Do itraconazole and HPMC really mix? A single glass transition against nanoscale chemical maps

Background and motivation

An amorphous drug dissolves to a higher apparent concentration than its crystal, but the advantage lasts only if crystallization from the solid and from the supersaturated solution is held back. In an amorphous solid dispersion (ASD) the drug is dispersed in a polymer that slows crystallization and, if hydrophilic, helps wetting and dissolution (Hancock 2000, ref 1; Janssens 2009, ref 6). It is "commonly thought" that drug and polymer must be miscible at the molecular level for the ASD to perform well, and a phase-separated dispersion "would be expected" to crystallize faster in storage and in dissolution (refs 3, 7, 8). C1·C2

Testing miscibility is the practical difficulty. The authors list a long set of methods already used: solid-state NMR, local thermal analysis, X-ray diffraction, mid-infrared spectroscopy, differential scanning calorimetry (DSC), atomic force microscopy (AFM), scanning and transmission electron microscopy and confocal Raman imaging (refs 9–33). Each has a blind spot. Bulk vibrational spectroscopy says nothing about the size of separated domains; spectroscopic imaging cannot resolve domains smaller than the diffraction limit of light; AFM, SEM and TEM resolve small features but cannot say what they are made of; and DSC, the most widely used method, is limited by being non-isothermal: the sample changes while it is being heated (refs 34–36). C3

Two newer tools address these gaps. AFM combined with nanoscale infrared spectroscopy (AFM-nanoIR) records IR spectra through an AFM tip and so gives chemical information at sub-micrometre resolution; the Taylor group had used it shortly before to judge the miscibility of polymer blends and of a drug–polymer blend (Van Eerdenbrugh 2012, refs 40, 41). Fluorescence spectroscopy with environment-sensitive probes is common for biological and polymeric systems and had recently been applied to amorphous pharmaceutical systems (refs 48–50, including Purohit 2015, ref 50). C4·C5

The test system is itraconazole (ITZ) with hydroxypropyl methylcellulose (HPMC). The pair is used in the marketed ASDs Sporanox (spray dried) and Onmel (hot melt extruded), and its miscibility at different drug loadings had been debated (Janssens 2008, ref 58; Six 2003, ref 59; Six 2005, ref 60). C6·R16 The paper sets out to evaluate AFM-nanoIR further, alongside an emerging approach based on fluorescence spectroscopy and imaging with two probes, pyrene and prodan, and to use these methods to see how polymer amount and temperature affect the miscibility and domain structure of ITZ–HPMC. A1·A2·A4

Background concepts

Environment-sensitive fluorescent probes. A small amount of a fluorescent molecule is added to the film, and its emission reports how polar its immediate surroundings are. Pyrene has a five-peak emission spectrum; the ratio of the first to the third peak, the I/III ratio, rises as the surroundings become more polar. Prodan has a single broad emission band whose maximum shifts to longer wavelength (a red shift) in more polar surroundings. R17 In this paper the hydrophilic polymer HPMC and the hydrophobic drug ITZ give clearly different readings, so each probe acts as a local polarity meter: if a blend reads like ITZ, the probe sits in an ITZ-like environment. R1·C7 The probes are hydrophobic, so the authors expect them to prefer a drug-rich phase if one exists. C19

Excimers and exciplexes. Pyrene can also give a broad, featureless band at longer wavelength, near 470 nm, when an excited pyrene molecule forms a complex with a neighbour that is within about 5–10 Å. With a second pyrene the complex is an excimer; with a different molecule it is an exciplex. R18 The appearance of such a band therefore signals that pyrene has close partners.

AFM-nanoIR. The film sits on a ZnSe prism and is illuminated from below by a tunable IR laser. Where the sample absorbs, it heats and expands very slightly; the AFM tip resting on the surface feels the expansion as a ring-down of the cantilever, whose amplitude, plotted against wavenumber, gives a local IR spectrum. Because the tip, not the light, defines where the signal comes from, the spatial resolution is well below the optical diffraction limit. C4·P9 Combined with the topography from the same tip, a raised domain can be identified chemically.

Nanoscale thermal analysis (nanoTA). A heatable AFM probe is placed on one spot and its temperature is ramped. While the material below is glassy, the probe rises slowly as the material expands; when it softens, the probe sinks in and the deflection signal turns down or levels off. The authors take this turning point as the local softening point, which for an amorphous phase corresponds to its glass transition temperature (Tg). C33·P10

The single-Tg criterion. A miscible amorphous blend has one Tg, between those of its components; two separate phases large enough to have their own bulk properties show two. A single calorimetric Tg is "generally considered" an indicator of miscibility. C22 The weakness, which this paper illustrates, is that DSC heats the sample through and above its Tg, so a blend that is separated at room temperature can mix during the measurement. C39

Flory–Huggins free energy of mixing. The thermodynamics of drug–polymer mixing is usually discussed with the Flory–Huggins equation, which the paper prints as Equation 1 E1:

Here $\Delta G_{\mathrm{m}\mathrm{i}\mathrm{x}}$ is the free energy of mixing, $R$ the gas constant, $T$ the absolute temperature, $n$ the number of moles and $\varphi$ the volume fraction of drug ($D$) or polymer ($P$), and $\chi$ the Flory–Huggins interaction parameter. In plain terms: the first two terms are the entropy of mixing, which always favours mixing; the third term carries the balance of drug–polymer contacts against drug–drug and polymer–polymer contacts. A negative $\chi$ means the drug and polymer prefer each other; a positive $\chi$ means each prefers itself, and the blend mixes only if the entropy gain is large enough. E1·C41 The paper uses the equation only in words; no $\chi$ was measured.

Upper critical solution temperature (UCST). In a UCST phase diagram (the paper's Figure 9) the components separate into two phases below a critical temperature $T_c$, inside a dome bounded by the binodal and, within it, the spinodal; above $T_c$ they are miscible at every composition. C47 For a glassy blend this offers one way in which heating could produce mixing that survives cooling: the blend mixes while hot and is then frozen in that state. C47

Methods in detail

The design is simple: one drug–polymer pair, five compositions, one casting solvent, and four film methods that see local composition set against DSC, each applied before and after heating. The flow below summarises it.

Samples

ITZ (Attix Pharmaceuticals) and HPMC grade 606 (Shin-Etsu) were dissolved at 50:50, 40:60, 30:70, 20:80 and 10:90 w/w in a 50:50 v/v mixture of methanol (MeOH) and dichloromethane (DCM), at 20 mg/mL total solids. M1·M2·P1 One set of solutions stayed probe-free for AFM work; a second set was split, and pyrene or prodan was added from a MeOH:DCM stock to 0.01% w/v (100 µg/mL). P1·M3·M4 A 50:50 solution was also made in 1:1 v/v MeOH:tetrahydrofuran (THF) to see whether the solvent matters. P1·A3

For fluorescence, 100 µl of solution was spin coated on a quartz coverslip at 3000 rpm for 30 s and dried under vacuum for 24 h; films of ITZ alone and HPMC alone served as references. P2 For AFM, probe-free solutions were spin coated on ZnSe prisms and vacuum dried. P9 For DSC, powders were made from the same solutions by rotary evaporation (water bath at 45 °C) and dried for at least 24 h under vacuum. P7 The DSC samples are therefore not the films that the other methods examined, a point that matters later. X1

Fluorescence spectroscopy and imaging

Emission spectra of the probe in each film were recorded on a Shimadzu RF-5301-PC spectrofluorophotometer, and films were imaged on an Olympus BX-51 fluorescence microscope with excitation at 330–380 nm and emission collected from 420 nm. P3·P4 Measurements were repeated after heating the films for 2 min at 165 °C, for pyrene only, because prodan is not stable at high temperature. P5·P3 The standard deviation was below 0.02 for the I/III ratio and below 2 nm for the prodan maximum; the number of films and the excitation wavelength of the spectrometer are not stated. R1·X4

AFM-nanoIR and nanoTA

To pick a marker band, bulk transmission IR spectra of amorphous ITZ and HPMC films on KRS-5 were recorded (64 scans, 500–4000 cm⁻¹). The ITZ carbonyl band at 1700 cm⁻¹ is strong and HPMC has no band there (Figure 5), so the authors use it to tell ITZ-rich from ITZ-lean regions. P6·R9·C27 The AFM-nanoIR instrument (Anasys) worked in contact mode with an IR laser at 1 mW; spectra from 1600 to 1800 cm⁻¹ were co-added over 256 scans at 4 cm⁻¹ resolution, with a tip scan rate of 0.06 Hz, 256 × 256 points and a contact frequency of 15 KHz. P9 Spectra were taken at marked points on raised domains and on the surrounding matrix, before heating (Figure 6) and after heating at 165 °C (Figure 7). R11·R12

For nanoTA, the probe temperature was calibrated with three crystalline polymers melting at 55 °C (polycaprolactone), 116 °C (polyethylene) and 235 °C (polyethylene terephthalate). On a 40:60 film the probe was heated at 2 °C/s on a discrete domain and on the continuous phase, and deflection was recorded against temperature (Figure 8). P10·M8

Differential scanning calorimetry

About 5 mg of powder in a Tzero aluminum pan with a pin-holed lid was heated at 2 °C/min to 170 °C with a modulation of ±1 °C every 60 s, held at 170 °C for 2 min, quench-cooled to 0 °C and heated again to 170 °C with modulation; indium and tin calibrated the instrument. P8·M9 The reversing heat flow (Figure 4) gave onset, midpoint and offset Tg for both runs (Table 2). R7·R8 The midpoint Tg values of pure ITZ and HPMC were 60 °C and 140 °C; the paper reports them as "data not shown". M1·M2

Experimental analysis

The probes report an ITZ-like environment in every film

Table 1 is the first piece of evidence. In pure HPMC pyrene gave I/III = 1.54 and prodan peaked at 452.9 nm; in amorphous ITZ the values were 1.39 and 439.0 nm. In all five blends before heating, pyrene read 1.39–1.42 and prodan 438.7–443.7 nm, close to ITZ. R1 The authors conclude that the probes are "preferentially present in an ITZ rich phase in all the films" ("This suggests"), which contradicts the single Tg from DSC. C8·C9

Table A. Probe readouts in ITZ–HPMC films (from Table 1; SD below 2 nm for prodan and below 0.02 for I/III). R1

Note by the AI writer (not discussed in the paper): "very close to ITZ" fits the 30:70 to 50:50 films best. At 10:90 the prodan maximum of 443.7 nm lies about a third of the way from ITZ toward HPMC, more than the stated standard deviation, and pyrene at 20:80 and 10:90 (1.42) lies about a fifth of the way. The low-loading films therefore register a slightly more polar environment than pure ITZ, consistent with some ITZ mixed into the matrix or some probe in it. X13·R1.f·R1.g

Pyrene adds a second signal. A broad band near 470 nm appeared in pure ITZ and in every blend film (Figure 2a shows the 10:90 film), but not in pure HPMC carrying the same amount of pyrene. R3 Because pyrene alone in HPMC gives no excimer, the authors exclude crowding of the probe and assign the band to an exciplex between excited pyrene and ITZ, concluding that the probe sits in drug-rich domains ("These observations indicate"). C10·C11 They argue that in a single miscible phase the distances "would most likely be greater than" about 10 Å and no exciplex would form, a supposition based on the HPMC control. C12 The band position is given as about 470 nm in the text and 475 nm in the Figure 2 caption, and its maximum in Figure 2a lies near 480 nm. X5 As cast, the films were translucent to opaque, whereas molecularly miscible films are typically transparent (Feng 1995, ref 71); the authors take this as support for phase separation. R5·R19·C13 The MeOH:THF 50:50 film gave I/III = 1.39, the same as the MeOH:DCM film, so phase separation also occurred with the other co-solvent ("suggesting"). R2·C18

Images and nanoIR spectra locate the drug

Fluorescence microscopy of the pyrene-containing 10:90 film showed discrete bright spots of about 1 µm on a darker background (Figure 3a); the non-homogeneous nature is "clearly evident" even at this low loading, and the authors report similar images at all ratios. R6·C20

AFM topography resolves the same structure much more finely. Every film, from 10:90 to 50:50, shows round raised domains in a flatter continuous matrix (Figure 6, left). The topography alone cannot say which phase holds the drug, but the nanoIR spectra can: spectra taken on domains carry a strong carbonyl band at about 1700–1708 cm⁻¹, while those on the matrix carry a weak band or none. R10·R11·C28·C29 The authors call this "excellent agreement" with the fluorescence results and a confirmation that the system "is not uniformly mixed, not even at high polymer loadings". C30 They size the domains at 500–1000 nm and note that ITZ, the lower-viscosity component, forms the discrete phase, as seen before in other blends (ref 40). C37

Table B. AFM-nanoIR before heating, read from Figure 6 (approximate). Relative band heights compare red (matrix) with blue (domain) spectra in the same panel. R10·R11

Two details sharpen the picture. First, at low loading the matrix is not drug-free: at 10:90 and 30:70 the matrix spectra still carry a clear, if weaker, carbonyl band, while at 40:60 there is none. R11 Second, domain size is not fixed: the table-of-contents graphic, the only nanoIR chemical map in the paper, shows ITZ-rich domains of about 1.5–2.5 µm in a 30:70 film, larger than the stated 500–1000 nm. R14·X8

nanoTA: two softening points

On the 40:60 film a probe placed on a domain softened at 71 °C (plot marker 71.26 °C), close to the 60 °C Tg of amorphous ITZ, while probes on the continuous phase softened at 108 °C (marker 108.94 °C), which the authors read as a polymer-rich phase ("indicates"). R13·C34 The domain trace was recorded once and two traces were taken on the matrix. R13 The single DSC transition of the 40:60 powder (onset 78, midpoint 87, offset 95 °C) falls between these two values; nanoTA ramps at 2 °C/s and DSC at 2 °C/min, so the numbers are not directly comparable. R7.b·X2

DSC: one broad Tg, narrower on reheating

Table C. Glass transition of rotary-evaporated powders in the first and second DSC heating (from Table 2). The width column (offset minus onset) was computed by the AI writer. R7

Every powder shows a single step, at temperatures that rise with HPMC content between the Tg of ITZ (60 °C) and HPMC (140 °C). R7·M1·M2 In the results section the authors therefore write that "DSC results indicate that a miscible system is formed" at all ratios, while noting that the transition is broad. C23·C24 After the first scan, a 2 min hold at 170 °C and a quench, the second heating still shows one Tg, now narrower (widths 6–16 °C against 11–25 °C) and lower (midpoints 13–26 °C below the first run). R7·C25 The authors read the narrowing as less compositional variation, which "helps support" heat-induced miscibility. C26 Figure 4a also shows, in the 50:50 first scan only, a broad upward hump near 105 °C and a small dip near 165 °C after the Tg step; the text does not discuss them. R8·X11

Heating removes the domains

After 2 min at 165 °C every method changed in the same direction. The films became transparent ("may indicate heat induced mixing"), pyrene I/III rose to 1.50–1.53, close to the value of unheated HPMC (1.54), and the exciplex band disappeared from the 10:90 spectrum (Figure 2b). R5·R1·R4·C14·C15 The authors read the loss of the exciplex as "the increased interaction between ITZ and HPMC (heat induced miscibility)" disrupting the pyrene–ITZ contact, and the three changes together as an increase in the extent of miscibility. C16·C17 Fluorescence images became more uniform, with faint residual spots (Figure 3b). R6·C21

The AFM picture changed too (Figure 7). Round domains were gone, and every one of eight spectra per film, taken along a line across the scan, carried the carbonyl band near 1700 cm⁻¹; the authors describe it as present "throughout the sample with almost equal intensity". R12·C31 Read from the figure, the band heights are close at 10:90 to 30:70 (about 0.7–1 of the highest), but span about 0.6–1 at 40:60 and about 0.3–1 at 50:50, where the heated topography still shows raised and depressed regions of ±100 nm. R12·X9 Across the paper the authors conclude that heating promotes mixing, "with a concurrent loss of structural and chemical heterogeneity". C38

Discussion

Why DSC and the film methods disagree

The authors' summary is direct: the results "clearly indicate" that ITZ and HPMC are phase separated over 10–50% drug loading when prepared by solvent removal and examined by fluorescence or AFM-IR, while DSC "was not able to show clear evidence of phase separation". C36 Their explanation for the single Tg is that the sample "most likely undergoes mixing during the heating encountered during the DSC scan", giving one, broadened transition; DSC had given ambiguous miscibility results before, including for HPMC–ITZ (refs 36, 59). C39·C40 The heated films support the premise that heating mixes the blend, and the narrower second-scan transition fits the same picture. C38·C26

Note by the AI writer (not discussed in the paper): the explanation is plausible but the paper does not show it directly. The DSC powders were made by rotary evaporation, not spin coating, and they were never imaged or probed; the films were never run by DSC. It therefore remains open whether the powders were phase separated before the scan in the same way as the films. X1 The nanoTA result is the closest link: in a 40:60 film, domain and matrix soften at 71 and 108 °C, while the 40:60 powder gives a single transition from 78 to 95 °C, between the two. X2·R13·R7.b

Kinetic or thermodynamic origin

Why does a solvent-cast ITZ–HPMC film separate in the first place? The authors set out two routes (Figure 10; the text cites it as Figure 9) and say plainly that "it is very hard to determine" which applies. C44·C48·X6 The flow below follows their argument.

The kinetic route is solvent-induced immiscibility. In a mixed solvent whose components differ in volatility and solvency, separation "could be even more pronounced". ITZ is freely soluble in DCM and poorly soluble in MeOH, while HPMC dissolves in the mixture; as the more volatile DCM leaves first, ITZ "might reasonably be expected to phase separate as it exceeds its miscibility limit with MeOH". C45·M1·M2 Once dry, both phases are glassy at room temperature (Tg 60 and 140 °C), so even if mixing were favoured thermodynamically, "a phase separated system would be expected to persist". C46

The thermodynamic route starts from chemistry. HPMC's substituted hydroxyls leave few donors for ITZ's carbonyl acceptor, and ITZ has no donor groups, so "any miscibility between the two components is likely to be entropically driven" (Rumondor 2011, ref 20). C42 In Flory–Huggins terms that is a positive $\chi$, balanced only by the entropy terms of Equation 1. C41·E1 ITZ and HPMC may then have limited miscibility at room temperature that grows with temperature, as in a UCST diagram; near the drug's melting point mixing is favoured, and on cooling the glass "will very likely trap" the miscible state. C47

The paper offers one observation that bears on the choice: a 50:50 film cast from MeOH:THF also gave an ITZ-like pyrene reading (I/III 1.39). R2·C18 Note by the AI writer: this shows that the separation is not specific to DCM, but it does not decide between the two routes, because methanol is present in both mixtures and the paper gives no information on how the solvent composition changes during drying. No $\chi$, temperature–composition diagram or solvent composition during drying was measured. C45·C47

What it suggests for processing and products

From the heating results the authors draw a processing implication: spray drying with a similar solvent system "might lead to some degree of phase separation" in ITZ–HPMC, while hot melt extrusion "would be more likely" to give a homogeneous dispersion, provided mixing efficiency in the extruder is not limiting (refs 59, 83). C49 Both routes exist on the market, as Sporanox (spray dried) and Onmel (hot melt extruded), and their microstructure "is not currently known". C50·R16 A partially miscible ASD with drug-rich domains "would be expected" to crystallize more readily, because less polymer surrounds the drug, and its dissolution "may" suffer, because the polymer-rich phase dissolves first and leaves a poorly wettable drug-rich phase. C51 The authors close by calling AFM-nanoIR and fluorescence "useful additions to the toolbox" that "offer promise for investigating process-induced variations in ASD structure". C52

Note by the AI writer (not discussed in the paper): for a formulator working with solvent processes, the solvent pair is part of the formulation. A co-solvent that evaporates last and dissolves the drug poorly can drive drug-rich domains during drying, even at 10% drug in HPMC. A first DSC scan that narrows and shifts on a second scan after a hold above Tg is a cheap hint that the as-made material was heterogeneous, but a sub-micron method that sees composition is needed to show it. No crystallization, stability or dissolution data were collected here, so the practical cost of these domains remains to be measured. C51·C35

Limitations

Limits the authors state. They cannot say whether the phase separation is kinetic or thermodynamic ("Unfortunately, it is very hard to determine"), and they do not know the microstructure of the marketed products. C48·C50 Their explanations of the single DSC Tg ("most likely") and of the processing difference ("might", "would be more likely") are given as reasoning, not as tested results. C39·C49

The AI writer's own judgement (1): what the evidence does and does not show. The qualitative conclusion that solvent-cast ITZ–HPMC films are phase separated is well supported: four independent film methods agree, and the nanoIR assignment rests on a clean marker band that HPMC lacks. R1·R6·R9·R11·R13 The evidence is, however, built on single films and a few spots per film. nanoTA was run on one domain of one 40:60 film; the number of films, replicate spectra and the heating time of the AFM films are not stated; and domain size and area fraction were not quantified. X2·X12·X8 The exciplex argument is indirect: the HPMC-only control excludes crowding of pyrene, but not an ITZ–pyrene exciplex inside a blend that holds 10–50 wt% ITZ, where ITZ molecules would surround many pyrene molecules even if fully mixed. X3 The after-heating I/III values have no same-condition reference, because pure ITZ and HPMC films were not remeasured after heating and pyrene's stability at 165 °C is not reported. X4

The AI writer's own judgement (2): comparisons and loose ends. The central claim that DSC misses the separation compares rotary-evaporated powders with spin-coated films. X1 The stated agreement with Six et al. (ref 59) compares this paper's solvent-cast films, separated from 10% drug, with melt-extruded material reported to separate at 40%; the authors themselves argue that melt processing mixes the blend more. X10·C43 "Almost equal intensity" after heating fits 10:90 to 30:70 better than 50:50, where band heights still vary about three-fold. X9 The second-scan midpoints are 13–26 °C lower than the first-scan ones, a shift the paper reports but does not explain, and the 50:50 first scan contains an undiscussed hump and dip. X14·X11 There are also small textual slips: the band position is given as 470 and 475 nm, the mechanism schematic is cited as Figure 9 instead of Figure 10, Figure 6c labels its axis "Wavelength (cm-1)", and the pyrene structure in Figure 1b is not drawn as fully aromatic. X5·X6·X7 None of these changes the main conclusion.

Conclusion

This short paper makes one point clearly. A single, broad DSC glass transition did not mean that itraconazole and HPMC were mixed: in films cast from methanol–dichloromethane, environment-sensitive probes, fluorescence imaging, AFM-nanoIR and nanoTA all found itraconazole-rich domains of roughly a micron in a drug-lean, HPMC-rich matrix at every loading from 10 to 50% drug. C36·C37·R11·R13 Heating to 165 °C largely removed the domains, which the authors take as the reason DSC sees one Tg: the sample mixes while it is measured. C38·C39 Whether the room-temperature separation comes from solvent evaporation or from limited thermodynamic miscibility was left open, and the implication for spray drying versus hot melt extrusion was not tested. C48·C49