Watching a copovidone dispersion dissolve: droplets, networks and the limit of congruency

Background and motivation

Amorphous solid dispersions (ASDs), molecular mixtures of an amorphous drug in a polymer, are among the most used ways to deliver poorly soluble drugs, and the number of approved ASD products has grown over the last decade (Section 1). Their weak point is drug loading (DL, the weight percent of drug in the dry dispersion). For many drug–polymer pairs, release from the dispersion is fast and complete only up to a certain loading. Above it, release drops sharply. The authors note in the abstract that this threshold can be as low as 5% DL, which raises the pill burden. C1

Earlier work from the same group and others had described two release regimes for copovidone (PVPVA, Kollidon VA 64) dispersions. At low DL the drug and the polymer leave the dispersion at the same percent rate and both release completely; this is called congruent release. Drug concentrations in the medium can then exceed the amorphous solubility, and drug-rich nanoparticles form by liquid–liquid phase separation (LLPS). C2 An isotope-scrambling study had placed the formation of these nanoparticles in a high-mobility environment, either in solution or in a low-viscosity region near the dissolving surface. C3 At higher DL the two rates diverge and both fall, the drug more strikingly, a behaviour called "falling off a cliff". At still higher loadings neither component releases. The DL at which congruent release ends is the limit of congruency (LoC); reported values span 5–40% DL. C4·R15

Why the cliff happens was not settled. Water taken up by an ASD can cause amorphous–amorphous phase separation (AAPS), thermodynamically the same phenomenon as LLPS but at higher concentrations. AAPS had been studied mostly for its effect during processing and storage. It had been proposed to occur at the surface of a dissolving ASD, but it had rarely been watched there as it happens. C5 One proposal was that AAPS produces an amorphous, insoluble barrier on the surface, which has been reported as drug-rich or as containing polymer, up to about 75% polymer for a phenolic compound in copovidone. C6·R15 Han and coworkers went further and proposed that the morphology of the separated phases decides the outcome: a continuous hydrophobic network across the interface gives incongruent release, while discrete hydrophobic domains inside a continuous hydrophilic phase give congruent release. C7 Factors suggested to move the LoC included the strength of drug–polymer interactions, the drug's glass transition temperature (Tg) and additives. C8

The authors chose two model compounds whose behaviour in copovidone was already known from their earlier work (Que and coworkers, 2021): phenolphthalein (PHPH), with an LoC of 5% DL, and its methoxy derivative (MePHPH, both phenolic OH groups methylated), with an LoC of 40% DL. R14 The two molecules share a scaffold, but only PHPH has hydrogen-bond donors that can bind to PVPVA. The study set out to develop an in situ confocal microscopy method that shows phase separation at the ASD/solution interface during hydration and release; to use it, together with release tests and an analysis of the gel layer, to explain the release profiles; to measure the Tg and composition of the hydrophobic phase formed during release; and to propose a framework for predicting the LoC of binary ASDs qualitatively (Section 1).

Background concepts

The gel layer of a dissolving glassy polymer. PVPVA is a linear, nonionic, glassy polymer whose molecular weight lies above the critical value for chain entanglement. Such polymers dissolve through a gel layer (Section 4.7.1, renumbered 5.1 by the correction; Figure 14). Water entering the dry glass plasticizes it and causes a sharp transition from glass to rubber; this narrow front is the glass/gel interface. Near it the chains are entangled and the gel is viscous, so whole chains move slowly even though their segments are mobile. As more water diffuses in, the gel swells, its viscosity falls and the number of entanglements per chain decreases. At a critical water content the chains disentangle and leave into solution; this is the gel/solution interface. Beyond it lie the diffusion boundary layer and the bulk solution. C42 When both interfaces move at the same constant rate, the gel thickness stays constant and a flat surface releases polymer at a constant (zero-order) rate. Water penetration into the gel is described as Fickian and slows as the gel thickens, while chain release speeds up as the surface becomes more dilute, until the two rates match; chain release is described by reptation, and water penetration into the glass itself as non-Fickian. C43

Gel residence time. The authors define the residence time of the gel layer as the time needed for one gel-layer thickness to release, calculated as the gel layer weight divided by the total surface release rate of drug plus polymer (Section 4.2). E3 A short residence time means the material in the gel is exchanged quickly.

Congruent and incongruent release. Release is congruent when drug and polymer leave at the same percent rate; it is incongruent when the drug lags or stops. In this paper an incongruent release is also detected by a rise in the drug share of the gel layer: if the drug cannot leave with the polymer, it accumulates in the gel (Section 4.2). C13

Water-induced phase separation and its two kinds. When water enters a dispersion of a hydrophobic drug in a hydrophilic polymer, the mixture can split into a drug-rich hydrophobic phase and a water- and polymer-rich hydrophilic phase. Where the polymer goes matters. If it prefers the hydrophilic phase, the separation is called segregative; if it prefers to stay with the drug in the hydrophobic phase, it is associative (Section 4.7.5). C65 An associative hydrophobic phase carries much polymer and is therefore larger in volume.

Phase inversion. In a two-phase amorphous system the minority phase usually forms droplets in a continuous majority phase. As the minority grows, the morphology passes through a co-continuous state and then inverts: the former droplets become the continuous phase. For a drug-rich phase in a gel, this inversion is the change from separate droplets to a connected network.

Reading the two-dye images. The ASDs used for imaging contained two fluorescent dyes in trace amounts: prodan (blue), which is relatively hydrophobic and qualitatively follows the drug, and rhodamine 6G (red), which is relatively hydrophilic and follows the polymer and water. C10 A magenta region means both dyes are still mixed at the 230 nm resolution of the microscope; separate blue and red domains mean phase separation. Black regions are pores or regions where scattering by separated domains removes the signal; the authors stress that such dark regions still contain drug, polymer and dyes. In the in situ images, a dashed blue line marks the glass/gel interface, a white line the gel/solution interface (roughly where the red signal vanishes) and a red line the onset of phase separation. C23

Ternary phase diagram and plait point. A drug–polymer–water phase diagram is a triangle whose corners are the three pure components. In the type I diagrams that PVPVA dispersions usually give, drug and polymer mix in all proportions, and so do polymer and water; the drug–water side shows a miscibility gap that extends into the triangle as a two-phase region. C66 Tie lines join pairs of coexisting compositions; the plait point is the point on the edge of the two-phase region where the two coexisting phases become identical. The hydration path is the line of compositions a dry ASD passes through as it takes up water. C67 No phase diagram of either system is measured or computed in this paper; the diagrams it shows are schematics (Figures 17 and 18). P12·X12

Methods in detail

The study combines four kinds of experiment on compacts of the same dispersions with a literature-based framework. The flow below shows how they fit together.

Materials and dispersions

PHPH was bought from MilliporeSigma; MePHPH was synthesized in-house; PVPVA was Kollidon VA 64 from BASF. M1·M2·M3 Drug and polymer were dissolved in ethanol and rotary-evaporated at a 70 °C bath temperature, then dried under vacuum overnight; dispersions for imaging also contained 0.05 w/w% rhodamine 6G and 0.1 w/w% prodan, dissolved with the components before evaporation. P1·M4·M5 Between 100 and 200 mg of dispersion was pressed at 1500 psi for 60 s in a Wood intrinsic die exposing a 0.5 cm² face, and the compact was used at once, still in the die. P2 MePHPH dispersions spanned 5–70% DL and PHPH dispersions 5–20% DL; neat PVPVA served as the reference. A2·A3·A5 The onset Tg values of the neat materials, measured by DSC, were 22.5 °C for MePHPH, 84.5 °C for PHPH and 101.3 °C for PVPVA (Table 3). M1·M2·M3

Contribution 1: release and the gel layer, measured on the same compacts

Problem. Release curves alone show that something goes wrong above the LoC but not where. The authors therefore analysed the gel layer itself at each time point.

What they did. Compacts were dissolved in 100 mL of pH 6.8, 50 mM phosphate buffer at 37 °C and 100 rpm, and drug and polymer in 0.5 mL aliquots were measured by HPLC at 10, 20 and 30 min, in triplicate per time point; percent release was calculated on a 100 mg compact weight. P3·P5 At each time point the run was stopped, the gel was scraped off the compact within 2 min, weighed in a capped vial and dissolved in ethanol for HPLC. Water was taken as the gel weight minus the drug and polymer found by HPLC. P4 The gel weight was converted to a thickness with Equation 1:

In plain terms: the gel volume $V$ is the sum of the volumes of drug, polymer and water, each its weight $w$ divided by its density $\rho$ (subscripts 1, 2 and 3), and dividing by the exposed area $A$ gives the thickness $h$ of a flat disc. The authors used 1.19 g/mL for PVPVA and the vendor's crystal density of PHPH, 1.25 g/mL, for both amorphous drugs, which they call "a reasonable approximation"; volumes are assumed additive. E1·C9 The thickness was also measured directly on cross-sectional images (below), at five positions per image. P9

From these data come the total release rate of the gel (drug plus polymer, in mg·min⁻¹·cm⁻²), the water-to-polymer weight ratio and the residence time defined above. R4·E3

Note by the AI writer (not discussed in the paper): with the printed thickness, water content and densities, the total gel weight implied by Equation 1 gives residence times longer than those in Table 2 (for neat PVPVA about 7.9 min against the printed 5.7 min, taking water as 1.00 g/mL). Counting only the drug and polymer in the gel gives about 5.9 min, close to the table; the printed values therefore seem to use the dry gel mass. X3

Contribution 2: three ways of looking at the interface with two dyes

Problem. Phase separation at the dissolving surface had been proposed but seldom observed in situ. The authors built an imaging set-up that shows it while it happens (Figure 2).

In situ imaging. A compact was taken out of the die and fused onto the glass bottom of a dish heated to about 120–200 °C with light pressure, so that the material stuck without deforming; a glass barrier greased onto its top stopped water from entering from above. The dish was filled with buffer at 37 °C, mixed by hand with a pipette, and the bottom edge of the compact was imaged every 30 s with a 60× oil objective (numerical aperture 1.49) as 15 stitched frames of 1024 × 1024 pixels (100 nm pixels, 230 nm optical resolution). Prodan was excited at 409.0 nm (emission 425–475 nm) and rhodamine 6G at 561.5 nm (emission 570–620 nm). The distance from the glass/gel interface to the onset of phase separation was measured by eye on the 3, 4, 5, 6 and 7 min frames and averaged. P6·P7 The authors caution that the gel/solution boundary in this set-up depends on the gentle manual mixing, so the in situ gel size "may not be" representative of the release experiments. C24

Surface imaging. Compacts were dissolved in the intrinsic apparatus for 2, 10 or 20 min, taken out of the die and pressed face-down against a viewing glass. Images were taken at three scales: native frames, digital magnification and 3 × 3 stitched frames. P8

Cross-sectional imaging. Compacts treated the same way were cut in half with a blade and imaged at the cut face with a 20× objective, stitching 4–10 frames and stacking images every 1–2 μm to a depth of up to 200 μm. P9

How this differs from earlier work. The dye pair had been used to study water-induced phase separation of miconazole–PVPVA dispersions exposed to humidity (Saboo and Taylor, 2017). C10 Here it is applied to compacts in liquid buffer, at the moving dissolving interface, and combined with release and gel data from the same compacts.

Contribution 3: the hydrophobic phase that leaves the compact

To learn what the drug-rich phase contains, the authors released whole compacts of 30% DL MePHPH or 5% DL PHPH, loadings they describe as at the LoC of each system, in 10 mL of buffer at 37 °C with stirring at 500 rpm for 30 min. The nanoparticles were spun down at 21100 g and 37 °C, excess water was wicked off, and the pellets were analysed by HPLC (dry basis). P10·C11 For DSC, pellets from one or two 200 mg compacts stirred for 60 min were sealed wet in hermetic pans, cooled to −90 °C and heated at 2 °C/min with a modulation of ±1.27 °C every 60 s, in triplicate. P11 The water content of these pellets was not measured. X14

Note by the AI writer (not discussed in the paper): the paper gives the MePHPH LoC as 40% DL from the earlier study but calls 30% DL "at the LoC" when choosing this loading; no release data exist for 30% DL. X1

Contribution 4: a framework from polymer dissolution and phase diagrams

In the Discussion the authors combine the classical picture of polymer dissolution with schematic type I phase diagrams to explain the LoC qualitatively (Figures 14, 17 and 18). P12 This part uses no new data; it is described in the Discussion below. Equation 2, the Fickian law they quote for water entering the gel, is:

In plain terms: the rate at which solvent mass $m_1$ enters the gel over time $t$ is set by the solvent diffusion coefficient $D$, the contact area $A$ and how steeply the solvent changes with position $x$ in the gel; as the gel thickens, the gradient flattens and penetration slows. E2 Note by the AI writer: as printed, the equation is not dimensionally consistent (the gradient should be of a concentration, and Fick's law carries a minus sign); the paper uses it only qualitatively. X16

Experimental analysis

Release falls off a cliff at the expected loadings

Figure 3 shows the release of drug and polymer from the compacts. For MePHPH, 20% DL released both components faster than neat PVPVA, and 40% DL still released at least as fast; at 50% DL release started only after a lag of more than 10 min and stayed low, and at 70% DL nothing was released within 30 min. For PHPH, 5% DL released faster than neat PVPVA, while 20% DL did not dissolve at all. R1·R2·C12 The authors describe this as "expected" from the LoCs of 40% and 5% DL measured earlier. C12·R14

Table A. Release at 30 min, read from Figure 3 (approximate). R1·R2

Two details matter later. At 40% DL the drug slightly lagged the polymer by 30 min (about 61 against 72%). R1·C61 And because percentages are calculated on a 100 mg compact weight while compacts weighed 100–200 mg, values above 100% (about 105% polymer at 20% DL) occur; the percentages are best read as milligrams released per 100 mg. X9

The drug piles up in the gel before release fails

Figure 4 gives the drug share of the scraped gel on a dry basis. For 20% DL MePHPH and 5% DL PHPH it did not change over 30 min. For 40% DL MePHPH it rose from 40% to about 45, 46.5 and 49% at 10, 20 and 30 min, and for 50% DL from 50% to about 52, 54.5 and 59.5%. R3 The authors argue that this is a more sensitive test of congruency than the release curves, because the drug builds up in the gel when it cannot leave with the polymer. They conclude: "This suggests that the true limit of congruency for the MePHPH-PVPVA system is below 40%", even though the earlier study had found at least 80% drug release at that loading. C13·C14·R14 Because only 20, 40 and 50% DL were analysed, this criterion places the LoC between 20 and 40% DL. X18

The gel layer below and above the LoC

Table 2 summarises the gel layer. Its numbers give the clearest quantitative picture of what changes with loading.

Table B. Gel layer during dissolution, from Table 2 (mean ± SD, n = 3; cross-section n = 5). * Not at steady state; values after 30 min. R4

Four patterns stand out. First, the two thickness methods agree reasonably well. C15 Second, below the LoC the MePHPH gel was thinnest at 20% DL and thickened as loading rose; above the LoC the 50% DL gel was still growing at 30 min, and the 70% DL value is not at steady state. R4·C16 Third, the water content fell with loading (from 25% for neat PVPVA to 11% at 50% DL and 1.5% at 70% DL), but the water-to-polymer ratio stayed between 0.25 and 0.41 except at 70% DL. R4·C18 The authors take this to mean that, at low loading, about as much water per polymer is needed for the chains to disentangle as in neat PVPVA ("it can be inferred"). C52 Fourth, the release rate depends on loading even below the LoC: for MePHPH it rose to a maximum at 20% DL, 70% faster than neat PVPVA, fell to about the neat-polymer value at 40% DL and dropped further above, while the residence time had its minimum (1.9 min) at 20% DL. R4·C20

PHPH behaved differently. The 5% DL gel was about 40% thicker than the neat PVPVA gel, which the authors find "somewhat surprising", yet it released faster, so its residence time (5.3 min) matched that of neat PVPVA. R4·C16·C21 The 20% DL PHPH gel was the thickest of all (586 μm); the authors attribute this to water still entering while neither drug nor polymer leaves ("can be attributed"). The near-dry 70% DL MePHPH gel is "likely water-transfer-limited", with a hydrophobic barrier potentially covering much of the surface. C17·C19 From the set of results they conclude that loading "can influence release rates both below and above the LoC", in line with one earlier study though not with others. C22·R17

MePHPH: droplets inside the gel, then networks, then inversion

In every in situ series of MePHPH dispersions, blue and red domains separated somewhere in the gel, which the authors read as segregative phase separation into a hydrophobic and a hydrophilic phase. R6·C25 At 5% (Figure S2), 20% (Figure S3) and 30% DL (Figure 5) the blue domains were discrete particles with diameters below 1 μm. Phase separation started 155 ± 43, 57 ± 14 and 45 ± 20 μm from the glass/gel interface: the higher the loading, the closer to the dry core. R6·R8 Particles were also seen beyond the gel/solution interface, which the authors take as nanoparticles being released into solution as the gel dissolves ("indicating"). R6·C27

At 40% DL (Figure 6) some hydrophobic domains connected into networks in the image plane; continuous regions formed on the compact at some time points and ruptured intermittently. At 50% DL (Figure 7) the morphology had inverted by 60 s: the hydrophobic phase was continuous and the hydrophilic phase formed separate pockets; this persisted and coarsened up to 2100 s. The networks at 40% and 50% DL did not deform under the flow. R6·C26·C33 The authors summarise that the proportion of interconnected hydrophobic domains increases with drug loading, which is "visually apparent". C26

PHPH: particles beyond the gel, then a network already at 10–20%

In PHPH dispersions both dyes moved into the hydrophobic phase after separation, so that phase stayed magenta while the hydrophilic phase lost both dyes and looked black. The authors read this as a hydrophobic phase much less hydrophobic than that of MePHPH, consistent with associative phase separation in which much polymer joins it ("This indicates"). R7·C28 At 5% DL (Figure 8) sub-micrometre domains appeared 601 ± 68 μm from the glass/gel interface. R7·R8 At 10% DL (Figure 9) the hydrophobic phase was partly interconnected, and at 20% DL (Figure 10) it was entirely interconnected, with the hydrophilic phase as separate pores, whereas 20% DL MePHPH gave discrete droplets. R7·C32 The PHPH network deformed under the flow, which the authors take as a sign that it "may have a lower mechanical strength", consistent with the low Tg of its hydrophobic phase. R7·C33

Note by the AI writer (not discussed in the paper): the text cites "Figure S2" for 20% DL MePHPH, but Figure S2 shows 5% DL and Figure S3 shows 20% DL. X7 The caption of Figure 10 swaps the colour key of the dashed lines relative to Figures 5–9. X5 And in the 1800 s frame of Figure 9 the partly connected domains of 10% DL PHPH lie beyond the marked gel/solution line, so it is not clear whether this network sits in the gel or in front of it. X19

Where phase separation starts, relative to the gel

Figure 15 puts the onset positions next to the gel thickness from Table 2. For MePHPH the onsets lie well inside the gel; for 5% DL PHPH the onset lies far beyond it. The authors conclude that MePHPH nanoparticles form within the gel layer ("suggesting") and that the 5% PHPH nanoparticles "appear to originate outside the gel layer". R8·C27·C31

Table C. Onset distances from the text, gel thickness from Table 2 (cross-sectional imaging unless noted), morphology from Figures 5–10 and S2–S3. R4·R5·R6·R7·R8

Two cautions apply. Figure 15 plots the 30% DL MePHPH gel/solution boundary at about 151 μm, whereas Table 2 gives 171 ± 15 μm; the other four values in Figure 15 match Table 2. X2 More importantly, the onset distances come from the in situ set-up with gentle hand mixing, and the gel thicknesses from the intrinsic apparatus at 100 rpm, a comparison the authors themselves warn about. The margin is large for 5% PHPH (about 245 μm beyond the gravimetric thickness the authors quote) and modest for 5% MePHPH (about 72 μm against an onset uncertainty of ± 43 μm). X4·C24

The hydrophobic phase: almost pure drug for MePHPH, mostly polymer for PHPH

Table D. Glass transition and composition of the released hydrophobic phase, from Table 3. R12·M1·M2·M3

The MePHPH-rich phase was almost pure drug, and its Tg was close to that of neat MePHPH; MePHPH, with a calculated log P of about 4, is "not expected" to absorb much water. R12·M2·C40 The PHPH-rich phase held about 75% PVPVA on a dry basis, and its Tg of −5.8 °C lay far below those of both neat PHPH and PVPVA, which the authors explain by water in the phase: "it is likely that a considerable amount of water is present". R12·C29 Both compositions resemble the equilibrated hydrophobic phases of the earlier study, and both Tgs are below 37 °C, which the authors take to mean that the phases are mobile enough to reach compositional equilibrium during dissolution ("suggesting"). R18·C39

Note by the AI writer (not discussed in the paper): the Tg was measured on wet pellets whose water content was not measured, so the water explanation is plausible but untested. X14 The two compositions also come from different loadings and from particles released in bulk at 500 rpm, not from the gel; for PHPH at 5% DL these particles form outside the gel, while the network that matters appears at 10–20% DL. X11 The earlier equilibrated compositions are attributed to reference 23 here but to reference 22 elsewhere in the paper. X15

What the surface and the cross-section show

Surface images after partial dissolution (Figure 11) repeat the in situ picture. At 30% DL MePHPH the surface carried hydrophobic domains of about 1 μm or less, with a few larger ones that "may have been due to coalescence or aggregation", and looked uniform at low magnification. At 40% DL large interconnected domains at 2 min looked discrete at 20 min, but the overview was patchy, with continuous hydrophobic regions next to regions of discrete hydrophobic domains in a continuous hydrophilic phase. At 50% DL continuous hydrophobic domains covered a larger area and enclosed discrete hydrophilic pockets, the inverse of 30% DL. R9·C34·C35 For PHPH (Figure 12), 5% DL looked homogeneous, 10% DL showed shades of magenta without features, and 20% DL showed a porous structure; the authors explain the first two by particles forming outside the gel and by smearing of the wet gel end on the glass ("likely"). R10·C36 Neat PHPH, MePHPH and PVPVA compacts showed no phase separation, while 20% DL MePHPH did (Figure S4). R10

Cross-sections (Figure 13) add the depth dimension. At 30% DL the gel was macroscopically homogeneous. At 40% DL after 2 min the upper part of the gel, near the solution, was partly separated with connected-looking hydrophobic domains while the lower part near the dry core was homogeneous, which "may indicate" too little water near the core; by 20 min the segregation looked complete and the gel more homogeneous, "likely" because domains moved and coalesced. At 50% DL separation started immediately after the gel formed, and by 20 min a thick hydrophobic layer covered the compact. PHPH cross-sections showed no macroscopic segregation of the dyes. R11·C37·C38 Crystallization was not observed in any formulation, although the paper does not say how this was checked. R13·X13

Discussion

How the authors connect the findings

The Discussion (Section 4.7, renumbered 5 by the correction) assembles the observations into one chain. A gel layer formed on every compact, as "perhaps expected" with PVPVA as the main component, which "suggests" that ASDs "may" dissolve like neat polymers. C44 The drug is too poorly soluble to diffuse out of the gel, so its release is "solely" by erosion and dissolution of the polymer, an assumption the authors call reasonable under the non-sink conditions used and supported by congruent release below the LoC. C45 Within this frame, "the morphology of the two phases following amorphous phase separation appears to differentiate congruent release from incongruent release". C46

The good branch. From the micrographs the authors infer that during congruent release the drug is "carried" into the diffusion boundary layer by the swelling and release of polymer chains. Either it separates into nanoparticles inside the gel, as for 5–30% DL MePHPH, and the particles do not have time to coalesce in the viscous gel within the residence time of about 2–6 min ("It appears"), or it stays dissolved through the gel and separates in the diffusion layer or the bulk, as for 5% DL PHPH. C47·R4

The bad branch. Above the LoC both systems formed continuous or semi-continuous hydrophobic phases (Figures 7, 9 and 10). C55 Because the hydrophobic phase is poor in water, it is "expected" to become more viscous during separation; once connected, it adds structure and mechanical strength to the gel and, being anchored to it, "persists at the ASD/solution interface and therefore acts as an insoluble barrier". C56 In top view this barrier is porous (Figures 11j,k and 12c), as reported for ritonavir–PVPVA and indomethacin–PVPVA dispersions by electron microscopy and for ritonavir–PVP by confocal microscopy. C57·R16 The pores are first filled with hydrophilic phase, through which water enters and from which some drug and polymer can still leave, which accounts for some release just above the LoC; water entering through the pores moves the glass/gel boundary deeper, so the gel keeps growing, as seen for 50% DL MePHPH and 20% DL PHPH. C58·R4 At 70% DL the pores are expected to shrink and the network to stiffen, which "may" explain the small, nearly dry gel. C59

The borderline case. At 40% DL MePHPH release was "marginally faster" than that of neat PVPVA, yet continuous hydrophobic regions formed at times. The network appeared weaker than at 50% DL and ruptured intermittently; "a possible explanation" is that swelling pressure from water overcame its strength. C60 The release was nevertheless not fully congruent: drug lagged polymer and accumulated in the gel. C61·R3 With time, the authors expect the surface to become covered by the accumulating network, an effect that "may have a greater impact on thicker compacts". C62

Their final statement is stronger than the hedged chain: "it was found that in both cases, the loss of congruency was due to the evolution of a continuous hydrophobic network in the gel at the surface of the ASD". The abstract words the same result as a correlation: the morphology "was found to correlate with the release behavior". C72

Why release rate changes with loading below the LoC

For MePHPH the authors give a mechanism for the rise and fall of the release rate below the LoC (Section 4.7.3). The rise up to 20% DL "must be attributed" to faster water penetration into the glassy core and faster release of chains from the gel surface. Faster water penetration "may be explained" by two effects: a thinner gel, through which water diffuses faster, and plasticization of the dry polymer by MePHPH (onset Tg about 23 °C), which increases free volume, as residual solvent does in polymer glasses. C50 Because the gel thinned, polymer release must have sped up more than water penetration; "a possible explanation" is faster initial water entry plus phase separation moving most of the water into the hydrophilic phase, which dilutes the chains so they reach the disentanglement point sooner. C51 Figure 16 illustrates the idea: the hydrophilic phase around the MePHPH-rich nanoparticles is compositionally practically the same as the neat PVPVA gel, so it erodes like it. C52 The fall above 20% DL "may be due" to MePHPH's lipophilicity, which reduces the thermodynamic drive for water to enter and offsets the kinetic gains; "likely, this balance" decides whether rates rise, fall or stay flat with loading below the LoC, and near the LoC the barrier adds its effect. C53 The authors summarise that "a congruent release rate is controlled by the penetration rate of water into the ASD and the disentanglement of the polymer chains" ("it was proposed"). C73 PHPH could not be analysed this way because its LoC of 5% left too few loadings. C49

Why phenolphthalein fails so early

Strong drug–polymer interactions, especially hydrogen bonding, are known to help physical stability, but the earlier study had attributed the low LoC of PHPH–PVPVA to "adhesive hydrogen bonding". C64 The authors now place the effect in the phase morphology: hydrogen bonding increases the distribution of PVPVA into the hydrophobic phase, making it "volumetrically larger" (associative separation), so the hydrophobic phase connects into a network at a lower loading. The conclusions state it directly: "The variation in the LoC was due to the extent of polymer present in the hydrophobic phase, where a high polymer content resulted in low LoC." C65 The supporting data are the composition (26.2% drug against 98.8%), the dye behaviour and the network already at 10–20% DL. R12·R7·R6

The plait point as a guide to the maximum LoC

To turn this into a rule, the authors use the drug–polymer–water phase diagram (Section 4.7.5; Figures 17 and 18). In congruently releasing dispersions the hydration path runs straight from the dry composition toward the water corner; in incongruent ones it bends inside the two-phase region. C67 Since the volume of the hydrophobic phase decides whether it forms a network, a hydration path that enters the two-phase region to the right of the plait point, through the hydrophobic side of the boundary, makes the hydrophobic phase the majority, and "intuitively, it is expected" to form a continuous network; a path entering to the left makes the hydrophilic phase the majority and the hydrophobic phase discrete. Hence "the position of the plait point serves as an arbiter for approximating the maximum LoC value for a binary ASD system". C68 Three scenarios follow: associative systems have the plait point to the left, which "can result in a low maximum LoC"; segregative systems have it to the right and a higher maximum LoC; and a smaller miscibility gap also favours a higher maximum LoC. C69

The authors name the main assumption themselves: phase separation must be fast enough for the two phases to reach near-equilibrium compositions along the hydration path. This "may be valid" if the hydrophobic phase is mobile, and because both hydrophobic phases have Tg below 37 °C, "the systems studied here may conform"; high-Tg hydrophobic phases need further work. C70

Note by the AI writer (not discussed in the paper): the framework is drawn on generic schematic diagrams; this paper measures no phase diagram, plait point or hydration path for MePHPH or PHPH, so here it is a hypothesis consistent with the observations rather than a tested prediction. X12

What it suggests for formulation

The authors draw several practical points. Strong drug–polymer interactions chosen for physical stability can be detrimental to release in PVPVA dispersions, and the study "highlights the importance of drug−polymer interactions" for release as a function of loading. C64·C74 Drugs whose crystallization induction time is shorter than the gel residence time (about 5 min here) may crystallize in the matrix and lose congruent release; if this cannot be inhibited, "it may be worthwhile to consider a polymer carrier with a short gel layer residence time". C48 Because a partial network accumulates over time, thinner tablets or small, well-dispersed ASD particles "may have more complete drug release" near the LoC than compacts, though this "may be challenging" in tablets, which tend to disintegrate poorly at moderate-to-high ASD loads. C63 Finally, modelling of drug–polymer–water phase diagrams could "potentially" turn the framework into quantitative LoC predictions for in silico formulation design. C75

Note by the AI writer (not discussed in the paper): the gel-layer drug content is a simple readout that flagged 40% DL MePHPH as incongruent while its 30 min release still exceeded that of neat PVPVA, so it may catch an LoC that short release tests miss. R3·R1

Limitations

Limits the authors state

• The in situ gel size depends on the gentle manual mixing and "may not be quantitatively representative" of the release experiments. C24
• The low LoC of PHPH left too few loadings to analyse how loading changes its congruent release rate. C49
• The framework assumes near-equilibrium phase compositions, which "may" hold here but needs further work for high-Tg hydrophobic phases. C70
• The drug densities in Equation 1 are approximated by the crystal density of PHPH. C9

The AI writer's own judgement

The paper's strength is that release, gel composition and three independent imaging modes agree on where things change for both drugs: separate droplets go with good release, and a connected drug-rich phase goes with failing release, at 40–50% DL for MePHPH and 10–20% DL for PHPH. R1·R2·R3·R6·R7·R9·R11 The gel-layer data in Table 2 are quantitative, with standard deviations, and the composition contrast of the two hydrophobic phases is large. R4·R12 Several points limit how far the conclusions reach:

• The barrier is inferred, not measured. Release (compact in the die at 100 rpm) and morphology (gentle in situ mixing, or a compact taken out of the die) come from separate experiments, and at 40% DL a partial network coexisted with good release; transport of drug or polymer through the network was not measured. The conclusions' "was due to" is stronger than the abstract's "was found to correlate". X10·C72
• Inside or outside the gel. These statements compare onset distances and gel thicknesses from differently stirred set-ups; they are firm for 5% DL PHPH and approximate for MePHPH. X4
• The polymer-content explanation rests on thin data. One composition per drug, at different loadings, measured on particles released in bulk rather than on the gel; hydrogen bonding and the volume fraction of the hydrophobic phase in the gel were not measured. X11
• The LoC of MePHPH is not stated consistently. It is 40% DL from the earlier study, "at" 30% DL in the methods and "below 40%" from the gel data, which bracket it between 20 and 40% DL. X1·X18
• The plait-point framework is untested here. It rests on schematic diagrams only. X12
• Smaller source issues. Figure 15 gives a different 30% DL boundary from Table 2; the residence times match the dry rather than the total gel mass; the Figure 10 caption reverses the line colours; the text cites the wrong SI figure for 20% DL MePHPH; release percentages are on a 100 mg basis; crystallization was checked by an unstated method; and the HPLC flow rates are printed in μL·min⁻¹. X2·X3·X5·X7·X9·X13·X8
• Scope. Two closely related drugs, one polymer, compacts at one stirring rate and 30 min experiments.

Conclusion

This paper put the dissolving surface of copovidone dispersions under the microscope. With two dyes marking the drug-rich and water-rich phases, it showed that water splits the gel layer of MePHPH dispersions into sub-micrometre drug-rich droplets at 5–30% DL, which form inside the gel and leave with the polymer, and into a connected, eventually inverted network at 40–50% DL, where release fails. R6·R1·C26 Phenolphthalein, whose drug-rich phase is about three-quarters polymer, separated beyond the gel at 5% DL but formed a connected network already at 10–20% DL, where its release fails. R7·R12·R2 The gel-layer measurements add a quantitative frame: below the LoC the release rate and the gel thickness depend on loading, with a maximum rate at 20% DL MePHPH, and above it the gel grows, dries out and accumulates drug. R3·R4

The authors conclude that in both systems the loss of congruency came from a continuous hydrophobic network acting as a barrier, that the low LoC of PHPH comes from the polymer carried into its hydrophobic phase, and that the plait point of the drug–polymer–water phase diagram can approximate the maximum LoC. C72·C65·C68 The first two are well supported as correlations; the causal barrier role, the role of hydrogen bonding and the phase-diagram rule remain to be tested with transport measurements, gel-phase compositions and measured or computed diagrams. X10·X11·X12