When water splits an amorphous solid dispersion: ritonavir, three polymers and the origin of drug nanodroplets

Background and motivation

Amorphous solid dispersions (ASDs) raise the apparent solubility of a poorly water-soluble drug by holding it in a higher-energy amorphous state, molecularly mixed with a polymer that slows crystallization and helps dissolution (Hancock 1997; Murdande 2010; Vasconcelos 2007). C8 Two kinds of phase change can spoil this picture once water arrives. In the solid, absorbed water can split the dispersion into an amorphous drug-rich phase and a drug-lean, polymer-rich phase; this is amorphous–amorphous phase separation (AAPS), first studied in detail for PVP dispersions stored at high relative humidity (Rumondor 2009, 2011). C9 AAPS was known as a possible first step toward crystallization, because the separated drug no longer benefits from the polymer's inhibition (Rumondor 2009; Marsac 2010). C9·C42 What it does to dissolution was, in the authors' words, "unknown". C9

The second kind of phase change happens in solution. When a dissolving ASD pushes the free drug concentration above the drug's amorphous solubility, the solution undergoes liquid–liquid phase separation (LLPS): the excess drug forms disordered, drug-rich droplets, typically 200–500 nm across, while the surrounding solution stays at the amorphous solubility (Ilevbare 2013; Raina 2014). C10·R25 For ritonavir, LLPS appears at a similar concentration whether it is triggered by solvent shift, pH change, cooling or ASD dissolution (Ilevbare 2013). R21 The droplets matter in practice because they can act as a reservoir that keeps the free drug, and so membrane flux, at its maximum as drug is absorbed (Indulkar 2016). C48·R22

Two earlier observations frame the question of this paper. First, the drug loading (DL) decides how an ASD dissolves: in classic work on sulfathiazole–PVP, release at about 10% DL was controlled by the dissolving polymer and was fast, while at about 50% DL a drug-rich layer formed on the surface and controlled a slower release (Simonelli 1969, 1976). R23·C49 Low-loading dispersions are therefore the ones that usually reach LLPS (Alonzo 2011; Jackson 2015). C49 Second, the origin of the nanodroplets that ASDs release was disputed. They could form in solution, by LLPS after the drug has dissolved; or, as Harmon et al. proposed for a copovidone dispersion in 2016, they could form inside the wetted matrix by AAPS into nanosized drug-rich domains that the polymer stabilises and then releases. C51

The authors therefore asked how drug loading and polymer type change phase transformations inside the ASD matrix and how those transformations in turn change dissolution and supersaturation. C12 They chose ritonavir (RTV) as a hydrophobic model drug and three polymers that differ in how much water they take up and in their hydrogen-bonding groups: polyvinylpyrrolidone K-90 (PVP), polyvinylpyrrolidone vinyl acetate (PVPVA) and hydroxypropyl methylcellulose acetate succinate (HPMCAS, MF grade). M1·M2·M3·M4·A1 The earlier finding from the same group, that 50% DL ritonavir dispersions in PVP or PVPVA phase separate within 10 s of contact with water, made ritonavir a natural test case (Purohit 2015). R20

Background concepts

Why water splits a drug–polymer mixture. The authors use the Flory–Huggins free energy of mixing for a water–drug–polymer system (their eq. 2):

Here $\Delta G_{mix}$ 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 water ($w$), drug ($D$) and polymer ($p$), and each $\chi$ is a Flory–Huggins interaction parameter for one pair of components. E2 In plain terms: the first three terms are the entropy gained by mixing, which always favours one phase, and the last three are the pairwise interactions, which can oppose mixing when a $\chi$ is large and positive. E2·C43 The authors note that a hydrophobic drug and water have a large positive ${\chi }_{WD}$, while a hydrophilic polymer and water, which can hydrogen bond, have a much smaller ${\chi }_{WP}$. C43 The difference between the two drives separation:

In plain terms: the larger the gap between how much water dislikes the drug and how much it likes the polymer, the stronger the tendency of the wet mixture to separate; this is the "Δχ effect" known from polymer solutions (Robard 1977). E4·C43 The paper states this relation in words, not as a printed equation. E4 The authors describe AAPS as "a classic case of solvent-induced immiscibility": the polymer and the sorbed water form a co-solvent in which the drug's miscibility limit is exceeded, so drug-rich domains form, provided the wet matrix is mobile enough. C44

Amorphous solubility and LLPS. The amorphous solubility is the highest concentration of free drug that a solution can hold before a second, drug-rich liquid phase appears; above it, the excess drug separates as droplets and the free concentration stays at the amorphous solubility. C10·C47 A practical consequence for measurement is that a UV probe in a solution that contains droplets reads an "apparent concentration", because the droplets absorb and scatter light; the free drug has to be measured after removing the droplets, for example by ultracentrifugation (Van Eerdenbrugh 2011). C15

Environment-sensitive fluorescent probes. Two probes do much of the work in this paper. Pyrene has five vibronic emission peaks, and the ratio of the first to the third peak (I/III) is higher in polar surroundings and lower in hydrophobic ones (Kalyanasundaram and Thomas 1977). C18 In solution, pyrene moves into drug-rich droplets once they form, so a drop in I/III marks LLPS (Raina 2015); in a film, a drop in I/III marks pyrene moving into drug-rich domains. C18·C20 Nile red is also hydrophobic; its emission is brighter and blue-shifted in hydrophobic surroundings, so in confocal images drug-rich domains appear brighter than polymer-rich regions (Greenspan 1985; Sackett 1987). C24

Polymer-controlled and drug-controlled dissolution. At low loading the polymer dissolves and carries the molecularly dispersed drug with it, so drug and polymer leave at the same fractional rate (congruent release) and the solution can overshoot the amorphous solubility. C49 At high loading, where the solid that remains is amorphous drug, the authors rationalise the outcome with the Noyes–Whitney equation (printed with the label "(2)" but called eq. 3 in the text):

Here $x$ is the drug concentration at time $t$, $S$ the solubility of the amorphous drug and $C$ a constant. E3·X2 In plain terms: the dissolution rate falls to zero as the concentration approaches the solubility of the dissolving solid, so a solid made of amorphous drug can bring the solution up to the amorphous solubility but not beyond it, and LLPS cannot occur. C50

Sizing particles in solution. Nanoparticle tracking analysis (NTA) follows the Brownian motion of individual scattering particles and converts their diffusion coefficient into a size with the Stokes–Einstein equation (eq. 1):

Here $D$ is the diffusion coefficient, $k_B$ Boltzmann's constant, $T$ the absolute temperature, $\eta$ the dynamic viscosity and $r$ the particle radius. E1 In plain terms: smaller particles diffuse faster, so measuring how fast each particle wanders gives its size. E1

Methods in detail

The study has three parts that are later laid side by side: a measurement of the ceiling (the amorphous solubility), a survey of what water does to the solid dispersions, and a dissolution study that records drug release, polymer release and the presence of nanodroplets for fresh and humidity-exposed films. A1·A2·A3 All dispersions were made at 10, 20, 30, 40 and 50% DL (RTV:polymer 10:90 to 50:50 by weight) by rotary evaporation from methanol at 45 °C. P4·P5 Bulk powders were vacuum-dried for at least 24 h and sieved to 150–250 µm; dissolution films were cast the same way on the inner wall of 20 mL glass vials; thin films for fluorescence, confocal imaging and AFM were spin-coated from methanol onto quartz coverslips (with 0.2% w/w pyrene or nile red) or ZnS substrates. P4·P5·P6

Part 1: Measuring the ceiling three ways

Problem. Every later statement about supersaturation and LLPS needs a reliable amorphous solubility, and it must be known whether dissolved polymer changes it. C17 What they did. A 5 mg/mL methanolic ritonavir stock was pumped at 20 µL/min into 100 mM pH 6.8 phosphate buffer stirred at 300 rpm, and the extinction at 400 nm, where ritonavir does not absorb, was recorded every 10 s; the concentration at which scattering rises marks the formation of a drug-rich phase. P1·C13 The same additions were made into buffer containing 0.4 ppm pyrene, recording I/III after each addition. P2 In the third method a solution brought to about 40 µg/mL, visibly cloudy, was ultracentrifuged at 35000 rpm for 30 min and the clear supernatant was measured by UV. P3 Each test ran in buffer alone and with 512 µg/mL PVP, PVPVA or HPMCAS, all at 25 °C. P1·P3 Why three methods. Dynamic addition can overshoot the amorphous solubility before the new phase is detected, whereas separating the phases of a solution that has already undergone LLPS gives a value "potentially" closer to the equilibrium co-existence concentration. C19 The authors chose 512 µg/mL as the most polymer a 10:90 dispersion can release; at the dissolution dose used later, a 10:90 film actually releases 540 µg/mL, so the tested level is about 95% of that maximum. C14·X3

Part 2: What water does to the solid

Problem. Before linking phase separation to dissolution, the authors needed to know which dispersions phase separate, how fast, and what the separated structure looks like. They combined four techniques because none answers all three questions: fluorescence and DSC report whether separation happened but not the microstructure, AFM shows structure but cannot tell which phase is drug-rich, and confocal imaging identifies drug-rich domains only down to about 300 nm. C28·C25 What they did. Humidity was set to 97% RH over saturated potassium sulfate in a sealed jar. P7 Pyrene-containing films were measured at intervals up to 96 h, with films of pure amorphous ritonavir and of each pure polymer as end points. P8·R2 Films with nile red were imaged by confocal microscopy (excitation 561 nm, emission 575–620 nm) before and after humidity exposure; the exposure time for these images is not stated. P9·X7 AFM topography (contact mode, 256 × 256 points, 0.5 Hz) was recorded on 10:90 films before and after 4 days at 97% RH. P10 For DSC, powders were weighed into pans, held at 97% RH for 24 h, vacuum-dried for 12 h and run in modulated mode (2 °C/min, ±1 °C every 60 s), so the traces show the phase structure that humidity left behind rather than a wet sample. P11 To mimic the first seconds of dissolution, 10:90 and 50:50 films were dipped in buffer for 10 s, blown dry, vacuum-dried overnight and examined by confocal microscopy and by SEM. P9·P13

Part 3: Dissolution and the three LLPS read-outs

Problem. Whether a dispersion reached LLPS cannot be read from the UV trace alone, because droplets distort it. C15 What they did. Each film vial received 18 mL of 100 mM pH 6.8 phosphate buffer at 25 °C; the ritonavir dose was fixed so that complete release gives 60 µg/mL, which means complete polymer release gives 540 µg/mL from a 10:90 film and 60 µg/mL from a 50:50 film. P14·R15 Ritonavir was followed in situ by a UV dip probe at 240 nm with the 400 nm baseline subtracted to correct for scattering, for about 4 h. P14 Polymer in filtered aliquots was assayed colorimetrically: HPMCAS by the phenol–sulfuric acid method at 490 nm (linear 5–35 µg/mL), PVP and PVPVA by potassium triiodide at 500 nm (linear 5–40 µg/mL). P19·P20 Films were dissolved fresh and after storage at 97% RH; the NTA section states 4 days of storage, while the storage time for the dissolution profiles of Figure 9 is not printed. P7·P12·X7 At the end of a run, LLPS was judged in three independent ways: free ritonavir in the ultracentrifuged supernatant (fresh films, Table II), pyrene I/III in the 1 µm-filtered solution (Table III; the filter passes droplets of about 250 nm but not undissolved solid), and the presence and size of scattering particles in NTA (Figure 7). P16·P17·P12·C15 Polarized light microscopy checked solution and residue for crystals. P18 Powders (100 mL of buffer, same dose) were also dissolved, but those data are not shown. P15·R19

Experimental analysis

The ceiling is about 27–28 µg/mL, and the polymers do not move it

All three methods placed the amorphous solubility of ritonavir in pH 6.8 buffer at 25 °C between 26.8 and 28.5 µg/mL, with a standard deviation of 0.5–2 µg/mL (Table I). R1 The extinction rose sharply from 27 µg/mL upward, pyrene I/III changed suddenly at about 28 µg/mL, and ultracentrifuged supernatants held about 27 µg/mL. R1·C16·C18·C19 With 512 µg/mL of polymer the values stayed within the same range, and the authors conclude that at this concentration none of the polymers has a solubility-enhancing effect ("It is evident"). R1·C17

Table A. Amorphous solubility of ritonavir (µg/mL) in 100 mM pH 6.8 phosphate buffer at 25 °C, alone or with 512 µg/mL polymer (from Table I; SD 0.5–2 µg/mL). The fluorescence method is described for polymer-containing buffer too, but no value is reported. R1·X14

Humidity splits the PVP and PVPVA dispersions, not the HPMCAS ones

Before exposure, the pyrene I/III ratio of every film lay between the values of the pure components (amorphous ritonavir about 1.27–1.30; PVP about 1.83, PVPVA about 1.75, HPMCAS about 1.52) and rose with polymer content, as expected for miscible films. R2·C21 At 97% RH the ratio of PVP and PVPVA films fell toward the value of pure ritonavir, while that of HPMCAS films hardly moved (Figure 2, Table B). R2 The fall was fastest at high loading: the 50:50 PVP film dropped from about 1.50 to about 1.30 within the first hour, while the 10:90 films declined slowly and reached only about 1.35 (PVP) and about 1.45 (PVPVA) after 96 h. R2.a·R2.e·R2.l The authors read the fall as pyrene partitioning into drug-rich domains formed by moisture-induced AAPS, with the value of about 1.30 at the end of day 3 meaning that pyrene is "predominantly surrounded by RTV molecules" in all PVP and PVPVA films except the 10% DL ones. C22 In the discussion they add that all PVP and PVPVA dispersions start to phase separate "as early as 1 h"; the data support that for 20–50% DL but not for the 10% DL films. C22·X9

Table B. Pyrene I/III in films at 97% RH, read from Figure 2 (approximate). The time axis of Figure 2 is broken after 6 h; the first point after the break is unlabelled and lies near 24 h. Pure amorphous ritonavir reads about 1.27–1.30. R2

The images agree. After humidity exposure, confocal images of PVP and PVPVA films showed bright, nile red-rich domains: densely packed domains about 1 µm across in 10:90 PVP, discrete round domains of about 1–3 µm in 10:90 PVPVA, and interconnected domains in both 50:50 films, while both HPMCAS films stayed uniformly fluorescent (Figure 3). R3 Because nile red is brighter in drug-rich surroundings, the authors take the discrete domains to be ritonavir-rich ("the images confirm"), and the uniform HPMCAS films to have no phase separation detectable at the confocal resolution of about 300 nm. C24·C25 AFM of 10:90 films after 4 days at 97% RH showed raised domains on PVP and PVPVA films, which the discussion sizes at 5–10 µm, against featureless surfaces before exposure; the HPMCAS film showed no such domains. R5·C27 In the results the authors are careful ("This may indicate"), because AFM height contrast does not say which phase is drug-rich. C27·C28

DSC of the humidity-exposed and re-dried powders gave the third line of evidence (Figure 6). Every PVP and PVPVA trace showed a step at about 47–50 °C, close to the 45 °C midpoint Tg of amorphous ritonavir, and the PVPVA traces showed a second step at about 99–108 °C from the polymer-rich phase; HPMCAS dispersions showed a single step that rose from about 63 °C at 50:50 to about 97 °C at 10:90 and, by the text, did not change with exposure. R6·M1 All fresh dispersions had a single Tg (data in the supplementary material, not supplied). R6 The authors read two Tgs as "consistent with" moisture-induced AAPS and the low first Tg as a sign that phase separation "appears to have occurred" extensively, and the single HPMCAS Tg as miscibility "at least within the resolution of the DSC measurement". C29 Note by the AI writer (not discussed in the paper): in the PVP traces only the lower step is distinct in all five loadings; a higher step is visible only at 20:80, so for PVP the DSC evidence rests on the ritonavir-rich step. X10

In buffer, the 50% DL PVP and PVPVA films split within seconds

After 10 s in buffer and drying, the 10:90 PVP film was uniformly fluorescent while the 50:50 PVP film had separated into dark domains of about 1–3 µm in a bright drug-rich phase (Figure 4). R4 The text reports the same behaviour for PVPVA films and uniform images for 10% and 50% DL HPMCAS films, which are not shown. R4·C26 SEM after the same treatment agreed: the 50:50 PVP surface was a rough porous network and the 50:50 PVPVA surface granular, while all three 10:90 films and the 50:50 HPMCAS film were smooth (Figure 8). R9·C31 The results paragraph introduces Figure 8 as films "after exposure to 97% RH", but the caption, the methods and the discussion all describe 10 s in buffer, which is how the figure should be read. X1 The 20–40% DL films were not imaged after contact with buffer. A2

Only some fresh dispersions pass the ceiling

Three independent read-outs at the end of dissolution pick out the same group of dispersions (Table C). R7·R10·R11 Free ritonavir after 4 h was 27–30 µg/mL, at the amorphous solubility, for every HPMCAS dispersion and for 10:90 and 20:80 PVP and 10:90 PVPVA, but 18–25 µg/mL for the other PVP dispersions and 10–22 µg/mL for the other PVPVA dispersions. R10 Pyrene I/III in the filtered solutions was 1.50–1.66 for the same group and 1.88–1.90 for the rest, against 1.90 in buffer. R11 NTA found nanodroplets for all HPMCAS dispersions and only the 10:90 PVP, 20:80 PVP and 10:90 PVPVA films, with mean sizes of 200 nm (PVPVA), 250 nm (HPMCAS) and 315 nm (PVP) for the fresh 10:90 films. R7·R8·C30 The authors read the I/III decrease as pyrene entering ritonavir-rich nanodroplets ("indicates"), and where there was no significant decrease "it can be assumed that LLPS did not occur". C32

Table C. Free ritonavir after 4 h of dissolution of fresh films (µg/mL, from Table II; SD less than 2 µg/mL) and pyrene I/III of the final solutions (from Table III; 1.90 in buffer). Table III does not state whether its solutions came from fresh or humidity-exposed films. R10·R11·X6

The dissolution profiles show the same split (Figure 9). Fresh LLPS-forming films spiked within minutes to apparent concentrations far above the ceiling: about 130 µg/mL (line maximum) for 10:90 PVP, about 80 for 20:80 PVP, about 116 for 10:90 PVPVA, and about 119, 67 and 49 µg/mL for 10:90, 20:80 and 30:70 HPMCAS, while 40:60 and 50:50 HPMCAS levelled off at about 30–33 µg/mL. R12·R13·R14 The other PVP and PVPVA films rose slowly and stayed below the ceiling, at about 9–26 µg/mL after 240 min. R12·R13 The spikes then decayed, for example to about 17 µg/mL for 10:90 PVP and about 21 µg/mL for 10:90 HPMCAS at 240 min, although the free drug stayed at about 28 µg/mL. R12·R14·R10 No crystals were found by polarized light microscopy in any solution or residue, and the authors explain the decay as an artefact of the UV method: nanodroplets "can absorb and scatter UV light and can also aggregate and/or coalesce with time". R18·C39

The polymer leaves early; the drug keeps pace only where nanodroplets form

Polymer release was measured for the 10:90 and 50:50 films of each polymer (Figure 10, Table D). R15·R16·R17 From the 10:90 films the polymer came out within minutes (about 450–500 of 540 µg/mL PVP by 5–10 min, about 565 µg/mL PVPVA and about 550 µg/mL HPMCAS by about 5 min) and ritonavir came with it, to apparent concentrations of about 83–116 µg/mL. R15·R16·R17 From the 50:50 films the picture splits by polymer. PVP reached about 57 of 60 µg/mL within 5 min while ritonavir crept to about 18 µg/mL; PVPVA dissolved more slowly, about 22 µg/mL at 5 min and about 45 at 60 min, reaching full release near 120 min, while ritonavir stayed at about 9 µg/mL; HPMCAS reached about 53 µg/mL by 20 min and ritonavir rose with it to about 32–34 µg/mL. R15.b·R15.d·R16.f·R16.h·R17.j·R17.l The authors call the HPMCAS release "congruent in the initial stages of dissolution even for 50% DL", "which can explain" why HPMCAS dispersions reached LLPS at high loading. C38

Table D. Polymer and ritonavir concentrations (µg/mL) during dissolution of fresh films, read from Figure 10 (approximate). "Full" gives the polymer / drug concentration at complete release. Drug values above about 28 µg/mL are apparent concentrations that include nanodroplets. R15·R16·R17

Two details in Figure 10 do not fit the text exactly. The discussion says that the polymer "in all instances dissolves almost completely in less than 10 min", and the results that all polymers dissolve "within a few minutes"; the 50:50 PVPVA film needed about 2 h. C40·C52·X4 And the 10:90 PVPVA film read about 565–575 µg/mL, a few percent above the theoretical 540 µg/mL, which gives a sense of the accuracy of the polymer assay. X15 Neither changes the qualitative point that the polymer leaves well ahead of the drug from the 50:50 PVP and PVPVA films.

After humid storage, PVP and PVPVA lose their nanodroplets; HPMCAS keeps them

Storing the films at 97% RH before dissolution changed the PVP and PVPVA results most at low loading (Figure 9, Table E). R12·R13 The 10:90 PVP film no longer spiked; it reached about 25 µg/mL after about 25 min and about 30 µg/mL at 240 min, and the 10:90 PVPVA film rose to only about 25–26 µg/mL. R12.j·R13.t All exposed PVP and PVPVA films stayed at or below the ceiling, the higher loadings changing little, and NTA found no nanodroplets from any of them. R12·R13·R7·C33·C35 The HPMCAS films, by contrast, still spiked after exposure, to about 131, 124 and 90 µg/mL for 10:90, 20:80 and 30:70, and still produced nanodroplets. R14·R7 The authors conclude, "very clearly", that a phase-separated matrix dissolves more slowly and no longer produces colloidal species. C53

Table E. Apparent ritonavir concentration (µg/mL) of fresh and humidity-exposed films, read from Figure 9 (approximate). The storage time before these runs is not stated. R12·R13·R14·X7

Note by the AI writer (not discussed in the paper): the authors state that for HPMCAS "exposure to high RH did not impact the dissolution kinetics". C37 That holds for the fast initial rise and for the LLPS outcome, but the exposed HPMCAS films reached markedly higher apparent peaks at 20:80 and 30:70 and stayed higher at 240 min, as Table E shows. X5 Because these are apparent concentrations, the difference may lie in the droplets (number or size) rather than in the free drug, which was not measured for exposed films. X5·R10

Controls and side observations

No crystallization was seen in any experiment, so every decrease in concentration and every phase-separated structure in this paper is amorphous. R18 Powdered dispersions gave LLPS for the same drug–polymer combinations as films, with rate differences that the authors attribute to surface area; the data are not shown. R19·C41 The highest-loading HPMCAS film is worth noting: at 50:50 it gave an apparent plateau of about 33 µg/mL, free ritonavir of 27 µg/mL and a pyrene I/III of 1.61, the basis of the authors' statement that all HPMCAS dispersions reached LLPS. R14.u·R10.a·R11.c·C4

Discussion

Why HPMCAS resists the split

The authors explain the polymer ranking with two properties that they take from the structures and from earlier work rather than measure here. PVP and PVPVA are more hydrophilic and hygroscopic than HPMCAS, so their ritonavir dispersions "can be expected" to be more prone to moisture-induced AAPS, a pattern the group had seen with other hydrophobic drugs (Rumondor 2009, 2010; Purohit 2015). C45 PVP and PVPVA carry hydrogen-bond acceptors only, whereas HPMCAS carries donors and acceptors, so its interactions with the donor and acceptor groups of ritonavir "can be expected to be more numerous/diverse"; together with the higher hydrophilicity of PVP and PVPVA, these are "expected to result in a higher Δχ value" for the pyrrolidone polymers. C46·M2·M3·M4 The observation itself, that only PVP and PVPVA dispersions split, is supported by four techniques; the explanation is not, because water uptake, hydrogen bonding and interaction parameters were not measured. C1·C46

A race between dissolving and splitting

The central interpretation is a competition between two rates. If the wet matrix dissolves faster than it phase separates, the polymer carries the drug into solution, the concentration passes the amorphous solubility and the solution undergoes LLPS; if the matrix phase separates first, the polymer leaves on its own, the solid that remains is mostly amorphous drug, release becomes drug-controlled and, by Noyes–Whitney, cannot exceed the amorphous solubility. C5·C50·C54 The authors put it in hedged form in the discussion ("appears to depend on the competition between the rate of dissolution and the rate of phase separation") and more firmly in the abstract ("is governed by"). C5 The chain below follows their argument.

The evidence for each branch is good at the end points: the three LLPS read-outs agree, and humidity storage switches PVP and PVPVA films from the good to the bad branch without crystallization. R7·R10·R11·R18·C53 What is not measured is the race itself. Matrix phase separation was observed after 10 s in buffer and after hours to days in humid air, and release was measured over minutes to hours, but no experiment followed both processes on the same dissolving film. R4·R9·R2·R12·X11 Polymer release, the evidence that the matrix switched from polymer to drug control, was shown for the fresh 10:90 and 50:50 films only, not for the intermediate loadings or for any humidity-exposed film. X12 The step "the matrix split, so release became drug-controlled" is therefore an inference for the humidity-exposed films, consistent with the data but not demonstrated by a polymer-release measurement. C54·X12

Where the nanodroplets come from

The paper's second contribution is its answer to the Harmon proposal. Colloidal species formed only where drug and polymer were released congruently; they did not form where the polymer dissolved much faster than the drug. C52 If nanodroplets formed in the matrix by AAPS and were then released, the authors argue, the low-loading films, which give the most droplets, should show nanoscale separation after contact with water; instead the 10:90 films were uniform after 10 s in buffer by confocal imaging and SEM, while the 50:50 films, which give no droplets, showed extensive phase separation within seconds. C56·R4·R9 They add that the domains formed in the matrix are microns across, whereas the colloidal species in solution are an order of magnitude smaller (200–315 nm), so the release of matrix domains "cannot account for" them. C56·R8·R5 Their conclusion is hedged: they "find no evidence" for the matrix route, and the nanodroplets are "likely" formed by rapid dissolution of a miscible matrix followed by LLPS. C56·C57

Note by the AI writer (not discussed in the paper): the argument against the matrix route depends on what the imaging can see. The droplets average 200–315 nm, which is at or below the confocal resolution of about 300 nm that the authors give for these images, and SEM shows the dried surface of a film after the water has been blown off. R8·C25·X11 Drug-rich domains of a few hundred nanometres inside a low-loading matrix would therefore probably not have been resolved, so the matrix route is argued against rather than excluded at the nanoscale. X11·C51

What it suggests for formulation

The authors draw one recommendation: "formulation scientists should design ASDs that are resistant to AAPS and exhibit polymer controlled dissolution". C58 Their data add two practical points. Humid storage can remove the supersaturation advantage of a PVP or PVPVA dispersion of a hydrophobic drug without any crystallization: the 10:90 films lost both their spike and their nanodroplets. R12·R13·R7·R18 And the polymer sets how much drug a dispersion can carry before it stops reaching LLPS: here up to 20% DL with PVP, 10% with PVPVA, and at least 50% with HPMCAS. R7·R10·R11 Note by the AI writer (not discussed in the paper): the PVP and PVPVA thresholds are the behaviour that later work from the same field calls a limit of congruency; this paper does not use that term, and all of its tests ran at 25 °C in phosphate buffer, so the thresholds may differ under physiological conditions. X18

Limitations

Limits the authors state. The authors note that fluorescence spectroscopy and DSC "provide no information about the resultant microstructure", that AFM topography cannot distinguish drug-rich from polymer-rich phases, that DSC detects miscibility only "within the resolution of the DSC measurement", and that confocal imaging cannot detect phase separation below about 300 nm. C28·C29·C25 They also state that the apparent UV concentration is not the free-drug concentration once droplets are present, which is why they measured free drug after ultracentrifugation. C15·C39

The AI writer's own judgement (1): what is measured and what is inferred. The strongest results are the observations: water splits the PVP and PVPVA dispersions and not the HPMCAS ones (four techniques), and only dispersions that stayed miscible reached LLPS (three read-outs, fresh and humidity-exposed films). R2·R3·R5·R6·R7·R10·R11 The mechanism that links them, a race between two rates, is an interpretation; neither rate was measured on a dissolving film, and polymer release, the direct sign of a switch to drug control, was shown only for fresh 10:90 and 50:50 films. C5·X12 The explanation of the polymer ranking (hygroscopicity, hydrogen bonding, Δχ) was argued from structures and earlier work, without water-uptake, spectroscopic or χ data. C46

The AI writer's own judgement (2): statements stronger than the data, and unstated conditions. Several sentences in the paper go beyond what its figures show: phase separation "as early as 1 h" holds for 20–50% DL films but the 10% DL films changed slowly X9; the polymer dissolving "in less than 10 min" in all instances does not hold for 50:50 PVPVA, which took about 2 h X4; humidity "did not impact" the HPMCAS dissolution kinetics, yet the exposed 20:80 and 30:70 films reached much higher apparent concentrations X5; and two Tgs are claimed for the exposed PVP dispersions, where only one step is distinct in most traces X10. Some conditions are missing or inconsistent: the storage time at 97% RH before the dissolution runs and before confocal imaging X7, whether the Table III solutions came from fresh or exposed films X6, the loading of the NTA images X8, and the exposure used for SEM, which the results text gives as 97% RH and the caption as 10 s in buffer X1. Replicate numbers are not stated for any test; only Tables I and II give standard deviations. R1·R10

Conclusion

The paper makes three contributions. First, it shows with pyrene fluorescence, confocal imaging, AFM and DSC that absorbed water splits ritonavir dispersions in PVP and PVPVA into drug-rich and polymer-rich amorphous phases, while HPMCAS dispersions stay miscible, and that at 50% DL the PVP and PVPVA films split within 10 s of contact with buffer. C1·R2·R3·R5·R6·R4·R9 Second, it maps which dispersions reach LLPS on dissolution, using free drug, pyrene in solution and NTA together: every HPMCAS dispersion, PVP up to 20% DL and PVPVA at 10% DL when fresh, and no PVP or PVPVA dispersion after humid storage. R7·R10·R11·C30 Third, it argues that dissolution depends on a race between matrix dissolution and matrix phase separation, and that the nanodroplets form by LLPS from a supersaturated solution rather than by release of matrix domains, in the authors' words "likely caused by rapid dissolution of miscible ASDs". C5·C57

The first two contributions are direct observations and hold well for this drug, these polymers and these conditions (films, pH 6.8 phosphate buffer, 25 °C). R7·R10·R11·X18 The third is an interpretation that fits the end points but rests on rates that were not measured and on imaging at about the size of the droplets it seeks to exclude. C5·X11 For formulation, the paper's message is that a dispersion only delivers supersaturation if it stays mixed long enough to dissolve under polymer control, and that humidity can quietly remove that ability before any crystal appears. C58·C53·R18