What sets the release rate of itraconazole–copovidone dispersions: drug loading, manufacturing route and spray solvent

Background and motivation

Spray drying (SD) and hot melt extrusion (HME, products labelled HM in the paper) are the two processes used to make amorphous solid dispersions (ASDs) commercially (Démuth et al., 2016). Each has its advantages for a given system, and when both are feasible the choice should rest on product performance (Vasconcelos et al., 2016; Patterson et al., 2007). The introduction argues that faster dissolution and supersaturation are not the only goals: process conditions such as the spray-drying solvent matter (Paudel and Van den Mooter, 2012), and so do test methods that can discriminate between products. This becomes pressing as generic ASD products appear, made by other manufacturers and possibly by other routes, and their equivalence to the reference has to be shown (Trasi et al., 2017a, 2017b).

A second strand is homogeneity. Drug and polymer should be mixed at the molecular level, but they can separate during processing and storage, which can lead to crystallization or poorer dissolution (Singh et al., 2016; Zhang et al., 2004); water in particular can make a hydrophobic drug demix from the polymer (Marsac et al., 2010; Rumondor and Taylor, 2010). Drug-rich and polymer-rich domains can be detected with AFM coupled with infrared spectroscopy, environment-sensitive fluorescent probes, solid-state NMR, transmission electron microscopy (TEM) and surface analysis. The authors note, however, that the effect of such phase separation on product performance is poorly understood and "may only be relevant" when it changes dissolution or crystallization. Very few studies had compared spray-dried and melt-extruded dispersions of the same composition with dissolution testing (Agrawal et al., 2013; Guns et al., 2011; Mahmah et al., 2014).

The paper takes itraconazole (ITZ) and copovidone (PVPVA 64, Kollidon VA 64) as the model pair. PVPVA extrudes easily because of its low glass transition temperature (109 °C, as stated in the paper) and spray dries well because it dissolves in many polar and non-polar solvents. ITZ is extremely poorly water soluble and is sold both as solvent-evaporation products (Sporanox, Lozanoc) and as an extruded product (Onmel). The study made dispersions at four drug:polymer ratios by both routes, measured how fast both drug and polymer dissolve, and then used surface analysis to understand the differences. Partway through, the results raised a specific hypothesis, that the spatial distribution of drug and polymer differs between batches, and the authors tested it by spray drying one composition from several solvents.

Background concepts

Intrinsic and surface-area-normalized dissolution. The intrinsic dissolution rate (IDR) is the mass of a pure compound that dissolves per unit time from a compact whose exposed area is fixed, usually in mg or µg per minute per cm². Fixing the area removes the effect of particle size and powder surface area, so that two solid forms of the same compound can be compared directly. The authors apply the same set-up to dispersions and call the result a surface-area-normalized dissolution rate. Here the compact is a 25 ± 5 mg disk with a face of about 0.50 cm², held in a die so that only that face touches the medium (Figure 2). The authors argue that this removes most of the particle-size effect and reduces porosity to a level relevant to manufacturing, so that formulations can be compared fairly (Section 5).

Rates and solubility. Initial dissolution rate is related to solubility (Hamlin et al., 1965), and for two polymorphs the ratio of their intrinsic dissolution rates is taken to reflect the ratio of their solubilities (Wadke and Reier, 1972). The authors apply this reasoning to amorphous and crystalline ITZ: the ratio of their IDRs "should be the same" as the ratio of their solubilities. The paper does not measure the solubilities, so this is used as an interpretation of the rate ratio, not as a test of it.

Drug-controlled and polymer-controlled release. Simonelli et al. (1969) measured both drug and polymer release from compacts of sulfathiazole–PVP dispersions and found three regimes as the drug loading falls: release controlled by crystalline drug at very high loading (the surface converts to crystals), by amorphous drug at intermediate loading, and by the polymer at very low loading. In the drug-controlled regime the drug dissolves from a drug-rich surface at roughly the rate of the pure amorphous drug, whatever the polymer does. In the polymer-controlled regime the drug is carried out as the polymer dissolves, so drug and polymer leave in about the ratio in which they are present, and drug release can be much faster than that of the amorphous drug. Chen et al. (2016) found, for ketoconazole dispersions at 20% drug loading, that the normalized drug release tracked the normalized polymer release, and that the presence of drug slowed the polymer. Both studies frame how this paper reads its data.

Loading-normalized rates. To compare drug and polymer on equal terms, the paper divides each release rate by the amount of that component in the dispersion ("normalized (to account for differences in drug and polymer content)", Figure 14). The operation is not written out; dividing the Table 2 drug rates by the drug weight fraction reproduces the drug bars of Figure 14 (for example 113 / 0.2 = 565 µg/min/cm² against about 555 in the figure). Equal normalized rates mean that drug and polymer leave the compact together.

Probing a dispersion's structure. Three techniques look at different depths. Pyrene, a fluorescent probe, emits five peaks between 350 and 420 nm; the ratio of peak I (373 nm) to peak III (383 nm) is higher in a polar, hydrophilic environment than in a hydrophobic one (Goddard et al., 1985), and has been used to detect phase separation in spray-dried ITZ–HPMC (Purohit and Taylor, 2015). TEM images the inside of thin particles; because each ITZ molecule carries two chlorine atoms, drug-rich regions should scatter electrons more and look darker (Li et al., 2017; Mugheirbi et al., 2017). X-ray photoelectron spectroscopy (XPS) measures the elemental composition of only the outer ~10 nm of a particle. Since PVPVA contains no chlorine, the chlorine (Cl 2p) signal relative to carbon (C 1s) indexes the amount of drug at the surface; the nitrogen (N 1s) and oxygen (O 1s) peaks, fitted with component shapes for drug and polymer, give drug and polymer percentages.

The Peclet number in spray drying. When a droplet dries, its surface recedes while dissolved components diffuse back toward the centre. The paper (Eq. 1) gives the Peclet number of a component $i$ as

where $k$ is the solvent evaporation rate and $D_i$ is the diffusivity of component $i$ in the liquid. In plain terms: the number compares how fast the solvent leaves with how fast a solute can move away from the shrinking surface. Simulations by Vehring et al. (2007) show that a solute with a large Peclet number accumulates at the surface. The paper uses the relation qualitatively and computes no value.

Zeotropic and azeotropic solvent mixtures. A zeotropic mixture changes composition as it evaporates, because the more volatile solvent leaves first; an azeotrope evaporates at constant composition. Table 4 lists DCM–methanol (azeotrope boiling at 38.1 °C, weight fraction of DCM printed as 92.7) and DCM–water (38.8 °C, 99.6) as azeotropic, DCM–n-hexane as forming no azeotrope, and gives no data for DCM–cyclohexane. In a zeotropic DCM–hexane droplet, DCM (boiling point 39.6 °C) leaves first and the droplet becomes richer in hexane, in which ITZ is much less soluble than in DCM (440 against 28,400 µg/mL at 25 °C) and PVPVA is insoluble (Table 3).

Methods in detail

The study has two parts. The first compares drug loadings and manufacturing routes with simultaneous drug and polymer release. The second holds the composition at 20% ITZ, the loading at which the routes differed, and varies the spray-drying solvent; the resulting particles are then examined by fluorescence, electron microscopy and XPS.

Making the samples

Pure forms. As-received ITZ was identified by PXRD as form I (CSD refcode TEHZIP), not form II. Amorphous ITZ was made by melting the drug on a hot plate at about 175 °C and quench-cooling the melt in liquid nitrogen, then crushing it; conversion back to the crystal is reported to be slow (Zhang et al., 2016). PVPVA was used as received ("PVPVA-Raw").

Spray drying. Drug and polymer were dissolved together to about 20% w/v solids, stirred for at least an hour and spray dried in a Büchi B-290 mini spray dryer with an inert loop, under nitrogen. The inlet was set at 70–75 °C, giving an outlet temperature of 48 ± 2 °C; atomization setting 40 mm, aspirator 70% and feed rate 5–7%. The powder was dried under vacuum for 24–48 h and kept in a desiccator over calcium sulfate. The standard solvent was 50:50 DCM–methanol, chosen because DCM dissolves both components well (28,400 µg/mL ITZ, more than 1000 µg/mL PVPVA, Table 3) but has a low permitted daily exposure (6 mg/day), DCM is commonly mixed with an alcohol, and methanol dissolves ITZ best among the alcohols (745 µg/mL). Dispersions were made at ITZ:PVPVA ratios of 1:1, 1:3, 1:4 and 1:9 by weight (50, 25, 20 and 10% ITZ; codes 50-50, 25-75, 20-80, 10-90; Table 1).

Hot melt extrusion. Physical mixtures of crystalline ITZ and PVPVA (3.5–4.0 g) were fed into a 2 mL twin-screw extruder (Xplore Pharma Melt Extruder PME2) whose six heating zones were held at 175 °C, above the reported melting point of ITZ (166 °C) and the softening of PVPVA (around 130 °C). The melt was recirculated for 3–4 min and discharged through a 3 mm orifice. The extrudates were milled with a pestle and mortar or in a Wig-L-Bug grinder and stored over calcium sulfate.

The solvent series. To test whether the spatial distribution of drug and polymer explains the release differences, the 20% composition was spray dried from solvents chosen to change the solubility of one component or the evaporation behaviour of the mixture. For the hexane batches, n-hexane or cyclohexane (non-solvents for PVPVA) was added to the DCM solution until the polymer began to precipitate (turbidity), and then enough DCM to redissolve it, giving about 55:45 (v/v) DCM–n-hexane and 40:60 (v/v) DCM–cyclohexane. For the water batch, water (an anti-solvent for ITZ) was added to 50:50 DCM–methanol up to 50:50:20 (v/v/v), "just enough water to keep the drug dissolved" but close to its precipitation point. One batch used DCM alone. The DCM–n-hexane batch was also dried at an outlet temperature of 70 ± 1 °C, slightly above the boiling point of n-hexane. Solids were 15–20% w/v. Table 1 codes these batches 20-80SD (the 50:50 DCM–methanol reference), 20-80SD water18, 20-80SD-DCM, 20-80SD Cyc-DCM, 20-80SD n-Hex-DCM and 20-80SD n-Hex-DCM-HH ("high heat"). The paper does not explain the "18" in "water18"; the composition it gives is 50:50:20, which is one part water to five parts DCM–methanol.Card check: Table 6 labels the n-hexane batches "5:4 n-hexane-CH2Cl2", which in the printed order would be a hexane-rich mixture, whereas the methods give about 55:45 DCM:n-hexane; the methods value is used here (card note X2).

The compact and the dissolution measurement

Each sample (25 ± 5 mg) was pre-compressed at 500 psi in an 8 mm circular die, giving a face of about 0.50 cm². Ethyl cellulose powder with 1% magnesium stearate (30–35 mg) was added behind it and the whole was compressed at 1000 psi, which the authors state corresponds to about 2900 kg/cm² on the sample, within the range used in tablet manufacture; the pressure was released after 60 s. A silicone septum closed the back of the die, so only the front face of the sample met the medium (Figure 2).

Dissolution was run in an Inform instrument (Sirius Analytical) in 70 mL of 0.1 N HCl at 37 °C in a flat-bottomed glass vial, stirred by a paddle at 300 rpm. A diode-array detector recorded the full spectrum from 187 to 729 nm about every 30 s, with baseline correction against 500–600 nm, where neither component absorbs. ITZ was quantified at 253 nm (linear calibration 0.4–45.0 µg/mL, R² above 0.999). PVPVA was quantified at 220 nm: because the ratio of ITZ absorbance at 253 and 220 nm was fairly constant (about 0.68), the ITZ contribution at 220 nm could be subtracted and the residual absorbance converted to PVPVA concentration with a second-degree polynomial calibration (2–140 µg/mL, R² = 1.000). Every sample was run in triplicate. Rates were taken from the slope of cumulative release per cm² over 0–20 min (Table 2), and for pure PVPVA over the first 4.5 min.

PXRD of a melt-quenched compact after the test showed no crystallinity, and neither did the slowly dissolving hexane-solvent compacts at the end of their runs ("these samples", Section 4.3), so the rates refer to amorphous material.

Structure and surface analyses

Pyrene was incorporated at 125:1 (w/w) ITZ:pyrene and the solid samples measured on a Shimadzu RF-5301PC fluorimeter (excitation 332 nm, emission 337–550 nm). TEM used an FEI Tecnai G2 20 at 200 kV, with powder touched onto carbon-coated copper grids. SEM used an FEI NovaNano instrument on platinum-coated powders, including fractured particles and a compact pressed at 1000 psi and extracted from the die. XPS used a Kratos Axis Ultra DLD with monochromatic Al Kα radiation; the C–C component of C 1s was set to 284.8 eV, a Shirley background was subtracted, and concentrations were computed with Scofield sensitivity factors assuming a homogeneous mixture within the ~10 nm information depth. Drug and polymer percentages were obtained by fitting the N 1s and O 1s regions, and the Cl 2p spectra were resolved into two spin-split pairs at about 200 and 197 eV (n = 3; details in SI Figure S1, not available). Only spray-dried samples were examined by these techniques; the extruded dispersions were characterized only by PXRD and dissolution.

Experimental analysis

The amorphous advantage of pure ITZ

The as-received drug matched form I, and the melt-quenched material showed no diffraction peaks between 5 and 40° 2θ, only a broad halo (Figure 3). ITZ is known to form liquid crystalline phases (Six et al., 2001); the absence of low-angle peaks rules out the smectic phase, but a nematic phase trapped in the glass on cooling "is still possible". Because the melt-quenched solid was birefringent under polarized light, the authors judge it "likely" a glassy phase with some orientational order. The 50-50 spray-dried and extruded dispersions gave halo patterns.

Over the first 20 min, crystalline ITZ dissolved at 2.8 ± 0.2 µg/min/cm² and melt-quenched ITZ at 51 ± 2 µg/min/cm² (Table 2), about 18 times faster (the abstract calls this an "18 fold dissolution advantage"). After 60 min about 2.8 mg/cm² of amorphous ITZ and about 0.16 mg/cm² of crystalline ITZ had dissolved (Figure 4). Applying the polymorph rule, the authors equate the factor of 18 with the amorphous:crystalline solubility ratio and find it in "reasonable agreement" with literature amorphous solubilities, which lie 10–20 times above the crystalline value (Matteucci et al., 2008; Remenar et al., 2003) and "vary quite widely". Whatever the exact value, they conclude, the IDR data confirm the dissolution advantage of the amorphous form.

Drug loading and manufacturing route

Table A. ITZ release rates from compacts in 0.1 N HCl at 37 °C (from Table 2; card R4). The last column is read from Figure 5 (approximate; card R6) and agrees with the rate divided by 2.8.

Table A carries the first main result. At 50% and 25% ITZ, both processes released ITZ at 47–64 µg/min/cm², about 17–23 times the crystalline rate and close to the 51 µg/min/cm² of pure amorphous ITZ. The 50-50 curves of both processes overlap the melt-quenched curve for the whole hour (about 3.0, 2.6 and 2.8 mg/cm² at 60 min for SD, HM and melt-quenched ITZ, Figure 4). The authors read these loadings as drug-controlled: the properties of the amorphous drug "appears to dictate the release rate" at 50-50, and release "seemed to be primarily dictated by the drug" at 25-75. At 10% ITZ the picture changes: both processes released 199 µg/min/cm², about four times the amorphous rate, with large scatter (±53 and ±75) and less linear curves (R² 0.957 and 0.983), which the authors suggest "might be due to initial hydration and swelling of the polymer". (The text points to "Table 4" for these R² values; they are in Table 2.)

The 20% loading is where the two processes part: the spray-dried compact released 113 ± 2 µg/min/cm², about 40 times crystalline ITZ, while the extruded one released 48.8 ± 0.9, still at the amorphous-drug level. In Figure 5 the points therefore form a plateau near the amorphous value up to 75% polymer for both processes and up to 80% polymer for the extruded material, followed by a rise. The abstract places the switch "when the drug loading was approximately 20% or lower"; Section 5 puts the amorphous-drug control "up to 75 wt.% polymer" and adds that spray-dried dispersions move into the polymer-controlled region at a lower polymer fraction than extruded ones.Card check: the switch is stated in four slightly different ways in the paper (abstract; "at and below 20%"; "up to 75 wt.% polymer"; "plateau ... from 0-80% polymer"). By the data, the last drug-controlled loading is 25% for SD and 20% for HM (card note X6).

Polymer release runs ahead of drug release

Because both components were measured in each run, the paper can show what the polymer does while the drug is released (Figure 6). Pure PVPVA dissolved very fast, at 10.4 mg/min/cm² over the first 4.5 min, which the authors find in "reasonable agreement" with 8.1 mg/min/cm² reported by Chen et al. (2016). In every dispersion the polymer dissolved far more slowly than this, but in most of them still far faster than the drug.

Table B. Amounts released after 20 min (mg/cm²), read from Figure 6 (approximate; card R8). Melt-quenched ITZ: about 1.0 mg/cm² at 20 min.

Three observations come out of Figure 6 and Table B. First, from 50-50 to 20-80 the polymer is released proportionally faster than the drug: at 50-50, after 20 min, about 5 (SD) and 12 (HM) times more polymer than drug had dissolved. Second, the manufacturing route changes the polymer rate much more than the drug rate. At 25-75 and 20-80 the spray-dried compacts released polymer faster than the extruded ones (the text says "twice as fast" at 25-75; the curves show about 1.5–1.7 times), whereas at 50-50 the order reverses, with polymer leaving "approximately twice as fast" from the extruded material (about 2.4 times at 20 min). Yet the drug curves of SD and HM coincide at 50 and 25%. Hence the authors' statement that, except at 10-90, "a faster polymer release rate did not necessarily correlate with faster drug release", which "further suggests" that the amorphous drug controls release at most loadings. Third, at 10-90 both components leave quickly: the polymer curves rise steeply (from the spray-dried compact somewhat later than from the extruded one, but with similar slopes) and leave the plotted range within about 10–13 min, and "it appears that the polymer dissolution is controlling the rate of drug release".

Table C. Loading-normalized release rates of the spray-dried series, read from Figure 14, and polymer release relative to pure PVPVA, read from Figure 13 (all approximate; card R11, R12). The extruded dispersions in Figure 13: about 0.10, 0.05, 0.05 and 0.29 at 50, 75, 80 and 90% polymer.

Table C makes the comparison on equal terms. At 50 and 25% loading the normalized polymer rate is about 3.5–4 times the drug rate; at 20% (about 555 against 810) and 10% (about 2000 against 2265) the two are similar, which the authors read as drug and polymer leaving together, in the polymer-controlled regime ("These figures clearly show"). For the extruded dispersions the authors state that the polymer is faster than the drug at all compositions, though less so at lower loading. The last column shows a further point the authors make: even in the polymer-controlled 10-90 dispersions the polymer dissolves at only a fraction of the pure-polymer rate, so "the polymer release rate itself has been modified by the presence of the drug", as Chen et al. (2016) also observed.Card check: the normalization in Figure 13 is not fully defined. Its spray-dried values at 80 and 90% polymer agree with the content-normalized polymer bars of Figure 14 divided by the PVPVA-Raw bar, but those at 50 and 75% agree only without that normalization; the figure also shows a spray-dried point at about 85% polymer for which no formulation is listed (card note X7).

The spray-drying solvent at 20% loading

Table D. Cumulative release (mg/cm²) of the solvent series, read from Figures 7 and 8 (approximate; card R13, R14).

Holding the composition at 20% ITZ and changing only the solvent changed ITZ release several-fold (Figure 7, Table D). The batches fall "roughly into two groups". The water, DCM-alone and reference batches dissolved considerably faster than amorphous ITZ; adding water gave the fastest release, "unexpectedly" faster than the reference, and DCM alone was "roughly similar" to the reference (its curve starts slower and crosses the reference at about 33 min). The DCM–n-hexane and DCM–cyclohexane batches released ITZ "at a significantly slower rate" than the reference, somewhat slower than amorphous ITZ but still well above crystalline ITZ. Drying the n-hexane batch at 70 °C instead of about 50 °C "did not affect the release rate" in the authors' reading.Card check: in Figure 7 the 70 °C curve runs below the 48 °C one (about 0.55 against 0.75 mg/cm² at 20 min); both stay in the slow group (card note X10). Polymer release, by contrast, showed "no discernible differences" between solvents (Figure 8): all batches had released about 15–18 mg/cm² by 30 min, the hexane batches being slightly ahead in the first minutes.

The authors interpret the two groups with the same drug–polymer comparison as before. For DCM–methanol with or without water, and for DCM alone, "the normalized release rate of the drug corresponds to the polymer release rate", so drug release is controlled by polymer dissolution. For the hexane solvents the polymer is released faster than the drug and the drug rate is similar to, in fact slightly below, that of amorphous ITZ, so release "is largely controlled by drug". The shortfall below amorphous ITZ "may be due to the pits" seen in the particle shells, which "may impact sample wetting".Card check: normalized drug and polymer rates are not shown for the solvent series; Figure 14 covers only the standard spray-dried series, so the assignment of each solvent batch to a regime is an inference from Figures 7 and 8 (card note X11).

Bulk structure: no phase separation detected

Pyrene gave a peak I/III ratio of 1.29 ± 0.06 in spray-dried ITZ and 2.06 ± 0.02 in PVPVA, and 1.75–1.79 in all the 20% spray-dried batches (Table 5). The probe therefore sits in a similar average environment in every batch, which "suggests that discrete drug-rich and polymer-rich phases are not formed". Table 5 lists 20-80SD-DCM, 20-80SD-18water, the two n-hexane batches and the cyclohexane batch, but no row labelled as the DCM–methanol reference.Card check: "20-80SD-DCM" in Table 1 is the DCM-alone batch, so it is unclear whether the reference was measured (card note X4).

TEM of the reference particles showed uniform contrast with no sign of phase separation, and the particles from the water-containing solvent looked the same, which "suggests that water did not cause separation of drug and polymer to drug-rich and polymer rich regions detectable using TEM" (Figure 9). The hexane batches showed small, lighter globules filling much of the particle. These were "initially assumed to be polymer-rich domains", but SEM showed them to be hollows in the particle shell.

SEM (Figures 10, 11) showed that all particles were roughly spherical and hollow. Spray-dried ITZ formed smooth spheres and spray-dried PVPVA raisin-like, buckled particles. The reference and water batches were raisin-like, like the polymer; the DCM–hexane batches were smooth spheres, like the drug, and their fractured shells were full of pits, also when dried at 70 °C. Some fractured shells of the water batch showed corrugations running across the wall; the other shells looked homogeneous. A compact pressed at 1000 psi still showed nearly spherical particles on its fractured edge and spaces between particles on its face (Figure 12), so compaction did not obviously destroy the particles.

Surface composition: where the batches differ

Table E. XPS of spray-dried powders (from Table 6, n = 3; card R20). A homogeneous 20% ITZ surface would give a polymer:drug weight ratio of 4.

XPS was the only analysis that separated the batches clearly. For a homogeneous surface with 20% drug, the polymer:drug ratio would be 4. The reference came close (about 2.8 by N 1s and 4.0 by O 1s). The n-hexane batches carried more drug at the surface (31.8–35.7% by N 1s, 35.0–39.0% by O 1s), "marginally" more when dried at 70 °C. Adding water to the solvent gave a drug-poor, polymer-enriched surface (4.2% by N 1s). The N 1s and O 1s fits differ in size but agree in order, a disagreement the authors put "most likely" down to the peak-fitting model; the small polymer signal in pure ITZ and drug signal in pure PVPVA show the size of the fitting error. The chlorine signal, which only ITZ contributes, follows the same order: Cl 2p/C 1s of 0.0067–0.0072 for the n-hexane batches, 0.0050 for the reference and 0.0011 for the water batch, against 0.0421 for pure ITZ.Card check: in the reference row the printed polymer/drug ratios (2.8 and 4.0) are not exactly the ratios of the printed percentages (71.9/28.1 = 2.56; 79.7/20.3 = 3.93); the other rows agree within rounding. Averaging ratios over the three replicates, with the large ±9.4 scatter of that row, could explain it, but the paper does not say (card note X1).

The Cl 2p spectra held one more difference. Pure ITZ shows one spin-split pair with Cl 2p_3/2 near 200 eV, the position for organic chlorine. The dispersions show a second pair near 197 eV. Ionic chloride from sodium chloride contamination would produce such a shift, but no sodium was detected, so the shift "suggests the existence of chlorine atoms of itraconazole in an altered electronic environment", which "could result from changes in interaction" of ITZ with drug or polymer molecules. The share of this component rose from 9.0–9.5% in the n-hexane batches to 15.8% in the reference and 62.5% with water, where, by the other XPS results, ITZ "appears to be" in a more polymer-rich environment.

Discussion

How the authors connect the results. The discussion builds one picture from the rate data. The melt-quenched IDR is the benchmark for drug-controlled release. When the polymer dissolves faster than the drug, the outer layer of the compact loses polymer and becomes richer in drug; drug release then proceeds from a drug-rich surface at about the amorphous-drug rate. The authors call this a "supposition" and note that it also explains why the PVPVA release from the 50:50 compacts starts fast and then slows: as the surface becomes richer in ITZ, the polymer has to pass through a thickening drug-rich layer. When the drug content is low enough, drug and polymer leave together and the drug can be released far faster than the amorphous drug alone.

Why the processes differ in polymer release. The authors argue that it is "likely" that the differences in PVPVA release between spray-dried and extruded dispersions come from different kinetics of the polymer solid–liquid boundary, which recedes at different speeds in compacts of the two materials. Because the drug controls release at 25% loading and above, this changes the polymer rate without changing the drug rate, and no correlation between the two is seen. Why the boundary recedes differently "is currently unknown" and "likely to be complex". Neither the boundary nor the composition of the dissolving compact surface was measured.

Why the solvent changes the surface. The authors offer two mechanisms, both hedged. For the water batch, which had a polymer-rich surface "somewhat surprisingly", they invoke the Peclet number: polymer enrichment "is thought to occur" when drying is fast relative to polymer diffusion, and "it appears that PVPVA has a large Peclet number in this particular solvent system" at the point where the polymer skin forms. DCM–methanol without water gives a surface close to the bulk composition, "suggesting a lower Peclet number". They note that a published case study (Lowinger et al., 2015) found the opposite, drug enrichment when water was added, but in a system containing a surfactant whose role is unknown. For the hexane batches, they invoke precipitation: a component less soluble in the drying solvent should precipitate first and concentrate at the surface because precipitated material diffuses slowly. As the zeotropic DCM–hexane droplet loses DCM, it becomes richer in hexane, in which ITZ is much less soluble, so "differential evaporation could conceivably lead to drug precipitation and surface enrichment as observed". Because PVPVA is also insoluble in hexane, they add that "this explanation may be an over simplification". No evaporation rate, diffusivity or precipitation sequence was measured.

From compacts to products. The authors are explicit that compact rates cannot be transferred directly to tablets, whose excipients change dissolution, but argue that "comparative trends between different candidate formulations can still be employed" to understand and optimize the release of formulated products. They also point out that a compact of packed particles presents a constantly renewed surface, so the measured rate reflects the combined dissolution of particle layers, not of one surface. In their conclusions they state that surface-area-normalized dissolution "can be a useful technique to differentiate subtle differences in the drug-polymer distribution" caused by manufacturing conditions, that drug loading "is likely to play a critical role in determining the in vivo performance for some ASDs systems", and that the work "highlights the importance of considering the solubility of both the drug and the polymer in a given solvent system".

Note by the AI writer (not discussed in the paper). The 20% loading of ITZ–PVPVA is the most informative composition in the study precisely because it sits at the switch. There the two processes differ by a factor of about 2.3 in ITZ rate (113 against 48.8 µg/min/cm²), and the spray solvent alone spans about 0.5 to 2.5 mg/cm² of ITZ released at 20 min. A product formulated near such a switch is sensitive to process details, while one formulated well inside either regime is not; for comparing products made by different routes, as in the generic-equivalence setting that motivates the paper, the drug loading relative to the switch matters as much as the route. The data also suggest that, for spray-dried particles near the switch, the surface drug:polymer ratio is a more release-relevant attribute than bulk homogeneity, which pyrene and TEM could not distinguish between batches.

Limitations

Limitations stated by the authors. The authors limit their findings to the one drug–polymer combination studied, while noting agreement with Simonelli et al. (1969). They state that n-hexane and cyclohexane are not typical spray-drying solvents and were chosen only to widen the design space so that their anti-solvent effect on the polymer could be evaluated. They note that XPS sees only about 10 nm of the surface, that the N 1s and O 1s fits disagree in size, that their explanation of drug surface enrichment "may be an over simplification", and that compact rates cannot be extrapolated directly to tablets.

The AI writer's own judgement (1): what carries the main conclusion. The drug release rates are well supported: ten samples in triplicate with small standard deviations, and the plateau at 50 and 25% loading and the polymer-controlled release at 10% are clear. The threshold itself rests on a single intermediate composition per process, and it differs between them (between 25 and 20% for spray drying, between 20 and 10% for extrusion), which the paper states in four slightly different ways. Polymer release rates for the two processes are given only as curves, and the "twice as fast" factors are larger than the plotted ones (card notes X6, X9).

The AI writer's own judgement (2): the surface link. The link between surface composition and release regime is a correlation across four batches analysed by XPS (two n-hexane batches, the reference and the water batch). The DCM-alone and cyclohexane batches, and all extruded dispersions, have no surface analysis. The hexane batches also differ from the others in particle shape (smooth spheres rather than raisin-like particles) and in shell porosity (pits), so surface chemistry, morphology and wetting change together and the design cannot separate them; the authors themselves use the pits to explain the small shortfall below amorphous ITZ. Normalized drug and polymer rates, which the authors use to assign each solvent batch to a regime, are not shown for the solvent series (card notes X11, X12).

The AI writer's own judgement (3): conditions and reporting. All dissolution was done in 0.1 N HCl, where ITZ, a weak base (pKa 3.7), is most soluble; the regimes and the switch may sit elsewhere at intestinal pH, under non-sink conditions or in a formulated tablet. Several reporting details need care: Figure 13's normalization does not match Figure 14 at every loading and shows an unexplained point at about 85% polymer; Table 5 lacks a row for the reference batch; the reference row of Table 6 gives ratios that do not follow exactly from its percentages; the n-hexane solvent ratio is printed in two orders; and the text contains small slips ("Table 4" for Table 2, "methane" for methanol, "methanol with n-hexane" for DCM with n-hexane, "90:10" for the 10-90 dispersion). None of these changes the direction of the findings (card notes X1–X5, X7).

Conclusion

The paper shows, for one drug–polymer pair, that surface-area-normalized dissolution of compacts, with drug and polymer measured together, can tell which component controls release. For ITZ–PVPVA in 0.1 N HCl, amorphous ITZ dissolves about 18 times faster than crystalline ITZ; dispersions at 50 and 25% loading release ITZ at about that amorphous rate whether spray dried or extruded, while the polymer runs ahead at a process-dependent pace; at 10% loading both release ITZ almost four times faster than amorphous ITZ, carried by the polymer. At 20% loading the two processes differ, and so do spray-dried batches made from different solvents: hexane-containing solvents gave drug-enriched particle surfaces and slow, drug-controlled release, whereas DCM–methanol with or without water gave polymer-controlled release. The authors conclude that a threshold drug-to-polymer ratio exists above which the amorphous drug controls dissolution ("suggests"), and that the solubility of both drug and polymer in the spray solvent should be considered when designing spray-dried dispersions. The mechanisms they propose for the process and solvent effects (a receding polymer front, the Peclet number, solvent-driven precipitation) remain to be tested.