal Tg − T with water, and humid felodipine films at the same T/Tg spanned about 30-fold marsac2006a C13; marsac2008a C23; konno2008 R13, C26 (Physical stability F5). In PEG the Tg of the drug did not matter either: nilutamide (Tg about 33 °C) behaved like the low-Tg drugs, and ketoconazole (Tg 44 °C) was excluded by its positive interaction parameter zhu2011 C28, C19, C30, C31 (PEG dispersions F6). For humid separation of PVP dispersions log P (1.92–5.34), the heat of fusion and the Hoffman driving force (4.1–11.0 kJ/mol) showed no trend rumondor2009c C27, C29, C32 (Physical stability F13).

Figure

At the same abscissa the rates scatter over more than a decade, with no single trend konno2008 R13 (Physical stability F5).

Points of every class sit in one band, so fragility on this scale carries no class information baird2012 R13, C42 (Crystallization tendency F12).

Why

The reduced Tg rests on the assumption that all liquids have the same viscosity at Tg, which the 2012 data contradict: compounds with similar Tm − Tg can differ in how fast viscosity rises on cooling baird2010 C1; baird2012 C36 (Crystallization tendency F12). In the dispersions an unchanged Tg means unchanged mobility, so "increases in Tg cannot account" for the slower growth kestur2010a C31, C34, C35 (Miscibility and drug-polymer interactions F11); a higher Tg "would certainly contribute" but "is not the only factor" konno2006 C27 (Physical stability F5); the estimates "do not forecast a large reduction" of the driving force, so the inhibition is read as kinetic marsac2006b C41, C42 (Miscibility and drug-polymer interactions F9). Where an additive did lower Tg strongly, the authors read the speed-up as mobility and found that 5–10% PHB lay close to the pure compounds when growth was plotted against T − Tg sato2016 C18, C19, R10, C21 (Physical stability F28); mosqueragiraldo2014 C29, C30 (Physical stability F22). (synthesis: every descriptor that failed is a bulk or whole-sample average, an equilibrium quantity or a single temperature; crystallization from a glass depends on local steps such as molecular orientation at the interface, specific drug–polymer contacts (F6) and nucleation at interfaces (F11), which none of them sees. This is a reading across the chapters, not a tested explanation.)

Conditions and limits

Tg was always measured on samples prepared differently from the films in which crystallization was counted konno2006 X12; konno2008 X6 (Physical stability F5). The polymer levels were 3–5 wt% in the growth studies, and at 20 wt% and more the blend Tg does change konno2006 C25 (Miscibility and drug-polymer interactions F11). No statistical test is named for the class comparisons, and 16 class I Tg values are estimates from molecular weight baird2010 R14, X17 (Crystallization tendency F12). The fragility category depends on whether m or D is used baird2010 X14 (Crystallization tendency F12).

Counterexamples and conflicts

For nimesulide the inhibition order does follow polymer Tg kestur2011 X8 (Miscibility and drug-polymer interactions F11). Antipyrin, 4-biphenylcarboxaldehyde and tolazamide lie below 0.01 Pa·s yet form glasses baird2012 X8 (Crystallization tendency F5). Where a miscible additive lowers the Tg of the mixture by about 10 °C, as TPGS and sucrose palmitate did for celecoxib, growth does speed up, but even after correction for Tg the rates still differed up to about 12-fold mosqueragiraldo2014 R5, R12, X4 (Physical stability F5). The Tg-based and mobility-based explanations therefore work for large differences and fail for small ones (synthesis: from the PHB and surfactant data against the 3–25% polymer data).

Transfer check

Do not screen candidates by Tg/Tm, glass-stability parameters or fragility alone; the heat–cool–heat test, a seeded growth rate and the melt viscosity at Tm carried more information in these data sets baird2010 C38; baird2012 C45; baird2010 C64 (Crystallization tendency F12). For polymers, if neither the Tg nor the melting point of a physical mixture moves by more than about 1 °C, do not select the polymer by them and test growth directly konno2006 C25 (Miscibility and drug-polymer interactions F11); use Tg only to compare the same formulation across storage conditions and do not screen by dispersion Tg or T/Tg at polymer levels below about 25% konno2006 X12; konno2008 X6 (Physical stability F5). For an additive that may plasticize, measure the mixture Tg first: a fall of 5 °C or more with 5% additive warns of faster growth (synthesis: from the PHB and surfactant data) sato2016 R7, X7 (Physical stability F28); mosqueragiraldo2014 R5, R1 (Physical stability F22).

History

2006: Marsac and co-workers found that a small polymer fraction changes the driving force little, and Konno and Taylor found equal nucleation despite different dispersion Tg marsac2006b R12 (Miscibility and drug-polymer interactions F9); konno2006 R6, R4 (Physical stability F5). 2010: Baird and co-workers found Tg/Tm, glass-stability parameters and fragility uninformative for 51 molecules, and Kestur and Taylor found that dispersion Tg and melting point did not explain the polymer order baird2010 R14, C37 (Crystallization tendency F12); kestur2010a R1 (Miscibility and drug-polymer interactions F11). 2011: Zhu and co-workers found no effect of drug Tg in PEG zhu2011 C28, C31 (PEG dispersions F6). 2012: Baird and co-workers added melt viscosity, which separated only the extreme classes baird2012 R1, C22 (Crystallization tendency F5). 2014: Trasi and co-workers found melting entropy to explain part of the growth spread, and Mosquera-Giraldo and co-workers showed that a surfactant that lowers Tg speeds growth trasi2014a R11 (Crystallization tendency F6); mosqueragiraldo2014 R1, R5 (Physical stability F22). 2016: Sato and Taylor traced the PHB speed-up mainly to the drop in mixture Tg sato2016 R7, R10 (Physical stability F28).

How it was shown

Marsac and co-workers computed χ from Fedors solubility parameters, from the depression of the drug melting point in mixtures with PVP K12 (1 °C/min) and, in 2009, from the offset of the melting endotherm and from the solubility at 25 °C in 1-ethyl-2-pyrrolidone (EP), a liquid analogue of PVP marsac2006b M1, M2, P1, R9 (Miscibility and drug-polymer interactions F6, Miscibility and drug-polymer interactions F7); marsac2009 P2, P7, E1 (Miscibility and drug-polymer interactions F8); marsac2009 P6 (Miscibility and drug-polymer interactions F7). Baird and Taylor and Zhu and co-workers placed the same logic on drugs in PEG 3350: χ from the liquidus, activity coefficients from the solubility in liquid PEG 400, the drug carbonyl band in the melt, and the solubility extrapolated to 25 °C baird2011 P8, E8, P5, E7, P4 (PEG dispersions F3); zhu2011 P5, E1, P6, E3 (PEG dispersions F3, PEG dispersions F4). Rumondor and co-workers fitted ternary Flory–Huggins equations to water-sorption isotherms of felodipine–PVP and indomethacin–PVP blends rumondor2010a E2, E3, E4, P5, P6 (Physical stability F8).

What the data show

Fedors parameters gave \chi = 2.0 for nifedipine–PVP and 0.5 for felodipine–PVP, both predicting immiscibility or marginal miscibility, whereas melting point depression gave −3.8 and −4.2 marsac2006b R4, R3, R9 (Miscibility and drug-polymer interactions F6). For indomethacin–PVP K12 the same sample gave −6.16, −1.94 and −0.82, a 7.5-fold spread (synthesis: 6.16/0.82) marsac2009 R6 (Miscibility and drug-polymer interactions F7). For five PVP pairs the solubility in EP gave χ from −0.08 to −1.83 against −0.82 to 0.02 from offset melting, with only felodipine equal, and the two routes ordered the compounds differently marsac2009 R7.a, R9, X7 (Miscibility and drug-polymer interactions F8). In PEG the model with no interaction energy predicted eutectics of 12.8 and 1.5 wt% for chlorpropamide and aceclofenac against 35 and 30 measured baird2011 R1 (PEG dispersions F2); chlorpropamide's χ was −2.53 from one paper's phase diagram and −0.50 from another's, and ketoconazole had an activity coefficient of 0.09 in PEG 400 but a positive χ, 0 wt% solubility and no spectral shift baird2011 R9.a; zhu2011 R5.c, R5.f, R20; marsac2008b R14.f (PEG dispersions F3). The sign was consistent: negative χ, activity coefficients below 1, a shifted carbonyl band and an estimated solubility above 20 wt% singled out chlorpropamide, aceclofenac, benzocaine and ibuprofen baird2011 R9, R10; zhu2011 R5, R20 (PEG dispersions F3). In water sorption the drug–polymer term was only 0.029–0.160% of the interaction adjustment, so \chi_{23} fitted to blend isotherms ranged from −9.6 to 26.9 for felodipine–PVP and from −20.4 to 22.0 for indomethacin–PVP rumondor2010a R13, R14, C30 (Physical stability F8).

The estimated solubilities did not decide stability: the predicted solubility of the crystal in PVP was 0.06–0.07 g/g for nifedipine and 0.25–0.31 g/g for felodipine, and 0.010–0.134 weight fraction for five compounds in PVP K29/32, so most dispersions are supersaturated and their stability was inferred to be kinetic marsac2006b R17, R18; marsac2009 R9, C34 (Miscibility and drug-polymer interactions F9). In PEG the estimate was 28, 26, 25, 11, 6 and 0 wt% for benzocaine, chlorpropamide, ibuprofen, nilutamide, fenofibrate, and haloperidol and ketoconazole, yet benzocaine and ibuprofen at 20/80 began to crystallize after 120–180 and about 100 min, and PEG 400 capacities of 2–24% did not predict which drugs lost crystallinity in PEO zhu2011 R5 (PEG dispersions F4); zhu2012 R23.a (PEG dispersions F4); marsac2008b C22, R16 (PEG dispersions F4).

Figure

Left columns: measured solubility, γ and χ in EP; right columns: predicted solubility in PVP marsac2009 R9 (Miscibility and drug-polymer interactions F8).

Negative χ and 11–28 wt% for four drugs, positive χ and zero for haloperidol and ketoconazole zhu2011 R5 (PEG dispersions F4).

Why

The authors give a reason for each disagreement. The geometric-mean rule behind Fedors parameters holds for van der Waals forces but not for directional ones, so the failure for nifedipine–PVP "most likely" arises from specific interactions marsac2006b C24, C25 (Miscibility and drug-polymer interactions F6); the lack of depression with a high-Tg PVP grade is "most likely" because it is not liquid-like below its Tg marsac2006b C29 (Miscibility and drug-polymer interactions F7); the solubility route gives the thermodynamics at room temperature and one composition, and the melting route near the melting point and at low polymer fractions, where slow mixing may bias χ marsac2009 C27, C28 (Miscibility and drug-polymer interactions F8). An activity coefficient below 1 suggests favourable interaction only if the entropy of mixing is near ideal, and the PEG 400 value describes supercooled liquid and not largely crystalline PEG baird2011 C23, C24; zhu2011 C8 (PEG dispersions F3, PEG dispersions F4). The ternary water model assumes that the binary interactions persist in the blend, which the authors consider "unlikely", so any change is absorbed by χ₂₃, which "may not be physically meaningful" rumondor2010a C26, C27, C29, C31 (Physical stability F8). (synthesis: each route measures a different temperature, composition range and physical state of the polymer, so the size of χ or of a solubility is a property of the protocol; the sign is what survives.)

Conditions and limits

The printed Fedors values 2.0 and 0.5 could not be reproduced (2.36 and 0.70 at 298.15 K), and only nifedipine–PVP K29/32 was simulated marsac2006b X1, X9 (Miscibility and drug-polymer interactions F6). The solubility inputs of the 2009 χ values are not printed, so neither set can be recomputed marsac2009 X4 (Miscibility and drug-polymer interactions F8). The printed interaction-energy equation of the PEG model does not reproduce, although the signs and ranking hold baird2011 C19, X3 (PEG dispersions F2). The PEG solubilities are extrapolations from 50–80 wt% drug data below the PEG melting point, and nilutamide's 11 wt% lies above its own 10 wt% eutectic zhu2011 C8, X4 (PEG dispersions F4).

Counterexamples and conflicts

The χ of nifedipine–PVP was −3.8 from the 2006 melting onset (PVP K12, 1 °C/min) marsac2006b R9 (Miscibility and drug-polymer interactions F6), 0.00 from the 2009 offset and −0.81 from EP marsac2009 R7.b, R9 (Miscibility and drug-polymer interactions F8), and 0.0 in the humidity paper rumondor2009c R26 (Physical stability F13). A negative χ was sometimes enough to separate drugs and sometimes not: ketoconazole is athermal with PVP and also had a favourable activity coefficient in PEG 400 marsac2009 R3, R7.c (Miscibility and drug-polymer interactions F4); marsac2008b R14.f (PEG dispersions F3). The authors write that agreement in sign is usable but agreement in size is not marsac2009 X7, X16 (Miscibility and drug-polymer interactions F8).

Transfer check

Treat χ from either route as a ranking within one method, measure it with a polymer whose Tg is well below the drug melting point and a slow scan, report onset, midpoint and offset, and compare only drugs measured the same way marsac2006b R11; marsac2009 R6, X7, X16 (Miscibility and drug-polymer interactions F7, Miscibility and drug-polymer interactions F8). Use at least two readings for a drug in PEG, and test the real dispersion if they disagree baird2011 R9.a; zhu2011 R5.c, R5.f (PEG dispersions F3). Take drug–polymer χ from melting point depression or from solubility in a liquid analogue, not from the water sorption of the blend rumondor2010a R13, R14 (Physical stability F8). Compare an estimated solubility with the intended load, expect the load to exceed it for PVP, and follow the real dispersion over hours to days by XRD or Raman, because loads below the estimate crystallized in PEG marsac2009 R9; zhu2012 R23.a, X13 (Miscibility and drug-polymer interactions F9, PEG dispersions F4).

History

2006: Marsac and co-workers showed that a solubility parameter predicts nifedipine–PVP wrongly and that melting point depression gives −3.8 and −4.2 marsac2006b R4, R9 (Miscibility and drug-polymer interactions F6). 2008: Marsac and co-workers took solubility in liquid PEG 400 as a stand-in for the polymer marsac2008b C18, R14 (PEG dispersions F4). 2009: Marsac and co-workers added the EP route and the onset and offset dependence, and Rumondor and co-workers computed drug–water χ marsac2009 R6, R9 (Miscibility and drug-polymer interactions F7, Miscibility and drug-polymer interactions F8); rumondor2009c R24 (Physical stability F13). 2010: Rumondor and co-workers showed that water-sorption fits cannot recover the drug–polymer χ rumondor2010a R13, R14 (Physical stability F8). 2011–2012: Baird and Taylor and Zhu and co-workers compared four readings of drug–PEG interaction and tested the estimated solubility against the time to crystallize baird2011 R9, R10; zhu2011 R5 (PEG dispersions F3, PEG dispersions F4); zhu2012 R23.a (PEG dispersions F4).

How it was shown

Konno and Taylor (2006) spin coated felodipine with 0–25% w/w PVP K29/32, HPMCAS or HPMC, stored the films at 0% RH and 22 °C and took the nucleation rate as the slope of birefringent site density against time, in triplicate konno2006 P1, P2, E1, A2, R18 (Physical stability F1). Kestur and Taylor (2010) grew felodipine form II from melt films at seven temperatures from 70 to 110 °C, in triplicate kestur2010a P2, P3, P4 (Miscibility and drug-polymer interactions F10). Trasi and Taylor (2012) melt-quenched acetaminophen with 10% w/w of seven polymers, measured spherulite growth on a hot stage and counted the nuclei that grew after one cycle to −30 °C; for flutamide with 10 mol% additives they read nucleation from the growth rate at the temperature of the DSC crystallization maximum trasi2012a P1, P2, P6, P7; trasi2012b P1, P6, P7 (Physical stability F17). Kestur and co-workers (2013) compared the early crystallization slope of felodipine powders with film growth rates at 30 °C kestur2013b R15, R12, E2 (Physical stability F17). For the neat drugs the nucleation zone was found by cooling the melt to successively lower end temperatures and reheating trasi2012a P6 (Crystallization tendency F3); trasi2014b P1, P7 (Crystallization tendency F4).

What the data show

In dry films the first 3% of polymer removed roughly a 40–60-fold factor (synthesis: 1.3 × 10⁵/3.0 × 10³ = 43 for PVP), and the three polymers lay within a factor of about 1.5 of each other from 3 to 15% konno2006 R4.a, R4.b, R4.g, X9 (Physical stability F1). In the melt the same felodipine showed an order, PVP > PVP/VA > HPMCAS > PVAc, at every temperature kestur2010a R6, R8 (Miscibility and drug-polymer interactions F10). For acetaminophen the nuclei counts ran in a different order from growth: HPMCAS 2.5, PVP 5.2, PVPh 9.2, HPMC 9.3, PVPVA 12.3, E100 13.1 and PAA 26, against 10.8 for the drug trasi2012a R12 (Physical stability F17). In flutamide, PVP slowed growth about tenfold but did not change the nucleation read-out, whereas PVAc and PIPA, weak growth inhibitors, lowered it trasi2012b R3, R11, C18 (Physical stability F17). In felodipine powders 3% PVP slowed the early crystallization 60–77-fold but surface growth only about 2-fold and bulk growth about 7-fold kestur2013b R15, R12, C34 (Physical stability F17). The polymer can also move the nucleation zone: acetaminophen nucleated at 50–60 °C alone, at −10–30 °C with PAA and at 10–45 °C with HPMCAS trasi2012a R12 (Physical stability F17). For neat drugs the windows were far from the growth range: nilutamide formed about 12.6 surface nuclei in 30 min at −30 °C against 0.8–2.5 from −20 to 40 °C, and its crystals grew fastest near 100 °C trasi2014b R9, R8 (Crystallization tendency F4, Crystallization tendency F3). A crystalline carrier is a third source of nuclei: a crystal grew at the interface of a PEG piece placed in supercooled ibuprofen or fenofibrate and nowhere in the bulk zhu2012 R15, R19 (PEG dispersions F9). Nuclei formed on slow cooling also persist in the glass: the heating rate needed to avoid crystallization fell from 300 to 135 °C/min for cinnarizine and from above 1000 to 300 °C/min for dibucaine when the melt was cooled at the maximum rate baird2013 R9 (Crystallization tendency F16).

Figure

On the log axis the drug-alone point sits about 50-fold above every 3% point, and the three symbols overlap at most levels konno2006 R4, X9 (Physical stability F1).

Read the last column against the growth ranking: HPMCAS (2.5) and PVP (5.2) below the drug (10.8), PAA (26) far above it trasi2012a R12 (Physical stability F17).

Why

Classical theory gives both rates a maximum between Tg and Tm, with nucleation favoured at larger undercooling, so ready glass formation could come from low nucleation, low growth, both, or poor overlap trasi2012a C35; trasi2014b C3 (Crystallization tendency F3). For the polymers, a lower driving force "probably cannot account" for the nucleation drop, and the authors "speculate" that the polymers raise the barrier to moving molecules onto the nucleus konno2006 C34, C37, C39, C40 (Physical stability F1). For growth, hydrogen bonds that lower drug mobility are proposed, weakening on heating kestur2010a C43, C44 (Miscibility and drug-polymer interactions F10). The effects on nucleation are "not particularly straightforward" to explain with classical theory, which does not encompass molecular recognition, and the authors of both 2012 papers conclude that the effect of a polymer on nucleation is not predictable from simple properties, unlike growth trasi2012a C48, C49, C52; trasi2012b C45 (Physical stability F17). Crystalline PEG is read as "a favorable heterogeneous nucleation site" zhu2012 C4 (PEG dispersions F9). (synthesis: a hot-stage growth screen measures the transport of drug to the crystal surface, which hydrogen bonding and polymer mobility can change (F6); nucleation also depends on the surface the nucleus forms on, which may be the polymer, a crystalline carrier or a free surface (F11). A ranking from one step therefore says little about the other.)

Conditions and limits

The acetaminophen counts add up nucleation over the time spent in each zone, and the PAA zone is four times wider trasi2012a X8 (Physical stability F17). The felodipine nucleation effect was not measured in the powders; it rests on rate ratios and a cited 150-fold value kestur2013b X7, R22 (Physical stability F17). Nilutamide nuclei counts are counts per sample, not rates, and zones were measured for three neat drugs only trasi2014b X7 (Crystallization tendency F3, Crystallization tendency F4). The dry equality of the polymers holds for nucleation and only when dry konno2008 C10 (Physical stability F1).

Counterexamples and conflicts

For felodipine and PVP both steps fell: the authors write that "it appears that PVP must" also inhibit nucleation strongly in the powders kestur2013b C35, C36 (Physical stability F17), so the decoupling is not universal. At 33–75% RH the three polymers that were equal when dry separated, PVP being weakest at low content konno2008 C10 (Physical stability F1). Polymers can also move the two steps in opposite directions, as PAA does in acetaminophen trasi2012a R6, R12 (Physical stability F17).

Transfer check

After a growth screen, test the best and the weakest polymers for nucleation as well, with a cooling–reheating DSC zone search and a nuclei count after one cycle trasi2012a R12, C46 (Physical stability F17). Locate the zone by cooling to successively lower end points, extending them well below Tg, and note that separated windows predict glass formation on cooling, not stability on reheating or in storage trasi2012a P6; trasi2014b R5; baird2010 C55 (Crystallization tendency F3). Before cold storage hold the glass at the lowest temperature it will meet, reheat, and identify any crystals by polymorph trasi2014b P8 (Crystallization tendency F4). Record and control the cooling path in melt processes baird2013 R9, C28 (Crystallization tendency F16). Count nucleation sites in dry films with 3, 7 and 15% of each candidate polymer, then repeat at the storage humidity konno2006 R4; konno2008 C10 (Physical stability F1).

History

2006: Konno and Taylor and Marsac and co-workers measured the dry nucleation drop with PVP, HPMC and HPMCAS konno2006 R4, C13 (Physical stability F1). 2010: Baird and co-workers proposed that class II glasses keep nuclei formed during cooling, and Kestur and Taylor ranked four polymers by growth baird2010 C55, C58 (Crystallization tendency F16); kestur2010a R6 (Miscibility and drug-polymer interactions F10). 2012: Trasi and Taylor found that polymers act on nucleation and growth in unrelated ways in acetaminophen and flutamide, and Zhu and co-workers found that crystalline PEG nucleates two slow crystallizers trasi2012a R12, C46; trasi2012b R11 (Physical stability F17); zhu2012 C4 (PEG dispersions F9). 2013–2014: Baird and co-workers tied cooling-rate effects to quenched-in nuclei, Kestur and co-workers inferred a strong nucleation effect of PVP in powders, and Trasi and Taylor found nucleation of nilutamide far below Tg baird2013 C28, C30 (Crystallization tendency F16); kestur2013b C34, C36 (Physical stability F17); trasi2014b R5, R9 (Crystallization tendency F4).

How it was shown

Trasi and co-workers (2014) melted 3–5 mg of each of 31 drugs between coverslips, quenched, seeded the film and filmed the growth front on a hot stage from about 30 to 90 °C above Tg; for five donor drugs they added 10 mol% PVP K-12 and measured at 40–120 °C trasi2014a P1, A1, P3, A2, C9 (Crystallization tendency F1, Crystallization tendency F19). Kestur and Taylor grew felodipine form II with 0.5–4.5 wt% PVP K-12, K-29/32 or K-90 at 70–110 °C kestur2010b P1, P2, P3 (Physical stability F3), felodipine form II with three further polymers at 3% kestur2010a P4 (Miscibility and drug-polymer interactions F10), bifonazole and nimesulide with 5% polymer at 18–100 °C kestur2011 P3, P4 (Miscibility and drug-polymer interactions F13), and the two felodipine polymorphs with 3% PVP kestur2013a P1, P2, P3, P6 (Physical stability F19). Tian and co-workers grew three indomethacin polymorphs with 10% w/w HPMC, HPMCAS or PVP from 80 °C to near each melting point tian2017 P1, P2, P3, P4 (Physical stability F29); Sato and Taylor blended six compounds with PHB sato2016 M1, P1, P2, P3 (Physical stability F28); Mosquera-Giraldo and co-workers cryomilled celecoxib with 10 wt% surfactant or PVP mosqueragiraldo2014 P1, P3, X1 (Physical stability F22).

What the data show

The neat rate peaks between about 0.90 and 0.96 of the melting temperature: vemurafenib α at 250 °C, indomethacin δ at 110 °C and α and γ at 135 °C, and felodipine form II near 105 °C lu2016 R12, R19.a; tian2017 R7, E3 (Crystallization tendency F2); kestur2010a R5.a, C10 (Crystallization tendency F1). Below the peak the rate falls steeply, about 51-fold from 105 to 70 °C for felodipine form II kestur2010a R5.a, C10, X1 (Crystallization tendency F1). Fast growth does not mean poor glass formation: among the class III drugs, which crystallized neither on cooling nor on reheating, the maximum growth rate still spanned 216 to 0.02 µm/s, so their glass formation reflects slow nucleation rather than slow growth trasi2014a R1, C14, C15 (Crystallization tendency F15). For felodipine, bifonazole and nimesulide the polymer factor is larger where the neat rate is lower: felodipine with 3 wt% PVP K-12, K-29/32 and K-90 was slowed 11-, 15- and 21-fold at 70 °C but 3-, 3- and 6-fold at 110 °C, and the logarithm of the rate fell linearly with concentration, each added wt% multiplying the rate by about 0.44 at 70 °C and 0.61 at 105 °C (synthesis: e^{-0.822} and e^{-0.486}) kestur2010b R8, R14, C19, E1 (Physical stability F3). With 3% of four polymers at 70 °C the ratios were 10.0 (PVP), 6.5 (PVP/VA), 4.4 (HPMCAS) and 1.2 (PVAc), falling to about 1.1 at 110 °C except PVP (2.9) kestur2010a R6, X2 (Miscibility and drug-polymer interactions F10). For bifonazole PAA gave about 15-fold at 18 °C and about 5-fold at 100 °C, PVP K-29/32 and HPMCAS about 2–6-fold kestur2011 R7 (Miscibility and drug-polymer interactions F13); for nimesulide PVP gave about 21-fold at 18 °C and 4.7-fold at 80 °C kestur2011 R9 (Miscibility and drug-polymer interactions F13). The factor was larger for a faster-growing polymorph only in the sense that the faster crystal stays faster: bulk form I of felodipine grew 15–18 times faster than form II at 60–70 °C and 3% PVP lowered the rate of the two forms by about 15.8 and 18.7 at 60 °C and 3.5 and 6.5 at 100 °C, which the authors read as "will require more polymer" for the faster form kestur2013a R8, R15, C14, C26 (Physical stability F19).

The factor depends on the crystal form and the additive: indomethacin polymorphs differed widely, with HPMC the ratios at 120 °C were about 2900 for α against 2.9 for γ, with PVP 34 against 3.9 and with HPMCAS 17 against 1.6, and HPMC moved the temperature of fastest α growth from 135 to 100 °C tian2017 R12.i, C37, R15, C34 (Physical stability F29). Additives that mix and lower Tg accelerated growth: 5% PHB raised the rate of griseofulvin from about 0.015 to 0.15 µm/s at 120 °C and of celecoxib about sixfold at 90 °C sato2016 R4.a, R4.c, R6 (Physical stability F28), and 10% TPGS raised that of celecoxib about 85-fold at 70 °C although all three surfactants were close to the drug near 120 °C mosqueragiraldo2014 R1, C9, C36 (Physical stability F22).

Figure

Each drug's PVP series runs roughly parallel to and below its neat series on the log axis, for the fast flutamide and the slow indomethacin alike; the one exception is the nilutamide point at 70 °C trasi2014a R6, X6 (Crystallization tendency F19).

In the lower panels the α series shoot up at high temperature while γ stays low tian2017 R12 (Physical stability F29).

Why

The authors place growth between two opposing factors: near Tm molecules are mobile but the driving force is small, and on cooling the driving force rises while mobility falls, so the product passes through a maximum trasi2014a C22 (Crystallization tendency F1). Below the temperature of fastest growth mobility limits growth and above it the driving force does tian2017 C22 (Physical stability F29). The polymer effect is attributed mainly to specific drug–polymer interactions: for nilutamide the interaction strength "correlates well" with the inhibition, and for the five 2014 drugs similar hydrogen bonding with PVP is assumed ("would be expected"), not measured trasi2014b C19; trasi2014a C9, C12 (Crystallization tendency F19). The weakening of drug–PVP bonds on heating "would also provide a potential explanation" for the weaker inhibition at high temperature, which Kestur and Taylor call a "speculation" kestur2010b C35, C37 (Physical stability F3); kestur2010a C43, C44 (Miscibility and drug-polymer interactions F10). Equal inhibition of two felodipine polymorphs points, "most likely", to an effect of PVP on the liquid and not on the crystal faces kestur2013a C1, C6 (Physical stability F19), whereas the indomethacin polymorphs, which also grow from one liquid, were slowed very unequally, so "it appears" the polymers act on each polymorph's crystal–melt interface tian2017 C39 (Physical stability F29). Miscible additives are read as small molecules that depress Tg and raise mobility mosqueragiraldo2014 C29, C30 (Physical stability F22). (synthesis: a multiplicative factor on the neat rate is what the mobility reading predicts when the polymer lowers drug mobility by the same proportion at each temperature; its fall near the growth maximum then follows from the weakening of the contact, but no paper measured the contact at growth conditions.)

Conditions and limits

The 31-drug set has no replicates, the MCV derivation is not stated and the molar basis of the PVP comparison is undefined, so the mass fraction of PVP differs between drugs trasi2014a X17, X7 (Crystallization tendency F19). Every rate is for melts above Tg and for the polymorph that the seed gave trasi2014a C6 (Crystallization tendency F1); storage is below Tg, where other growth modes occur (F11). The PVP concentration law is fitted to 70–105 °C for one drug and one form kestur2010b A1, C21 (Physical stability F3). The indomethacin ratios of 2900 rest on single, very slow points tian2017 X16 (Physical stability F29).

Counterexamples and conflicts

The factor did not always rise on cooling: for acetaminophen it was about 4.7 at 70 °C and about 9 at 100 °C, and for flutamide and celecoxib it showed no trend (synthesis) trasi2012a R6; trasi2012b R3; mosqueragiraldo2014 R1 (Physical stability F3). PAA, the best inhibitor of bifonazole, did not differ significantly from neat flutamide trasi2012b R3, C7 (Miscibility and drug-polymer interactions F13). The similar factor for felodipine's two forms differs from the unequal factors for indomethacin's three kestur2013a R15; tian2017 R12.i (Physical stability F19, Physical stability F29). Additives can lower the rate by more than an order (PVP with celecoxib) or raise it about 85-fold (TPGS) mosqueragiraldo2014 R1, C9 (Physical stability F22).

Transfer check

Measure the neat growth rate of the polymorph of interest at three or more temperatures, the reduction factor of the candidate polymer at two temperatures including one close to storage, and confirm the form by Raman or XRPD trasi2014a C7, C11; kestur2010b R13; lu2016 P4 (Crystallization tendency F1, Crystallization tendency F19). Measure the polymer effect on the polymorph that forms in the product, below about Tg + 15 °C as well as above, and on both sides of the temperature of fastest growth when the product meets high temperature, as in hot melt extrusion kestur2013a X9, X11 (Physical stability F19); tian2017 R15 (Physical stability F29). Before using a low-Tg polymer or surfactant with a high-Tg drug, measure the mixture Tg sato2016 R7, X7 (Physical stability F28).

History

2010: Kestur and Taylor measured the polymer order and the log-linear concentration law for felodipine kestur2010a R6; kestur2010b R14 (Miscibility and drug-polymer interactions F10, Physical stability F3). 2011: Kestur, Van Eerdenbrugh and Taylor showed that the best polymer changes with the drug kestur2011 R7, R9 (Miscibility and drug-polymer interactions F13). 2013: Kestur and Taylor found similar factors for two felodipine polymorphs kestur2013a R15, C26 (Physical stability F19). 2014: Trasi and co-workers extended the similar factor to five drugs and four orders of neat rates, and Mosquera-Giraldo and co-workers found surfactants that accelerate growth trasi2014a C10, C11 (Crystallization tendency F19); mosqueragiraldo2014 R1 (Physical stability F22). 2016–2017: Lu and Taylor and Tian and co-workers located the growth maximum near 0.94–0.96 Tm and showed polymorph-specific factors, and Sato and Taylor showed PHB accelerating high-Tg compounds lu2016 R12 (Crystallization tendency F2); tian2017 R12.i, R15 (Physical stability F29); sato2016 R4, R11 (Physical stability F28).

How it was shown

The designs fall into four kinds. Growth of drug melts at 3–10% w/w polymer: felodipine, bifonazole and nimesulide kestur2010a P4; kestur2011 P3 (Miscibility and drug-polymer interactions F17), flutamide and acetaminophen with simple polymers carrying only acceptors, only donors or both trasi2012b P1, M3, M4, M5, M6, M7 (Physical stability F18), and the enantiomers of 2-amino-1,1,3-triphenyl-1-propanol with five polymers at 5% w/w sato2015 M1, P1, P3 (Physical stability F27). Spin-coated films of 8 and 15 drugs with 7 polymers at several ratios, stored dry for 7 days and scored by polarized light as an amorphicity index vaneerdenbrugh2010b P1, P2 (Miscibility and drug-polymer interactions F15); vaneerdenbrugh2011a P1, P2, P3, E1 (Miscibility and drug-polymer interactions F24). Storage of dispersions of naringenin, quercetin, resveratrol, curcumin and lumefantrine at humid and hot conditions wegiel2013a P1, P2, P3 (Miscibility and drug-polymer interactions F23); wegiel2013b P1, P3, P4 (Physical stability F24); hiew2022a P8 (Miscibility and drug-polymer interactions F23). Phase diagrams and spectra of nine drugs with PEG 3350 and the melting and crystallization of PEG with drugs baird2011 P1, P2, P3 (PEG dispersions F1); zhu2012 P4, R8 (PEG dispersions F10). Infrared bands of drug and polymer were read in spin-coated films at 20–50% polymer, not at the 3–5% used for growth kestur2010a P6, P7, C6 (Miscibility and drug-polymer interactions F12).

What the data show

Felodipine, an NH donor, was inhibited most by the acceptor polymer PVP, and its NH band moved to 3294 cm⁻¹ with PVP and 3298 with PVP/VA but up to 3348 and 3354 cm⁻¹ with HPMCAS and PVAc kestur2010a R6, R9 (Miscibility and drug-polymer interactions F12). Bifonazole, which has an imidazole acceptor and no donor, was slowed most by PAA, whereas for nimesulide, whose NH is held in an intramolecular bond (NH stretch 3345 cm⁻¹ at all concentrations in dichloromethane), PVP, PVP/VA and PVAc differed only modestly kestur2011 R7, R9, C27 (Miscibility and drug-polymer interactions F13, Miscibility and drug-polymer interactions F14); kestur2011 R11 (Miscibility and drug-polymer interactions F14). Flutamide's NH band fell from 3315 to 3287 cm⁻¹ with PVP, and the acceptor polymers slowed growth most (PVP about 0.13 against 1.4 µm/s for the drug), whereas the donor polymers PVPH and PAA did little trasi2012b R3, C7, R14, R15 (Physical stability F18). For acetaminophen, whose carbonyl is free, the donor PAA was the best growth inhibitor trasi2012a R6, C31 (Physical stability F18). In the spin-coating screens the carboxylic acids and acidic N–H drugs did best with E100, PVP and PVPVA (amorphicity index about 54–100) and worst with PAA and PSSA (about 4–31), the amides and bases the reverse (PSSA about 68–100, PAA about 61–71, E100 about 13–58) vaneerdenbrugh2011a R3 (Miscibility and drug-polymer interactions F24); vaneerdenbrugh2010b R3 (Miscibility and drug-polymer interactions F15). In storage, naringenin and quercetin with E100 or PVP showed no crystals for more than 200 to more than 450 days at all five conditions, and every dispersion with HPMCAS, HPMC, CMCAB or PAA crystallized within 1–3 days at 84% RH/25 °C and at 75% RH/40 °C wegiel2013a R9 (Miscibility and drug-polymer interactions F23); for resveratrol PVP and E100 were amorphous for more than 400 days at 84% RH, and for curcumin E100 kept the dispersion amorphous for more than 350 days while PVP, which gave the largest IR shifts, lasted 22 days wegiel2013b R12 (Physical stability F24); wegiel2014 R13, R3.f, R4.a (Physical stability F24). Lumefantrine, a weak base, was 19–55% protonated by four enteric polymers and 0% by PVPVA, which crystallized after 1–8 weeks while the enteric dispersions crystallized only at 50% loading hiew2022a R20, R6 (Miscibility and drug-polymer interactions F21, Miscibility and drug-polymer interactions F23). A contact other than a hydrogen bond was weaker: for four halogenated compounds without donors the polymer carbonyl moved by only 2–6 cm⁻¹ and the binding constant with a PVP analogue was 0.56–1.11 M⁻¹, against shifts of 25–45 cm⁻¹ for donor drugs que2020 R11, R12 (Miscibility and drug-polymer interactions F20); marsac2009 R12 (Miscibility and drug-polymer interactions F2). The cleanest test used a chiral polymer: with HPMCAS-HF ATP(S) grew 1.9–2.5 times faster than ATP(R), with HPMCAS-LF ATP(R) 1.5–1.9 times faster than ATP(S), and the enantiomer that lost more of its free NH2 band was slowed more sato2015 R7, R8, R13, R14 (Physical stability F27, Physical stability F18).

Outside amorphous polymers, the drugs with hydrogen-bond donors had the highest eutectic compositions in PEG 3350 (30–40 wt% for chlorpropamide, aceclofenac, benzocaine and ibuprofen) and the same donor drugs slowed PEG spherulite growth most at 40 °C (1 µm/s with benzocaine and 3 µm/s with ibuprofen against 10 µm/s without drug) baird2011 R1.a, R1.b; zhu2011 R5.b, R5.d (PEG dispersions F1); zhu2012 R22 (PEG dispersions F10). For amorphous salts, every counterion raised the Tg of propranolol and nicardipine, but crystallization on heating and at 79% RH and moisture uptake did not follow pKa or Tg towler2008 R4, R5, R12, R8, C14, C34 (Crystallization tendency F20).

Figure

Compare polymers for one colour (one drug), not drugs vaneerdenbrugh2011a R3 (Miscibility and drug-polymer interactions F24).

The PVP and HPMCP pairs overlap, while the HPMCAS-HF and HPMCAS-LF pairs split apart in opposite directions sato2015 R5 (Physical stability F27).

Why

The common reading is that a polymer slows crystallization better when it forms drug–polymer bonds more favourable than those in the drug crystal, or disrupts drug self-association, and that a drug molecule must first demix from the polymer at the growth front ("This speculation") vaneerdenbrugh2011a C6 (Miscibility and drug-polymer interactions F24); kestur2010a C42 (Miscibility and drug-polymer interactions F12). The screen authors attribute the reversal with drug chemistry to interactions that were not measured ("reasonable to speculate") vaneerdenbrugh2010b C33 (Miscibility and drug-polymer interactions F17). Flutamide's carbonyl keeps part of its charge in an intramolecular CH···O contact, which makes it "much less favorable" as an acceptor, and curcumin's intramolecular bond leaves an ionic interaction as the effective one ("appear to be the reason") trasi2012b C35, C37 (Physical stability F18); wegiel2014 C3, C37 (Physical stability F24). For the enantiomers the differential inhibition is "most likely due to" differences in hydrogen bonding of the amino group with the polymer carbonyls, and because everything but chirality is the same the authors read it as "strongly" supporting compound–polymer interactions sato2015 C26, C29, C27 (Physical stability F27). For lumefantrine the enteric polymers' resistance is rationalized by an acid–base interaction and the PVPVA failure by the absence of any specific contact hiew2022a C61, C62 (Miscibility and drug-polymer interactions F23). (synthesis: complementarity is a statement about which polymer has an unmatched group to offer the drug's free group; when the drug's group is engaged within the molecule, nothing is on offer and the polymers converge.)

Conditions and limits

Most designs are dry and the interaction was measured by IR for felodipine, nimesulide, flutamide and the polyphenols but not for bifonazole or the screen drugs kestur2011 X9; vaneerdenbrugh2010b C33 (Miscibility and drug-polymer interactions F17). The IR spectra were recorded at 20–50% polymer, while growth was measured at 0.5–5% kestur2010a C38, X5, X6 (Miscibility and drug-polymer interactions F12). The spin-coating "prediction" is qualitative and the exceptions were explained afterwards vaneerdenbrugh2011a C55, X5 (Miscibility and drug-polymer interactions F24). Polymers were compared at unequal content (30–80%) in the polyphenol storage study wegiel2013a X3 (Miscibility and drug-polymer interactions F23). The protection can cost release: with HPMCAS and HPMCP the drug release rate of lumefantrine compacts fell about 10-fold between 5 and 20% loading, while PVPVA released drug in step with the polymer up to 35% loading hiew2022a R13 (Miscibility and drug-polymer interactions F26).

Counterexamples and conflicts

The size of the IR shift did not rank the polymers: HPMCAS shifted the felodipine NH band up like PVAc yet slowed growth 4-fold at 70 °C kestur2010a R9.b, R9.c, R6 (Miscibility and drug-polymer interactions F12); HPMCAS, third by IR, gave the least stable dry naringenin dispersion wegiel2013a X4, R9 (Miscibility and drug-polymer interactions F25); and within the resveratrol and curcumin papers the IR ranking did not follow the storage ranking wegiel2013b X10; wegiel2014 X7 (Physical stability F24). A complementary polymer did not guarantee stability: PAA, which bonds to itself, mixed poorly with felodipine at 30–70% PAA rumondor2009a C44, C45 (Miscibility and drug-polymer interactions F17), and the 50% drug-loading HPMCAS dispersion of lumefantrine, 1.61 mol drug per mol polymer acid groups, crystallized within a week hiew2022a R6, R26, C63 (Miscibility and drug-polymer interactions F21). For ionic contacts the extent of protonation correlated with stability in rotary-evaporated lumefantrine dispersions but not in powders precipitated into water, where residual water and drying time decided, and the counterion pKa did not rank the salts of propranolol and nicardipine trasi2020 C14, C36 (Physical stability F34); bhujbal2021 C17, C50 (Physical stability F34); towler2008 C34 (Crystallization tendency F20). In PEG a donor group alone did not decide: haloperidol has an OH group but is monotectic baird2011 M4, R10, C32 (PEG dispersions F1). For the effect of humid storage on the polymer order see F9.

Transfer check

List the drug's donors and acceptors, remove those involved in intramolecular hydrogen bonds (a dilution series in a non-polar solvent, a strong added acceptor, amorphous against crystalline bands), choose polymers whose groups are complementary and one that is not, rank them in a spin-coated screen and confirm by growth at 3–5% polymer and a bulk preparation vaneerdenbrugh2011a C50, C53 (Miscibility and drug-polymer interactions F24); kestur2011 C21, C27, R11, R17 (Miscibility and drug-polymer interactions F14); kestur2011 C15, C29 (Miscibility and drug-polymer interactions F17). For a phenolic drug try a very strong acceptor and a basic polymer first and read the IR for the type of interaction rather than the size of the shift wegiel2013b R17 (Physical stability F24). For a basic drug screen acidic polymers for salt formation, count moles of drug per mole of polymer acid group, measure residual water and drying time of any process that uses water, and store one portion dry as a control hiew2022a R19, R20, R26 (Miscibility and drug-polymer interactions F21); trasi2020 R11 (Physical stability F34). For a single-enantiomer drug screen on the enantiomer itself and test at least two grades of HPMCAS sato2015 R5, X5 (Physical stability F27).

History

1997: Taylor and Zografi showed that the indomethacin acid OH bonds to the PVP carbonyl at the expense of the acid dimers taylor1997 R10, C33, C35 (Miscibility and drug-polymer interactions F1). 2009: Marsac and co-workers put the PVP carbonyl shifts of four compounds on one scale (45 cm⁻¹ for indomethacin, 25 for felodipine, 0 for ketoconazole) marsac2009 R12, R14 (Miscibility and drug-polymer interactions F2). 2010–2011: Kestur and co-workers tied the polymer order to hydrogen-bond strength for felodipine and found it flattened for nimesulide, while Van Eerdenbrugh and Taylor showed in screens of 8 and then 15 drugs that no polymer was best for all kestur2010a R6, R9 (Miscibility and drug-polymer interactions F12); kestur2011 R7, R9 (Miscibility and drug-polymer interactions F13, Miscibility and drug-polymer interactions F14); vaneerdenbrugh2010b C20 (Miscibility and drug-polymer interactions F15); vaneerdenbrugh2011a R3 (Miscibility and drug-polymer interactions F24). 2012–2014: Trasi and Taylor found donor polymers failing on an intramolecularly held acceptor, and Wegiel and co-workers found the acidic polyphenols stable with PVP and E100 trasi2012b C35 (Physical stability F18); wegiel2013a R9 (Miscibility and drug-polymer interactions F23); wegiel2013b R12 (Physical stability F24); wegiel2014 R13 (Physical stability F24). 2015: Sato and Taylor gave the cleanest test with enantiomers sato2015 R13, R14, C27 (Physical stability F27). 2022: Hiew and co-workers and Trasi and co-workers showed the ionic contact for lumefantrine and its limits hiew2022a R20 (Miscibility and drug-polymer interactions F21); trasi2020 R11, C36 (Physical stability F34).

How it was shown

Konno and Taylor (2006) and Marsac and co-workers (2006) counted nucleation sites in felodipine and nifedipine films with 0–25% polymer konno2006 P1, P2, E1 (Physical stability F1). Kestur and co-workers (2010) grew felodipine form II with 0.5–4.5 wt% PVP K-29/32 at 70–105 °C and fitted the logarithm of the growth rate against concentration kestur2010b P1, P2, P3 (Physical stability F3); Kestur and Taylor (2010) grew it with four polymers up to 4.5% w/w at 90 °C kestur2010a P2, P3, P4 (Miscibility and drug-polymer interactions F10). Trasi and co-workers (2014) compared five donor drugs with 10 mol% PVP trasi2014a P3, A2 (Crystallization tendency F19). Rumondor and co-workers (2009) stored felodipine dispersions with 10–50 wt% PVP or HPMCAS at 0–93% RH for up to 724 days and quantified crystalline drug by PXRD rumondor2009b P1, P3, P5, P7 (Physical stability F11, Physical stability F15). Tian and co-workers (2016) read indomethacin's own fluorescence in drop-cast films with 5–95% drug in HPMC, HPMCAS or PVP and stored rotary-evaporated powders with 10–90% drug over desiccant at 40 °C tian2016 P1, P3, P4, P5, P2, P7, P8 (Physical stability F30). Hiew and co-workers (2022) stored lumefantrine dispersions at 5–50% drug loading at 40 °C/75% RH hiew2022a P8 (Miscibility and drug-polymer interactions F23).

What the data show

Nucleation: the drug-alone rate was about 1.3 × 10⁵, 2.3–3.0 × 10³ with 3% polymer and about 3 × 10² m⁻³ s⁻¹ with 25% PVP, so 3% did most of the work; nifedipine and felodipine with PVP fell about two orders at 7 wt% and 3–4-fold further to 25 wt% konno2006 R4, C13 (Physical stability F1); marsac2006a R16, C18 (Physical stability F1). Growth: the logarithm of the rate fell linearly with concentration at 90 °C, the same polymer order at every level kestur2010a R8, C15 (Miscibility and drug-polymer interactions F10), with slopes from −0.82 per wt% at 70 °C to −0.49 at 105 °C, and longer chains added little (K-29/32 about 1.55 times as effective as K-12, K-90 only 1.4 times as effective as K-29/32) kestur2010b R14, R15, R8, C32 (Physical stability F3). The five donor drugs were slowed by a similar factor whatever their neat rate, and the faster felodipine polymorph stayed faster, so more polymer is needed for a given rate trasi2014a R6, C10, C11 (Crystallization tendency F19); kestur2013a C26 (Physical stability F19).

Ceiling: indomethacin's emission maximum fell to the drug-rich range between 50 and 60% drug with HPMC, 20 and 30% with HPMCAS and 70 and 80% with PVP tian2016 R2.f, R3.c, R4.h (Physical stability F30). HPMC powders at 60–90% drug crystallized in 6–15 days and stayed amorphous to 50% drug for 14 months, HPMCAS powders at 40–90% drug crystallized in 2–10 days and stayed amorphous to 30% drug for 13 months, and PVP powders at 10–90% drug were amorphous at 3 months tian2016 R17, R16 (Physical stability F30). Telaprevir films carried drug-rich domains from 10% drug with HPMC and PVPVA and from about 40% with HPMCAS li2016c R4, R5, R6 (Miscibility and drug-polymer interactions F19). For nifedipine the predicted solubility in PVP is so low that more than 90% polymer would be needed to stay below saturation, and in PEG 20/80 dispersions of benzocaine and ibuprofen lay below the estimated solubility yet crystallized within hours marsac2006b R18 (Miscibility and drug-polymer interactions F9); zhu2012 R23.a, X13 (PEG dispersions F4).

After separation: the three drug–PVP dispersions that separated at 94% RH crystallized after 22 h, 48 h and 6 days, the two that stayed mixed after 46 and 125 days rumondor2009c R7, R9, R12, R14, R16 (Physical stability F15). In felodipine–PVP powders at 93% RH the initial rates for 10, 25 and 50% PVP were about 6.4, 5.8 and 5.5 g/day and crystallization stopped within about 20 days at plateaus of about 70, 80 and 75% rumondor2009b R21, R11 (Physical stability F15); growth in 2.5–10% PVP films converged from 64% RH rumondor2010 R5, C21 (Physical stability F15). At 75% RH 3% PVP gave a nucleation rate only about threefold below the drug alone konno2008 R3.b, C12 (Physical stability F10).

Figure

Crystallization onsets appear only at loadings to the right of the miscibility band (50–60% and 20–30% drug), and loadings that stayed amorphous carry no point tian2016 R17, R21 (Physical stability F30).

Nearly the same initial rate for the three PVP levels, falling to zero within about a month rumondor2009b R21 (Physical stability F15).

Why

The authors rule out a dilution effect on nucleation, because a pure diluent would scale with polymer fraction and a 1.5-order drop at 3% does not konno2006 C35, C37 (Physical stability F1). For growth the effect is "most likely kinetic in origin", through hydrogen bonds that lower drug mobility kestur2010b C23, C34, C35 (Physical stability F3). Above the miscibility limit a drug-rich phase forms, and with less polymer in that phase crystallization "would be anticipated" to occur more readily tian2016 C4 (Physical stability F30). After separation the phase compositions are set by the extent of immiscibility, so the overall rate "would be expected" to become independent of the amount of polymer, "exactly" as seen at 93% RH rumondor2009b C21, C22 (Physical stability F15). (synthesis: the three limits come from different mechanisms. A few percent polymer addresses the barrier to nucleation. More polymer lowers the growth rate in step with the contact between drug and polymer, so the drug with the larger neat rate needs more. The loading ceiling is a phase boundary and does not respond to adding polymer to a phase that already holds little of it.)

Conditions and limits

The nucleation data are dry thin films at 22 °C for two dihydropyridines konno2006 P1, P2 (Physical stability F1). The growth data are for form II of one drug at 70–105 °C kestur2010b A1, C21 (Physical stability F3). The ceiling is for one drug with three polymers; miscibility was read on drop-cast films and stability on powders, and PVP powders were followed for only 3 months tian2016 P8, X7, X8, X6 (Physical stability F30). Separation was not measured on the crystallizing powders rumondor2009b X7 (Physical stability F15).

Counterexamples and conflicts

The 30% HPMCAS powder held a drug-rich phase and stayed amorphous for 13 months tian2016 X5 (Physical stability F30). Separation does not always lead to crystals: nilutamide–PVP showed the separation pattern at all five PVP contents without crystalline nilutamide rumondor2011 R12, C24 (Physical stability F15). A complementary polymer at 50% loading did not keep lumefantrine amorphous, a case with 1.61 mol drug per mol of polymer acid groups hiew2022a R6, R26, C63 (Miscibility and drug-polymer interactions F21). Below the limit PVP kept every loading amorphous for 3 months tian2016 R16 (Physical stability F30), but at 3% PVP and 75% RH the nucleation rate was only threefold below the drug alone konno2008 R3.b, C12 (Physical stability F10).

Transfer check

Choose a loading at least one step below the miscibility limit, found with a method that resolves domains of a few hundred nanometres on a sample made by the intended route, not by DSC alone tian2016 X5, X7 (Physical stability F30). Measure the neat growth rate and the rate with 1.5 and 3–5% polymer at two temperatures, one near the storage temperature kestur2010a R6, M1 (Miscibility and drug-polymer interactions F10). Compare crystallization at two polymer levels at the storage RH: equal rates, as at 93% RH here, point to separation, and adding more of the same polymer will not then add stability rumondor2009b R21, C21 (Physical stability F15). Compare the nucleation of the dispersion and of the neat drug at 0 and 75% RH; a rise of two decades or more, as for 3% PVP or nifedipine–PVP, means the polymer level does not hold the water effect konno2008 R3.b, C12 (Physical stability F10).

History

2006: Konno and Taylor and Marsac and co-workers showed that a few percent polymer removes most dry nucleation konno2006 R4 (Physical stability F1). 2009: Rumondor and co-workers found crystallization at high RH independent of PVP content, with plateaus rumondor2009b R21, C20 (Physical stability F15). 2010: Kestur and co-workers found the log-linear concentration law and Rumondor and co-workers the convergence of growth with RH kestur2010b R14 (Physical stability F3); rumondor2010 R5, C21 (Physical stability F15). 2014: Trasi and co-workers concluded that fast growers need more polymer trasi2014a C11 (Crystallization tendency F19). 2016: Li and Taylor and Tian and co-workers tied the loading to the onset of a drug-rich phase and to storage li2016c R4, R5, R6 (Miscibility and drug-polymer interactions F19); tian2016 R17 (Physical stability F30). 2022: Hiew and co-workers added the loading dependence of lumefantrine stability hiew2022a R6 (Miscibility and drug-polymer interactions F23).

How it was shown

Purohit and Taylor (2015) spin coated itraconazole and HPMC from methanol–dichloromethane at 50:50 to 10:90, read the films with pyrene and prodan fluorescence, AFM-nanoIR and nanoTA, ran DSC on rotary-evaporated powders and heated films for 2 min at 165 °C purohit2015a P1, P2, P9, P10, P7, P8, P5 (Miscibility and drug-polymer interactions F18). Li and Taylor (2016) read telaprevir films with AFM height images and local AFM-IR band ratios li2016c P1, P2, P4, P5, P7, P8, P9 (Miscibility and drug-polymer interactions F19). Rumondor and co-workers (2009) compared infrared spectra, X-ray pair distribution functions and DSC of four dispersions rumondor2009a P1, P4, P5, P8, P9 (Miscibility and drug-polymer interactions F3). Tian and co-workers (2016) used indomethacin's own fluorescence in drop-cast films with 5–95% drug and stored powders with 10–90% drug at 40 °C over desiccant tian2016 P1, P3, P4, P5, P2, P7, P8 (Physical stability F30). For water, Rumondor, Marsac and co-workers stored drug–PVP films at 54–94% RH, dried them and compared their spectra with a calculated physical mixture, ran DSC on powders wet and re-dried, and added AFM and TEM rumondor2009c P1, P4, P7, E7, P9; marsac2010 P11, P14, P15; rumondor2010c P1, P7; rumondor2011 P1, P3 (Physical stability F12); Saboo and Taylor and Yang and co-workers varied how the water reached the film saboo2017 M1, M2, P1, A1, P4, P7; yang2022b P2, P4 (Physical stability F31).

What the data show

In the itraconazole films pyrene I/III was 1.39–1.42 against 1.39 for itraconazole and 1.54 for HPMC, a pyrene–itraconazole exciplex band appeared in every blend, nanoIR showed round domains of about 0.4–1.4 µm, DSC of powders gave one Tg per composition, and heating the films at 165 °C raised I/III to 1.50–1.53 and removed the exciplex band purohit2015a R1, R3, R10, R11, R7, R4 (Miscibility and drug-polymer interactions F18). In telaprevir films discrete domains appeared at 10% drug with HPMC and PVPVA (PVPVA about 280 nm wide at 10% and 380 nm at 50%) but only at about 40% with HPMCAS (about 120 nm) li2016c R4, R5, R6, R7 (Miscibility and drug-polymer interactions F19). For indomethacin the images became heterogeneous from 60% drug (HPMC), 30% (HPMCAS) and 70% (PVP) while DSC gave one Tg throughout tian2016 R7, R8, R9, R13 (Physical stability F30). Ketoconazole–PVP showed one Tg at every composition while IR changed only slightly and PXRD pointed to structure change at 70% PVP but not at 30%, so its phase state is unresolved rumondor2009a R3, R4, R10, R16 (Miscibility and drug-polymer interactions F4). Felodipine–PAA gave two Tgs only at 10–30% PAA and single Tgs that moved by 30–35 °C between two heating regimens at 70 and 90% PAA, while IR and the pair distribution function called it practically immiscible rumondor2009a R1, R13, R16, R22 (Miscibility and drug-polymer interactions F3). In dextran–PVP blends a pair that DSC called miscible (DEX6–PVP90) showed domains of about 0.3–0.6 µm in films vaneerdenbrugh2012a R13; vaneerdenbrugh2012c R16, R22 (Miscibility and drug-polymer interactions F22).

Water made the state in dispersions that were mixed when dry: pimozide–PVP 50:50 had one Tg dry (94 °C), two wet (−21 and 44 °C) and two after drying (61 and 112 °C), the lower at the Tg of amorphous pimozide; felodipine–PVP showed an event at about 45–50 °C instead of one Tg near 80 °C, and AFM roughness rose from 0.22 to 17 nm within 1 h at 94% RH rumondor2009c R17; marsac2010 R14, R19 (Physical stability F12). Felodipine–PVPVA, quinidine–PVP, quinidine–PVPVA, pimozide–PVPVA and pimozide–HPMCAS showed two Tgs in place of one after humid storage and drying, and five further drugs with PVP lost the bonded PVP carbonyl band at 70 and 90% PVP rumondor2010c R3, C10; rumondor2011 R14, C16 (Physical stability F12). The separated state stayed after a week of drying, and at 125 °C the drug–polymer to drug–drug band ratio went from 0.70 at 5 min to 1.41 at 1500 min against 1.47 for a film never exposed marsac2010 R8, R17, C47 (Physical stability F14). Miconazole–PVPVA films spin-coated at 45–60% RH instead of 20% held drug-rich pits from the start, and the dry powder kept one Tg after 60 min at 100 °C saboo2017 R6, R8, R22 (Physical stability F14); films stored at 75% RH formed drug-rich domains of 1–3 µm within a week saboo2017 R15, R16 (Physical stability F31).

Surfactants changed the timing and the pattern: in ritonavir–PVPVA films at high humidity and 37 °C, SDS and Tween 80 films were still homogeneous at 5 min and later held round drug-rich domains, whereas Span 20 and Span 85 films had separated by 5 min into coarse irregular regions, as the binary film did yang2022b R4, R5, R6, R7, R8 (Physical stability F32).

Water uptake also reports the state: the three drug–PVP blends that separated took up within 7.4% of the sum of their components at 95% RH, whereas the two that stayed mixed took up far less rumondor2009c R22, C24 (Physical stability F7).

Consequences for crystallization: the separating dispersions crystallized within 22 h (droperidol), 48 h (nifedipine) and 6 days (pimozide) at 94% RH, the mixed ones after 46 (indomethacin) and 125 days (ketoprofen) rumondor2009c R12, R7, R9, R14, R16 (Physical stability F15). In felodipine–PVP powders at 93% RH the initial rates for 10, 25 and 50% PVP were about 6.4, 5.8 and 5.5 g/day rumondor2009b R21 (Physical stability F15). Indomethacin powders crystallized in 2–15 days above the loading at which a drug-rich phase appeared and stayed amorphous for 13–14 months below it with HPMC and HPMCAS tian2016 R17 (Physical stability F30).

Figure

One step at 94 °C in the dry dispersion; two steps in the wet and in the re-dried sample, the lower re-dried one at the Tg of pure amorphous pimozide rumondor2009c R17 (Physical stability F12).

Round raised domains stand in a continuous matrix at each loading purohit2015a R10, R11 (Miscibility and drug-polymer interactions F18).

Why

A physical mixture is the weighted sum of its components, while molecular mixing adds new contacts, which is why IR and PDF can see what DSC leaves open, and domains below about 15–30 nm may give one Tg rumondor2009a C7, C9, C10, C6 (Miscibility and drug-polymer interactions F3). For the itraconazole powders the single Tg is explained as mixing during the first scan ("most likely") purohit2015a C39 (Miscibility and drug-polymer interactions F18). Sorbed water turns the binary into a ternary system and raises mobility; it interacts well with PVP and badly with the hydrophobic drug, so a lower free energy is reached by separating, and the authors state that the event is thermodynamically driven with mobility setting only its rate rumondor2009c C34, C37, C35, C38, C41 (Physical stability F12, Physical stability F13). For miconazole the dry system showed no separation even at 100 °C, so the gap is "not inherent", and water acts as an antisolvent saboo2017 C41, C42, C44 (Physical stability F31). After separation the drug-rich phase holds little polymer, so crystallization "would be anticipated" to be much faster rumondor2009c C5, C6, C3 (Physical stability F15). Above the miscibility limit of indomethacin the same applies tian2016 C4 (Physical stability F30). (synthesis: whichever step makes the drug-rich phase, the casting solvent, the water or the loading, its crystallization is that of a drug with little polymer, so the route and the loading enter the stability through the phase state and not through the average composition.)

Conditions and limits

DSC, IR and PXRD used samples with different thermal histories, and domains below 15–30 nm may give one Tg rumondor2009a X4, C6 (Miscibility and drug-polymer interactions F3). Miscibility of indomethacin was read on drop-cast films and stability on powders tian2016 P8, X7 (Physical stability F30). The humid-storage samples of 2010–2011 were all analysed after drying, wet two-phase evidence exists only for pimozide, and several key spectra are not shown rumondor2010c X12; rumondor2011 X7 (Physical stability F12). The telaprevir onset loadings come mainly from topography and humidity was not controlled li2016c X3, X9 (Miscibility and drug-polymer interactions F19).

Counterexamples and conflicts

Separation does not always lead to crystals: nilutamide–PVP showed the separation pattern at all five PVP contents without crystalline nilutamide rumondor2011 R12, C24 (Physical stability F15), and the 30% HPMCAS indomethacin powder held a drug-rich phase yet stayed amorphous for 13 months tian2016 X5 (Physical stability F30). The PVP limit of 70–80% drug may follow indomethacin dimers instead of a separate phase tian2016 C46, C50 (Physical stability F30). Water is not the only cause: itraconazole–HPMC was miscible as a dry binary and separated from a water-containing solvent mugheirbi2017 C39 (Miscibility and drug-polymer interactions F18). Indomethacin– and ketoprofen–PVP stayed mixed for 40 and 68 days at 94% RH rumondor2009c R14, R16, C20 (Physical stability F12).

Transfer check

Do not accept one Tg as proof of one phase for a new pair: add a method that reads local composition on a sample prepared as the product will be, and make the film or powder by the intended route rumondor2009a P4, P8, X14 (Miscibility and drug-polymer interactions F3); purohit2015a X1 (Miscibility and drug-polymer interactions F18). When drug and polymer Tg are close, look for domains by AFM at 10, 30 and 50% drug li2016c C22, C46, C49 (Miscibility and drug-polymer interactions F19). After any humidity excursion re-test the phase state after drying, and keep the water content of process solvents low, since 5 vol% water in the casting solvent was enough here marsac2010 C53, C48 (Physical stability F14). Store a film at 75% RH for 1 day and 1 week and at 94% RH for 1 h and 1 day, dry it and compare its spectrum with a calculated physical mixture rumondor2009c C11, X10 (Physical stability F12).

History

2009: Rumondor and co-workers showed that IR and PXRD complement DSC and that sorbed water splits PVP dispersions into two amorphous phases before crystallization rumondor2009a C47 (Miscibility and drug-polymer interactions F3); rumondor2009c R7, R17 (Physical stability F12). 2010–2011: Marsac and co-workers followed felodipine–PVP by IR, DSC, AFM and TEM and showed that the state is held after drying; Rumondor and Taylor and Rumondor and co-workers found it in nine further systems marsac2010 R14, R17, R19 (Physical stability F12, Physical stability F14); rumondor2010c R3 (Physical stability F12); rumondor2011 R14 (Physical stability F12). 2012: Van Eerdenbrugh and co-workers showed a DSC-miscible polymer pair with domains in films vaneerdenbrugh2012c R16 (Miscibility and drug-polymer interactions F22). 2015–2016: Purohit and Taylor and Li and Taylor imaged domains at the nanoscale in films where DSC showed one Tg, and Tian and co-workers tied the miscibility limit of indomethacin to storage purohit2015a R10, R11 (Miscibility and drug-polymer interactions F18); li2016c R4 (Miscibility and drug-polymer interactions F19); tian2016 R17 (Physical stability F30). 2017: Saboo and Taylor showed that the route of the water sets the pattern saboo2017 R5, R15, R18 (Physical stability F31).

How it was shown

Konno and Taylor (2008) counted nucleation in felodipine films with 0–25% PVP, HPMC or HPMCAS at 0–75% RH konno2008 P3, P4, P8, R14 (Physical stability F10, Physical stability F11). Rumondor and co-workers (2010) measured felodipine growth in films with 2.5, 5 and 10% w/w PVP or HPMCAS at 13–93% RH rumondor2010 P2, P4, P5 (Physical stability F11), and Rumondor and co-workers (2009) stored dispersions with 10–50 wt% PVP or HPMCAS at 0–93% RH for up to 724 days rumondor2009b P1, P3, P5 (Physical stability F11). Xie and Taylor (2016) stored celecoxib dispersions with five polymers and PVP-major ternaries at three conditions xie2016b P1, A1, P5, A4 (Physical stability F11, Physical stability F35). Wegiel and co-workers stored naringenin, quercetin, resveratrol and curcumin dispersions with six or seven polymers at five or six conditions, one composition per pair (30–80% polymer) wegiel2013a P1, P2, P3 (Miscibility and drug-polymer interactions F23); wegiel2013b P3, P4 (Physical stability F24); wegiel2014 P5, P6 (Physical stability F25). Hiew and co-workers and Trasi and co-workers stored lumefantrine dispersions with PVPVA and enteric polymers at 40 °C/75% RH hiew2022a P8 (Miscibility and drug-polymer interactions F23); trasi2020 P1, P5, P6 (Physical stability F34).

What the data show

At 75% RH pure PVP held about 33 g water per 100 g, HPMC about 10–11 and HPMCAS about 6, and the dry Tg advantage of 25% PVP (about 54–55 °C against 43–48 °C) was gone at 75–76% RH, where all systems sat near 28–36 °C konno2008 R8; rumondor2010a R1 (Physical stability F6); konno2008 R5, C15; marsac2008a R6, C17 (Physical stability F6). In nucleation the PVP/HPMCAS ratio at 75% RH fell from about 66-fold at 3% to 7-fold at 25% polymer konno2008 R3.b, R3.g, C10 (Physical stability F11). In growth PVP was the better inhibitor below 52% RH (10%: about 5.8 × 10⁻⁵ against 1.3 × 10⁻³ mm/min at 13% RH) and HPMCAS above about 64% RH (about 2.4 × 10⁻³ against 7.3 × 10⁻² mm/min at 93% RH) rumondor2010 R3, R5, C14 (Physical stability F11). For celecoxib at 40 °C/75% RH the PVP binary lasted 70 days against 33 for HPMCAS, and the ternaries with an eighth to a fifth of the PVP replaced by HPMCAS or HPMC behaved like the PVP binary at all three conditions xie2016b R24 (Physical stability F11); xie2016b R22, R23, R24, C38, C41, C42 (Physical stability F35). For the polyphenols, the same storage showed the opposite: no crystals for more than 200 to more than 450 days with E100 and PVP at all five conditions, crystals in 1–3 days with HPMCAS, HPMC, CMCAB and PAA at 84% RH/25 °C and 75% RH/40 °C wegiel2013a R9, C22 (Miscibility and drug-polymer interactions F23); curcumin with PVP lasted 22 days at 84% RH, the shortest of five polymers except the drug alone, although PVP gave the largest IR shifts wegiel2014 R13, R3.f, R4.a, C24 (Physical stability F24). Lumefantrine–PVPVA crystallized after 1, 2, 4 and 8 weeks at 50, 25, 10 and 5% drug loading, while enteric dispersions crystallized only at 50% loading hiew2022a R6, C10 (Miscibility and drug-polymer interactions F23). Water uptake alone did not give the order: celecoxib–PVPVA lasted 150 days at room temperature/94% RH against 115 for HPMCAS, against the water order xie2016b R23, C50 (Physical stability F23); the order of separation PVP > PVPVA > HPMCAS for the same drug did follow it rumondor2010c R18, C20 (Physical stability F23).

Figure

Read across the 75–93% RH rows: PVP columns 51–82%, HPMCAS columns 9–28%; at 0–52% RH every value is about 11% or less except 10% PVP, read at 724 days rumondor2009b R9, X10, X12 (Physical stability F11).

The E100 and PVP rows are all ">" (no crystals); the other rows are mostly 1–3 days at the humid and 90 °C conditions wegiel2013a R9 (Miscibility and drug-polymer interactions F23).

Why

For felodipine nucleation the differences between polymers "appear to be reasonably explained" by the water content of the dispersions; for growth, absorbed water "may result in disruption" of the specific interactions that PVP forms with felodipine; for bulk crystallization "It is hypothesized" that the hygroscopic PVP dispersions undergo moisture-induced immiscibility; none of these was measured on the stored samples konno2008 C30, X4 (Physical stability F11); rumondor2010 C40, C42, X12 (Physical stability F11); rumondor2009b C22, X7 (Physical stability F11). For the polyphenols PVP forms the strongest hydrogen bonds and E100 an ionic interaction, which "appear to be the reason" for the extremely good stability, while the cellulosics have only medium acceptors wegiel2013a C31, C35 (Miscibility and drug-polymer interactions F23). For celecoxib the small amount of HPMCAS did not disrupt the drug–PVP interactions and so did not change the kinetics, and in humid storage the hygroscopic PVP dominates xie2016b C45, C52 (Physical stability F35). (synthesis: polymer hygroscopicity brings water, and the drug–polymer contact has to persist against it. PVP gave stable dispersions with the polyphenols, whose OH donors meet its strong acceptor, and lost its advantage for felodipine, whose single NH donor is bonded to it by a weaker contact that sorbed water can displace; the data do not test this, because the contact was measured dry and in films, not in the stored powders.)

Conditions and limits

Water content and Tg come from bulk powders at 25 °C without replicates, while nucleation and growth were measured in films konno2008 X6; marsac2008a X9, X12 (Physical stability F6). The crossover RH depends on polymer content rumondor2010 X8 (Physical stability F11). The polyphenol study compared polymers at unequal content (30–80%) and measured IR on films, not on the stored powders wegiel2013a X3 (Miscibility and drug-polymer interactions F23). Within each polyphenol paper the IR ranking did not follow the storage ranking wegiel2013b X10; wegiel2014 X7 (Physical stability F24).

Counterexamples and conflicts

The ranking of PVP against the cellulosic polymers is opposite in two sets of studies: for felodipine and celecoxib PVP is the weaker inhibitor at high RH, for resveratrol, naringenin and quercetin it is among the two best, and for curcumin the weakest rumondor2009b R9 (Physical stability F11); wegiel2013a R9 (Miscibility and drug-polymer interactions F23); wegiel2014 R13 (Physical stability F24). The separation rule of the next finding puts OH and NH donors with a PVP carbonyl shift of 15–24 cm⁻¹ among the dispersions that separate under humidity rumondor2011 R15, C31, C32 (Physical stability F13), while the PVP carbonyl shift of naringenin and quercetin was 21 and 24 cm⁻¹ and these dispersions stayed amorphous for more than 200–450 days wegiel2013a R4, R9 (Miscibility and drug-polymer interactions F25). Phase separation was not measured in the polyphenol study and the polymer content differed (30–80% against 50%), so the two sets cannot be compared directly (synthesis). Sorbed water alone does not rank the polymers either: PVPVA, with less water uptake than PVP, lasted longer than HPMCAS for celecoxib at 94% RH xie2016b R23, C50 (Physical stability F23).

Transfer check

Rank candidate polymers at the intended storage humidity, at least 75% RH for open storage, not dry, and expect a hygroscopic polymer to lose ground for a drug whose only hydrogen-bond donor is an NH group, as felodipine konno2008 R3.b, C10 (Physical stability F11). Measure the isotherm at the storage temperature and the Tg of samples equilibrated at the storage RH, and treat a wet Tg near or below the storage temperature as a warning konno2008 R5, X6 (Physical stability F6). Test a ternary itself in one dry-heat and one humid condition and compare it with the binary of its main polymer, not with the binary of the minor one xie2016b R22, R23, R24 (Physical stability F35). For a phenolic drug, try PVP and a basic polymer under humidity first wegiel2013a R9, R6 (Miscibility and drug-polymer interactions F23, Miscibility and drug-polymer interactions F21); for a basic drug test acidic enteric polymers against PVPVA hiew2022a R6 (Miscibility and drug-polymer interactions F23).

History

2006: Konno and Taylor found PVP, HPMC and HPMCAS equal when dry konno2006 C14 (Physical stability F1). 2008–2009: Konno and Taylor and Rumondor and co-workers found PVP weaker at 64–75% RH and above for nucleation, growth and bulk crystallization of felodipine konno2008 R3, C10 (Physical stability F11); rumondor2009b R9, C13 (Physical stability F11). 2010: Rumondor and Taylor tied the separation order PVP > PVPVA > HPMCAS to water uptake rumondor2010c R18, C20 (Physical stability F23). 2013–2014: Wegiel and co-workers found PVP and E100 giving stable dispersions of resveratrol, naringenin and quercetin under humidity and PVP the weakest polymer for curcumin wegiel2013a R9 (Miscibility and drug-polymer interactions F23); wegiel2013b R12 (Physical stability F24); wegiel2014 R13 (Physical stability F24). 2016: Xie and Taylor showed that a small part of a cellulosic polymer does not change the behaviour of the PVP binary xie2016b C45 (Physical stability F35). 2020–2022: Trasi and co-workers and Hiew and co-workers showed enteric polymers outlasting PVPVA for lumefantrine trasi2020 R3, C6 (Physical stability F34); hiew2022a R6 (Miscibility and drug-polymer interactions F23).

How it was shown

Rumondor and co-workers (2009) tabulated log P, Tm, heat of fusion and the Hoffman driving force of six drugs, computed the drug–water χ from the uptake of the amorphous drugs at 95% RH, and read the hydrogen-bond strength from the shift of the PVP carbonyl in the dry dispersions rumondor2009c M1, M6, P10, P13, E5 (Physical stability F13); in 2011 they tabulated 20 drug–polymer systems rumondor2011 A1, P7, P9, R15, R16 (Physical stability F13). Rumondor and co-workers (2010) fitted binary Flory–Huggins isotherms to felodipine, indomethacin, PVP, HPMC and HPMCAS rumondor2010a E1, P4 (Physical stability F8). Marsac and co-workers pressed 500 mg compacts of PEO (molecular weight about 100000) with 10 wt% of a crystalline drug, stored them at 0% and 94% RH for up to 28 days and followed drug crystallinity by FT-Raman and XRPD; activity coefficients came from the ideal solubility and the solubility in PEG 400 and in water marsac2008b P1, P2, M7, P3, P4, P5, P8, P9, E2 (PEG dispersions F11).

What the data show

In PVP the nifedipine, felodipine, droperidol and pimozide dispersions separated and indomethacin and ketoprofen did not; log P (droperidol 1.92, felodipine 2.24, nifedipine 2.31 against ketoprofen 3.31 and indomethacin 3.58, pimozide 5.34) and the Hoffman driving force (4.1–11.0 kJ/mol) showed no trend, while the drug–water χ was pimozide 3.58, felodipine 3.18, droperidol 2.85 and nifedipine 2.84 against ketoprofen 2.71 and indomethacin 2.50 rumondor2009c R24, C27, C32, C46 (Physical stability F13). The PVP carbonyl shifted 20–24 cm⁻¹ with the separating NH-donor drugs and 42–44 cm⁻¹ with the two acids rumondor2009c R29, C48 (Physical stability F13). Binary fits gave water χ of 3.27 for felodipine, 2.86 for indomethacin and 0.36 for PVP, so the gap between the water–drug and water–PVP parameters is large rumondor2010a R11, C18 (Physical stability F8); the authors of the felodipine–PVP study call it the Δχ effect marsac2010 R22, C43 (Physical stability F12, Physical stability F13). For the polymer, less water gave less separation: for the same drug PVP separated most, PVPVA less and HPMCAS least, with the most hydrophobic drug, pimozide (log P 5.34), the exception, which separated slightly even with HPMCAS rumondor2010c R3, C35, R18 (Physical stability F23). In PEO, the loss of drug crystallinity at 94% RH was complete for acetaminophen in about 14–17 days, moderate for ketoprofen, naproxen and ibuprofen and negligible for felodipine and ketoconazole, following the activity coefficient in water (1.2 × 10¹ to 5.3 × 10⁵), whereas all six drugs had coefficients of 0.02–0.64 in PEG 400, including the two that stayed crystalline marsac2008b R2, R10, R14, R16 (PEG dispersions F11). PEO itself took up about 10 wt% water at 94% RH after several hours and about 37 wt% after 5 months, and its two main diffraction peaks lost intensity and shifted marsac2008b R11, R9 (PEG dispersions F12).

Figure

Only the last column, χ drug–water, puts the four separating drugs (nifedipine, felodipine, droperidol, pimozide) above the two that stayed mixed rumondor2009c R24 (Physical stability F13).

Why

For PVP the authors reason that separation needs a positive enthalpy of mixing, expected when the three pair interactions differ strongly, as for water, PVP and a hydrophobic drug, and that the stronger COOH–PVP bonds "might lead to a greater resistance" rumondor2009c C42, C43, C47, C48, C49 (Physical stability F13); water interacts well with PVP and badly with the drug, so separating lowers the free energy rumondor2009c C37 (Physical stability F12). For PEO, compaction puts crystalline drug against amorphous PEO, which "can be dissolved by PEO" (a hypothesis); water lowers the driving force because every drug is far less soluble in water than in PEG 400, yet speeds the loss by plasticizing the amorphous PEO and slowly deliquescing it; the authors conclude that the transformation "was related to the ability of the drug to dissolve in the water-amorphous PEO" cosolvent marsac2008b C16, C17, C24, C25, C26, C30, C38 (PEG dispersions F11). The same activity-coefficient reasoning is used in both: a value below 1 in the analogue solvent reads as exothermic mixing and the large values in water as endothermic mixing marsac2008b C14 (PEG dispersions F3). (synthesis: in the PEO compact the drug dissolves in the water-containing carrier only if its mixing with water is not too unfavourable; in the PVP film a drug whose mixing with water is the most unfavourable, and whose contact with PVP is weaker, separates from the water-swollen polymer. The variable common to both is the drug–water interaction, and the reasoning is not a tested unification.)

Conditions and limits

The set of six drugs in PVP is small, the χ gap is 0.13, and bond strength and drug–water χ cannot be separated, because the drugs with weaker PVP bonds also have the highest drug–water χ rumondor2009c R24, R29 (Physical stability F13). The PEO work used one loading, one PEO grade (not PEG 3350), one high humidity and up to 28 days; the controls that make water necessary are "data not shown" marsac2008b X7, X6, X8 (PEG dispersions F11). The isotherm method and temperature of the PEO uptake are not printed marsac2008b X6 (PEG dispersions F12).

Counterexamples and conflicts

Indoprofen, a carboxylic acid with a 41 cm⁻¹ shift, separated slightly before crystallizing, against the donor rule rumondor2011 R11, C33, C46 (Physical stability F13). In PEO acetaminophen lost crystallinity at 75% RH although the cited deliquescence point lies above it marsac2008b R22, R3, X10 (PEG dispersions F12). Ketoconazole had an activity coefficient of 0.09 in PEG 400, a favourable value, and stayed crystalline in PEO marsac2008b R14.f, R14 (PEG dispersions F3, PEG dispersions F11). For the PVP groups, log P did not separate the drugs, yet the most hydrophobic drug separated even with the least hygroscopic polymer rumondor2010c C35 (Physical stability F23).

Transfer check

For a new drug in PVP or PVPVA check its hydrogen-bond donor, the drug–water χ from the sorption of the amorphous drug at 95% RH and the PVP carbonyl shift in the dry dispersion, and do not use log P, Tm or heat of fusion as a guide; if log P is about 5 or higher, test even an HPMCAS dispersion for separation rumondor2009c R24, R29 (Physical stability F13); rumondor2010c C35 (Physical stability F23). For a crystalline drug in a PEG or PEO matrix compute its activity coefficient in water: near 10¹ is the case to test first, near 10⁵ stayed crystalline here marsac2008b R14, R16 (PEG dispersions F11). Follow water uptake and the carrier's diffraction peaks over weeks to months, not hours marsac2008b R11 (PEG dispersions F12).

History

2008: Marsac and co-workers showed the loss of drug crystallinity in PEO and tied it to the activity coefficient in water marsac2008b R10, R14 (PEG dispersions F11). 2009: Rumondor and co-workers showed that the drug–water χ and the donor type, and not log P or melting data, grouped the PVP dispersions that separate rumondor2009c R24, C46 (Physical stability F13). 2010: Marsac and co-workers and Rumondor and co-workers tied the water–drug and water–PVP gap and the polymer's water uptake to separation marsac2010 C43 (Physical stability F13); rumondor2010a R11 (Physical stability F8); rumondor2010c R18 (Physical stability F23). 2011: Rumondor and co-workers added the 20-system table and a model that weighted the COOH donor, the carbonyl shift and pKa and dropped log P rumondor2011 R15, R16, R17 (Physical stability F13).

How it was shown

Kestur and Taylor (2013) grew felodipine form I at the free surface and in the bulk of melt-quenched films from 18 to 100 °C, three samples per point, and with and without 3% PVP at 30 °C; Kestur and co-workers ground and sieved amorphous felodipine and its 3% PVP dispersion into two fractions and followed crystallinity by Raman and PXRD kestur2013a R8, R9 (Crystallization tendency F21); kestur2013b P2, P3, P4, P5, P6, P7 (Physical stability F21). Trasi and co-workers (2014) compared surface and bulk growth of 14 drugs trasi2014a P2, R4, R5 (Crystallization tendency F21). Zhu and co-workers put a piece of crystalline PEG into supercooled ibuprofen or fenofibrate zhu2012 P7 (PEG dispersions F9). Baird and co-workers (2013) cooled melts of ten compounds at 20 °C/min or at the maximum rate and reheated them at a series of rates baird2013 P2, P4, X5 (Crystallization tendency F16). Van Eerdenbrugh and co-workers (2010) spun 200 µL of solution into films and, for class III combinations, let 400 µL evaporate slowly in a vial vaneerdenbrugh2010a P2, P7 (Crystallization tendency F14). Powders of 50/50 drug–polymer were made by rotary evaporation vaneerdenbrugh2010b P6, C38 (Miscibility and drug-polymer interactions F16). Trasi and co-workers (2017) stored six marketed tacrolimus capsules and nine model formulations (extra croscarmellose, milling, seeds, four ethanol granulations) trasi2017 P7, P14, P15, P16, P17, P18, P19, P20, A6 (Physical stability F33); Bhujbal and co-workers (2021) sprayed a dimethylacetamide solution into cold dilute HCl and dried the precipitate bhujbal2021 P1, P4, P6, P8 (Physical stability F34). Marsac and co-workers (2010) cast felodipine–PVP from solvent with 0.1 mL water per 1.9 mL, and Saboo and Taylor (2017) spin-coated at 20% or 45–60% RH marsac2010 P13 (Physical stability F14); saboo2017 P1, A4 (Physical stability F14).

What the data show

Surface and bulk rates of felodipine form I did not differ above 70 °C, and the surface was faster by about 6-fold at 18 °C kestur2013a R8, R9 (Crystallization tendency F21); with 3% PVP at 30 °C the ratio rose from 5.7 to 22.8 because PVP cut bulk growth about 7-fold and surface growth about 2-fold kestur2013b R12, C32 (Physical stability F21). Powders of the smaller sieve fraction crystallized first, about 46 against 12% at 24 h kestur2013b R3.e, R5 (Physical stability F21), and in the 14-drug comparison felodipine was fast at the surface but slow in the bulk trasi2014a R4, R5, C8 (Crystallization tendency F21). Coating spray-dried ezetimibe particles delayed crystallization at 40 °C/75% RH from 19 days (50% loading) and 12 days (70%) for uncoated particles to mostly crystal-free for two years duong2022c R16, R17, R18 (Crystallization tendency F21). A crystal grew at the interface of a PEG piece in supercooled ibuprofen and next to the PEG particle in fenofibrate, while the bulk liquid stayed clear zhu2012 R15, R19 (PEG dispersions F9). PEG 3350 itself passed 50% crystallinity at about 22–24 s at 25 °C and was complete within about 2 min, so a drug that did not crystallize first had to crystallize inside an already crystalline carrier, and only haloperidol crystallized before PEG zhu2012 R4, R7 (PEG dispersions F7); while amorphous, the interacting drugs sat in the interlamellar region, the long period being 16.1–16.5 nm against 11.6 nm for pure PEG, until crystallization returned it to 11.4–11.9 nm zhu2011 R18, R8 (PEG dispersions F5). Dibucaine needed more than 1000 °C/min reheating after 20 °C/min cooling and 300 °C/min after maximum cooling, and clofoctol cooled at a constant 20 °C/min lost its melting endotherm when reheated at 500 instead of 200 °C/min baird2013 R9, R12 (Crystallization tendency F16).

The route changed the outcome: under fast spin coating only 4 of 33 drugs changed class with the solvent and 68% of combinations matched the melt class, but slow evaporation made 22 of 31 class III combinations fully crystalline at the first look vaneerdenbrugh2010a R11, R16, R12, R13 (Crystallization tendency F14); film rankings carried to 50/50 rotary-evaporated powders for six of seven drugs, with phenacetin–PAA the exception (about 80–95% crystalline against a film only slightly crystalline at day 7) vaneerdenbrugh2010b R9.c, R2.ay, C44 (Miscibility and drug-polymer interactions F16). Of six marketed tacrolimus capsules of about 1:1 drug:HPMC only one generic crystallized, from 27% at 2 days to 88% at 3 weeks; extra croscarmellose, seeds and milling did not destabilize the innovator product, while an ethanolic granulation on lactose and croscarmellose dried at 60–70 °C with HPMC added afterwards crystallized within 14 days and with HPMC in the liquid or the bed it did not trasi2017 R6, R7, R12, R18, R19, C35 (Physical stability F33). In lumefantrine powders precipitated into water the CAP dispersion with 9% water showed crystals from the first day and the Eudragit L100 dispersions, dried 4 days instead of about 12 h, were crystalline as made, whereas HPMCP and HPMCAS at 20–40% drug stayed amorphous for 3 months bhujbal2021 R1, R4, R8, R10, R11, C17 (Physical stability F34). Water at casting also changed the outcome: miconazole–PVPVA films spin-coated at 45–60% RH held drug-rich pits of 0.3–0.7 µm, and felodipine–PVP cast with water in the solvent lost most of the drug–polymer band saboo2017 R5, R6, R8 (Physical stability F14); marsac2010 R13, C16 (Physical stability F14).

Figure

The dashed circle marks the ibuprofen crystal at the PEG edge; the surrounding supercooled ibuprofen stays clear zhu2012 R15 (PEG dispersions F9).

Only the bottom trace (Accord) carries the arrowed tacrolimus reflections at 8.5, 10.3 and 13.7° 2θ; the extra peaks in the Mylan and Panacea traces are lactose trasi2017 R6, R8 (Physical stability F33).

Why

Faster surface growth "may be attributed" to mobile molecules at a more liquid-like surface, and where such modes occur, extrapolating high-temperature growth rates to storage "may not be predictive" kestur2013a C16, R17, C7, C18 (Physical stability F21, Crystallization tendency F21). The particle-size effect of the pure drug "can be attributed in large part" to the surface-to-volume ratio kestur2013b C24, C25 (Physical stability F21). Crystalline PEG is read as "a favorable heterogeneous nucleation site" zhu2012 C4 (PEG dispersions F9). Reheating through Tg "should in principle" erase the cooling history, so a cooling-rate effect points to nuclei trapped in the glass baird2013 C36, C37 (Crystallization tendency F16). Slow evaporation keeps the solution supersaturated longer, so crystallization becomes "an obvious result" vaneerdenbrugh2010a C27, C28 (Crystallization tendency F14). For tacrolimus the result "strongly suggests" a processing cause, and the authors "believe" the generic holds amorphous drug coated on lactose and "speculate" that it was made by ethanol granulation with drying at elevated temperature and HPMC mixed in afterwards trasi2017 C28, C47, C49 (Physical stability F33). For the precipitated lumefantrine powders the authors conclude that stability was reduced by moisture during manufacture or storage bhujbal2021 C27, C43, C56 (Physical stability F34). During casting ethanol and dichloromethane are "expected" to evaporate first and concentrate the water, which disrupts the drug–polymer bonds marsac2010 C15, C16 (Physical stability F14). (synthesis: each interface or nucleus that the sample brings in lets crystallization start without the nucleation barrier of the bulk dispersion, and each water or heat history sets the phase state (F8) before storage begins, so the contact between drug and polymer (F6) has nothing left to act on.)

Conditions and limits

The surface and bulk data are for one polymorph of one drug, and in the 14-drug comparison surface and bulk growth were measured in different T/Tg windows kestur2013a R8; trasi2014a X16 (Crystallization tendency F21). The tacrolimus work is one lot per brand, open-dish storage outside the package, and the model granulations changed drying temperature and HPMC contact together trasi2017 A1, P7, X6, X7 (Physical stability F33). Slow evaporation was run on class III combinations only vaneerdenbrugh2010a P7 (Crystallization tendency F14). The dibucaine and clofoctol scans lasted minutes on neat drugs, and the highest reheating rates exceed the calibrated range baird2013 A1, X15, X10 (Crystallization tendency F16). The lumefantrine explanation varies water together with polymer bhujbal2021 X15, C56 (Physical stability F34).

Counterexamples and conflicts

The route did not always matter: under fast spin coating the solvent changed the class of only 4 of 33 drugs and the spin-coating class equalled the melt class in 68% of combinations vaneerdenbrugh2010a R11, R16, R17 (Crystallization tendency F14). Seeds, extra croscarmellose and milling did not destabilize the innovator tacrolimus product, so the seeds suggested in the fresh generic were not shown to be the cause trasi2017 R18, C30, C32 (Physical stability F33). Surface and bulk growth coincide above 70 °C kestur2013a R9 (Crystallization tendency F21). The dry powder of miconazole–PVPVA kept one Tg after 60 min at 100 °C, so its water-made separation is not inherent to the pair saboo2017 R22, C44 (Physical stability F14).

Transfer check

Test open-surface films or powders as well as sandwiched films, at storage-relevant temperatures near or below Tg, and store at least two particle-size fractions of the dispersion kestur2013a R9; kestur2013b R12 (Crystallization tendency F21, Physical stability F21). Place a piece of any crystalline carrier in the supercooled neat drug and look for interface crystals before relying on a neat-drug class zhu2012 R15, R19 (PEG dispersions F9). Record and control the cooling path in melt processes baird2013 R9, C28 (Crystallization tendency F16). Test the drug at the process drying rate, use a spin-coat film as a cheap ranking step and confirm the top and bottom candidates in a slower bulk preparation of the intended route vaneerdenbrugh2010a R13 (Crystallization tendency F14); vaneerdenbrugh2010b C5, C39, C44 (Miscibility and drug-polymer interactions F16). Stress-test the finished product and each batch, and if a product crystallizes faster than the pure amorphous drug look at where the drug was dried, at what temperature, and whether it ever met the polymer in solution trasi2017 A1, C28 (Physical stability F33). Measure the residual water and drying time of any process that uses water and store one portion dry as a control trasi2020 R11 (Physical stability F34); keep the humidity of casting low saboo2017 R5, R6 (Physical stability F14).

History

2010: Van Eerdenbrugh and co-workers showed that under fast spin coating the solvent rarely changed the crystallization class, unlike slow evaporation, and that film rankings mostly carry to powders; Marsac and co-workers showed water in the casting solvent preventing mixing vaneerdenbrugh2010a R12, R16 (Crystallization tendency F14); vaneerdenbrugh2010b C44 (Miscibility and drug-polymer interactions F16); marsac2010 R13 (Physical stability F14). 2012–2013: Zhu and co-workers found the PEG interface nucleating slow crystallizers, Baird and co-workers tied cooling-rate effects to nuclei, and Kestur and co-workers measured the surface and particle-size effects zhu2012 C4 (PEG dispersions F9); baird2013 C30 (Crystallization tendency F16); kestur2013a R9 (Crystallization tendency F21); kestur2013b R3.e (Physical stability F21). 2014: Trasi and co-workers extended surface growth to 14 drugs trasi2014a R4, R5 (Crystallization tendency F21). 2017: Trasi and co-workers traced the generic tacrolimus to its route, and Saboo and Taylor showed the humidity of casting setting drug-rich domains trasi2017 C28 (Physical stability F33); saboo2017 R6 (Physical stability F14). 2021–2022: Bhujbal and co-workers showed residual water and drying time deciding the stability of lumefantrine powders, and Duong and co-workers coated particle surfaces to delay crystallization bhujbal2021 C43, C56 (Physical stability F34); duong2022c R16, R17, R18 (Crystallization tendency F21).

How it was shown

Baird and co-workers (2010) cooled class II and III molecules at 1 °C/min and reheated them at 20 or 2 and 10 °C/min, and in 2013 used a rapid heat/cool calorimeter on 30–50 µg crystals baird2010 P4 (Crystallization tendency F10); baird2013 P2, A1, R5, P4 (Crystallization tendency F10). They fitted critical cooling rates to undercooling measured at 5–20 °C/min and, in 2013, up to about 800–1100 °C/min baird2010 P2, P9, P6 (Crystallization tendency F11); baird2013 P3, R1 (Crystallization tendency F11). Rumondor and co-workers stored felodipine powders at six RH levels for up to 724 days and compared PVP and HPMCAS by an F-test at the last common time point rumondor2009b P2, P6 (Physical stability F16). Wegiel and co-workers stored curcumin dispersions at six conditions, from 25 °C/0% RH to 60 °C/98% RH, for up to 350 days, and resveratrol dispersions dry at 50, 90 and 120 °C and at 84% RH wegiel2014 P5, P6 (Physical stability F25); wegiel2013b P3, P4 (Physical stability F25). Kestur and co-workers stored a 3% PVP felodipine powder at 30 and 50 °C kestur2013b P4, R9 (Physical stability F25). Marsac and co-workers equilibrated PEO at each RH for several hours and for 5 months marsac2008b P7, X6 (PEG dispersions F12). Zhu and co-workers followed neat melts for about 500 h and comelts for hours zhu2012 P3, P5, P6 (PEG dispersions F8).

What the data show

On slow cooling four class II molecules crystallized and no class III molecule did, while within class II the heating rate needed ran from 12.5 to above 1910 °C/min baird2010 R7 (Crystallization tendency F10); baird2013 R9, C22 (Crystallization tendency F10). Of 19 fits of the critical cooling rate over 5–20 °C/min, seven had r^2 of 0.74–0.99 and twelve had r^2 < 0.55, while six compounds gave 0.89–0.99 over the wider range baird2010 R13, C18 (Crystallization tendency F11); baird2013 R7 (Crystallization tendency F11). In the χ protocols the same sample gave −6.16, −1.94 and −0.82 depending on the point read, and DSC, IR and PXRD used samples with different thermal histories marsac2009 R6 (Miscibility and drug-polymer interactions F7); rumondor2009a X4 (Miscibility and drug-polymer interactions F3).

In storage, felodipine dispersions at 0, 33 and 52% RH did not separate PVP from HPMCAS, while every PVP dispersion stored at 66% RH or above crystallized more than at lower RH; separation was seen after 3 days at 54% RH (pimozide–PVP, partial), after 1 week at 75% RH (felodipine–PVP) and within 1 h at 94% RH rumondor2009b R19, R9, R10 (Physical stability F16); rumondor2009c R10, X9, R9 (Physical stability F16). A tacrolimus generic crystallized to about 11% in 4 weeks at 40 °C/54% RH and about 84.5% at 40 °C/75% RH trasi2017 R14 (Physical stability F16). Growth of neat felodipine rose only about twofold from 13 to 80% RH and then reached 2.3–2.8 mm/min at 87–93% RH with a change of crystal habit that a 75% RH test would not show rumondor2010 R1, C8, C9 (Physical stability F9). For curcumin at 50 °C/0% RH E100 and HPMC stayed amorphous beyond 350 days while PVP crystallized by 6 days; dry at 50, 90 and 120 °C resveratrol–HPMC reached about 5, 20 and 100% crystallinity and then did not change; a felodipine–PVP powder crystallized about 8 times faster at 50 than at 30 °C and to about 85% instead of 45% wegiel2014 R13 (Physical stability F25); wegiel2013b R14, R13 (Physical stability F25); kestur2013b R9, R17 (Physical stability F25). The polymorph also depends on the stress: curcumin gave form I alone or with E100 or PVP but the metastable form II from HPMC, HPMCAS and CMCAB at 98% RH and 60 °C wegiel2014 R15, R17, C28 (Physical stability F20), and vemurafenib gave two or three of four polymorphs together at 120–240 °C lu2016 R10, C59 (Crystallization tendency F7).

Forming a glass and keeping it are different tests: flufenamic acid made glassy at 25 or 1025 °C/min still crystallized on reheating at 1900 °C/min baird2013 R10, C25 (Crystallization tendency F15). Storage crystallinity also costs dissolution: after 24 weeks, when the small particles were about 45% and the large ones about 17% crystalline, the dissolution peaks of the 3% PVP felodipine dispersions fell from 6.97 and 5.28 to about 2.5 and 3.4 µg/mL kestur2013b R20, R21, R8 (Physical stability F26).

PEO equilibrated for hours at 94% RH held about 10 wt% water and after 5 months about 37 wt%, whereas at 75–85% RH the two series differed little; acetaminophen lost crystallinity at 75% RH, though the cited deliquescence point lies above it marsac2008b R11, R22, R3 (PEG dispersions F12). In PEG, neat ibuprofen and fenofibrate had crystallized in only 10 and 17 of 50 samples after 3 weeks, so the tests stopped with 40 and 33 of 50 samples uncrystallized zhu2012 R2, R3, X5, C3 (PEG dispersions F8). The storage test that started the classes used glasses between two cover glasses, assumed to suppress surface crystallization, and felodipine form I grew about 6 times faster at the free surface at 18 °C baird2010 C45 (Crystallization tendency F21); kestur2013a R8, R9 (Crystallization tendency F21).

Figure

Near zero at 0–52% RH for about 493 days; about 81–89% within two weeks at 93% RH and a slower climb at 75% RH rumondor2009b R7 (Physical stability F16).

The curves part at 94% RH, about 10 against 37% marsac2008b R11 (PEG dispersions F12).

Why

The authors tie each limit to what the test changes: a class boundary is a fixed rate, so a drug whose own critical rate lies near it can move between classes with the protocol (synthesis) baird2010 P1; baird2013 R9 (Crystallization tendency F10). The extrapolation fails because crystallization temperatures scatter, "most likely" from the stochastic nature of nucleation and from heterogeneous nucleation in DSC pans baird2010 C18, C19; baird2013 C18 (Crystallization tendency F11). Separation at RH as low as 54 or 75% "highlights" a problem with accelerated conditions, because in susceptible systems the route of crystallization "may be altered at high RH" and need not represent lower RH rumondor2009c C52 (Physical stability F16). For resveratrol humidity promoted crystallization "to a much greater extent" than heat and "We do not fully understand" why, and the different rate and extent at 50 °C "highlights the difficulty" of predicting lower temperatures from accelerated studies wegiel2013b C39, C40; kestur2013b C37, C40 (Physical stability F25). Partially crystalline polymers "may" defeat Arrhenius extrapolation because PEO keeps taking up water and may slowly deliquesce marsac2008b C40, C30 (PEG dispersions F12). Surface growth "is generally faster" and extrapolation of high-temperature rates "may not be predictive" trasi2014a C4; kestur2013a C7, C18 (Crystallization tendency F21).

Conditions and limits

Values from two instruments are mixed in the critical-rate table baird2013 X9 (Crystallization tendency F10). The felodipine and curcumin tables give no sampling intervals or replicates, and crystallinities of about 5–10% are at the detection limit wegiel2014 X10; wegiel2013b X16, X15 (Physical stability F25). The humidity results are for NH-donor drugs with PVP at room temperature, and for HPMCAS dispersions high RH changed little rumondor2009b A3, C10 (Physical stability F16). The PEO isotherm method and temperature are not printed marsac2008b X6 (PEG dispersions F12).

Counterexamples and conflicts

The extreme classes held across protocols: no class III molecule crystallized on slow cooling and all eight class I-A molecules crystallized up to 800–1100 °C/min baird2010 R7; baird2013 R1, R3 (Crystallization tendency F10). A wider rate range improved the extrapolation for six compounds, though three table values are misprinted or not reproducible baird2013 R7, X2 (Crystallization tendency F11). High RH did separate polymers that dry storage could not rumondor2009b R9, C13 (Physical stability F16). A test can be too mild or too harsh to rank: the mildest condition separated no curcumin dispersion and neither did the harshest wegiel2014 R13, C23 (Physical stability F25).

Transfer check

Test the neat drug at the cooling rate of the intended process and treat a drug that changes behaviour with rate as borderline; answer the practical question with a quench test or a fast-scanning DSC run rather than an extrapolation from conventional DSC baird2010 X12, R8; trasi2012a C7 (Crystallization tendency F10); baird2013 C21 (Crystallization tendency F11). Compare χ only within one protocol and report onset, midpoint and offset marsac2009 R6 (Miscibility and drug-polymer interactions F7). Store at the intended condition and at two stress conditions that differ in kind, one hot and dry and one humid near room temperature; rank polymers only where the neat amorphous drug crystallizes but not every dispersion does; follow crystallinity to a plateau and name the polymorph wegiel2014 R13, C23 (Physical stability F25); wegiel2014 R15 (Physical stability F20). Run stability at the intended storage RH and at 75% RH, check the phase state as well as crystallinity at each pull, and do not extrapolate crystallization rates between a condition that separates the dispersion and one that does not rumondor2009b R9; rumondor2009c C52 (Physical stability F16). For a PEG or PEO dispersion follow the real dispersion by XRD and Raman for at least 24 h and the carrier's water uptake and peaks over weeks to months zhu2012 P5, P6 (PEG dispersions F8); marsac2008b R11 (PEG dispersions F12). Test open-surface films as well as sandwiched ones kestur2013a R9 (Crystallization tendency F21). Measure the dissolution of the dispersion after the stability pull, not only its crystallinity kestur2013b R21, R8 (Physical stability F26).

History

2008: Marsac and co-workers showed that PEO takes up far more water over months than in an equilibration of hours and warned that partially crystalline polymers may defeat Arrhenius extrapolation marsac2008b R11, C40 (PEG dispersions F12). 2009: Rumondor and co-workers showed that only high RH separated PVP from HPMCAS and that high RH also changes the route rumondor2009b R19 (Physical stability F16); rumondor2009c C52 (Physical stability F16). 2010: Baird and co-workers defined the classes by one protocol and showed their rate dependence, and tested the extrapolation of critical rates baird2010 R7, R13 (Crystallization tendency F10, Crystallization tendency F11). 2013: Baird and co-workers widened the rate range; Wegiel and co-workers showed that heat and humidity give different orders, and Kestur and co-workers that 50 and 30 °C give different rates and extents baird2013 R7, R9 (Crystallization tendency F11, Crystallization tendency F10); wegiel2013b R12, C39 (Physical stability F25); kestur2013b R9 (Physical stability F25). 2014: Wegiel and co-workers showed that only three of six curcumin storage conditions (50 °C/0% RH, 25 °C/84% RH and 40 °C/75% RH) ranked the polymers wegiel2014 R13, C23 (Physical stability F25). 2017: Trasi and co-workers showed 40 °C/54% RH separating from 40 °C/75% RH for a marketed product trasi2017 R14 (Physical stability F16).