When drug and polymer leave together: congruent release, drug-rich droplets and the loading cliff in PVPVA dispersions of nilvadipine and cilnidipine

Background and motivation

An amorphous solid dispersion (ASD) disperses a poorly water-soluble drug, in its amorphous form, in a hydrophilic polymer. Its dissolution advantage is twofold: the amorphous drug dissolves to a supersaturated concentration, and a polymer that inhibits crystallization keeps that supersaturation for a useful time (Brouwers et al., ref 1). C2 More recently it has been noticed that some ASDs, typically those with a low drug loading, dissolve to a concentration above the drug's amorphous solubility. The solution then undergoes liquid–liquid phase separation (LLPS): a colloidal, amorphous drug-rich phase appears, while the surrounding water holds drug at the amorphous solubility (refs 2–4). In vivo this colloidal phase is thought to act as a reservoir: as dissolved drug is absorbed, the colloids redissolve and keep the free drug, and hence the membrane flux, at its maximum until they are used up (refs 5–8). C1

The obstacle the authors set out to address is what they call the drug:polymer ratio paradigm. Good release and exposure often need a low drug loading, but a low loading makes the dosage form large, which hurts patient compliance and can be a marketing disadvantage (refs 2, 3, 10, 11). C3 Earlier work described two dissolution regimes: release controlled by the polymer at low loading (refs 12–15) and by the drug at high loading (refs 16, 17). The authors call this picture "likely" oversimplified and note that little is known about what decides the switch between the two regimes, in particular at what loading it happens for a given drug–polymer pair. Studies of polymer type and of drug–polymer interactions or phase behaviour gave drug-specific results (refs 18–22), and a study of several factors at once (ref 23, by Chen and co-workers, which the introduction calls "Qian et al.") found no clear trends. C4

The authors pick one candidate factor: the difference in hydrophobicity between drug and polymer. Poorly soluble drugs are lipophilic; the polymers used in ASDs are hydrophilic. Their reasoning was that the hydrophilicity of a drug–polymer blend depends on the properties and amounts of both components, so a more hydrophobic drug should erode the blend's hydrophilicity faster as loading rises, and release should fail at a lower loading. C6 To test this they chose two dihydropyridines, nilvadipine (Nil, ClogP 3.04) and cilnidipine (Cil, ClogP 5.54). The two are structural analogues of the same therapeutic class, are non-ionizing and crystallize slowly, and the authors assume they hydrogen bond with the polymer to a similar extent, so that log P is the main difference. M1·M2·C7 The polymer was poly(vinylpyrrolidone-co-vinyl acetate) (PVPVA, Kollidon VA 64), a hydrophilic polymer widely used in commercial ASDs made by hot-melt extrusion and spray drying. M3

The paper states two hypotheses. First, that congruent (simultaneous) release of drug and polymer from PVPVA dispersions gives superior dissolution. Second, that the dispersion of the more hydrophobic drug releases congruently over a narrower range of loadings. C5·C6 To test them the authors measured drug and polymer release separately from compacts of constant area, checked the solution for drug-rich droplets, and examined partially dissolved compacts by imaging and surface analysis.

Background concepts

Congruent and incongruent release. In a dispersion, drug and polymer are mixed at the molecular level. When the compact dissolves, the two can leave at the same rate relative to their content (congruent release) or at different rates (incongruent release). The paper expresses each component's release rate per unit area and normalized to that component's content, so a congruent dispersion has a polymer/drug rate ratio of about 1. P9·E4·R5 Congruent release means the drug is carried out at the rate of the hydrophilic polymer, which is fast; in the authors' terms it is "polymer-controlled". Incongruent release, with the polymer leaving much faster, means the drug is left behind and its release approaches that of the amorphous drug on its own, which the authors call "drug-controlled". C11·C12

Amorphous solubility and LLPS. The amorphous solubility is the highest concentration of dissolved drug that a supersaturated solution can hold before a second, drug-rich liquid phase appears. Above it, LLPS produces drug-rich droplets a few hundred nanometres across. Below it, no droplets can form. The ratio of amorphous to crystalline solubility gives the highest supersaturation the amorphous form can reach. E3·R1

Amorphous–amorphous phase separation. A dispersion that is one phase when dry can split into a drug-rich and a polymer-rich amorphous phase when it takes up water, because water is a poor solvent for the drug and a good one for the polymer. This water-induced amorphous–amorphous phase separation (AAPS) is known to occur in some ASDs at high relative humidity (RH) (refs 26–28). C15 The question in this paper is whether it also happens at the surface of a dissolving compact, and whether it controls release.

How the techniques read phase behaviour. The paper relies on several indirect probes, so it helps to know what each one reports.

• 4-Di-2-Asp fluorescence. This dye's emission depends on the polarity of its surroundings. In buffer it emits at a longer wavelength; when drug-rich droplets are present, it partitions into them and its emission shifts to shorter wavelengths (a blue shift). The authors used it both to find the amorphous solubility and to detect droplets in dissolution media (ref 25). P4·P12·C9
• Prodan and rhodamine 6G. In confocal fluorescence microscopy (CFM) of films, the less polar prodan stains drug-rich domains (shown blue) and the more polar rhodamine 6G (R6G) stains polymer-rich domains (shown red), following the group's earlier method (ref 26). P18·C9
• DLS and NTA. Dynamic light scattering gives a size distribution of particles in the medium; nanoparticle tracking analysis images individual scattering particles and sizes them from their diffusion (Stokes–Einstein). P12·E5
• EDX N/O ratio. Nilvadipine (C₁₉H₁₉N₃O₆) has a nitrogen-to-oxygen atom ratio of 0.5; the vinylpyrrolidone–vinyl acetate repeat pair of PVPVA (C₁₀H₁₅NO₃) has about 0.3. The local N/O ratio measured by energy-dispersive X-ray analysis therefore indicates how drug-rich a spot is. For cilnidipine the difference from PVPVA is too small to use. E2·M1·M3·P16
• XPS. X-ray photoelectron spectroscopy samples the top 10 nm or so; fitting the carbon 1s peak with the shapes of the pure components gives the surface drug fraction. P17
• Micro-CT. X-ray micro-computed tomography gives density cross-sections of a whole compact at 3 µm pixels, showing pores below the surface. P14

Methods in detail

The study combines five experimental strands that all hinge on the same set of dispersions. The flow below shows how they connect.

Solubility benchmarks

All solubility and dissolution work used 100 mM phosphate buffer, pH 6.8, at 37 °C. M4 Crystalline solubility was measured by stirring excess drug for 48 h, removing the solid by ultracentrifugation (35,000 rpm, 30 min) and assaying the supernatant by reverse-phase HPLC. P1·P2 The amorphous solubility of nilvadipine was measured four ways, each with 50 µg/mL PVPVA to stop crystallization and again with 1 mg/mL to see whether more polymer changes the value: the drug concentration at which light scattering at 450 nm starts to rise while a methanol stock is pumped in (UV extinction); the inflection of the 4-Di-2-Asp emission versus drug concentration; the drug concentration above which mass flow across a 6–8 kDa regenerated cellulose membrane stops increasing (mass flow); and the drug left in the supernatant after ultracentrifuging a supersaturated solution. P3·P4·P5·P6 For cilnidipine, whose amorphous solubility is very low, only ultracentrifugation was used; it reproduced a value of 0.5 ± 0.3 µg/mL reported earlier (ref 24), with and without polymer. P6·R2

Dispersions, films and compacts

Nilvadipine dispersions were made at 5, 10, 15 and 20% w/w and cilnidipine dispersions additionally at 25%, by dissolving drug and PVPVA in 1:1 methanol/dichloromethane (50 mg/mL solids) and rotary-evaporating at 45 °C. The powders were vacuum-dried, cryo-milled and sieved to 106–250 µm, and all were free of crystals by powder X-ray diffraction (PXRD) and polarized light microscopy (PLM). P7·R3 Thin 10:90 nilvadipine films for confocal imaging were spin-coated on quartz in a dry nitrogen glovebox (below 20% RH) with prodan and R6G at 0.01% w/v each. P8

Separate drug and polymer release from a constant-area compact

This is the central measurement. 100 mg of powder was pressed in an 8 mm die at 1500 psi for 1 min, and the die was mounted on a stirrer shaft so that only one face, 0.50 cm², dissolved in 100 mL of buffer at 100 rpm and 37 °C (a Wood's intrinsic dissolution apparatus). P9 Conditions were non-sink with respect to the amorphous solubility, so the medium could become supersaturated. Samples of 2 mL were taken over 2 h and replaced with fresh buffer: 100 µL went straight to HPLC for the drug, and the rest was filtered (0.2 µm nylon) for PVPVA by size-exclusion chromatography with UV detection at 205 nm. P9·P10 The release rate of each component came from a linear fit to the initial linear part of its % release versus time, normalized to its content in the compact, n = 3. The authors assume the exposed area stays constant, with fresh compact surfacing as material dissolves. E4·C8 Neat PVPVA and neat amorphous nilvadipine compacts served as the two limits; how these neat compacts were prepared is not described. R5·X12 Compacts were checked for crystallization by PXRD and PLM during and after the runs. P11

The authors also express drug loading in mol%, "normalized to account for the number of moles of each component", using an average monomer molecular weight for PVPVA. No formula or molecular weights are printed. E1

What is in the solution

At about 1 h of compact dissolution, the medium was examined three ways: 4-Di-2-Asp fluorescence (1 µg/mL probe, excitation 488 nm), DLS (volume median diameter d₅₀ and polydispersity index) and NTA (30 s videos at 37 °C). P12 Powder dissolution was run separately: powders equivalent to 100 µg/mL drug were stirred at 300 rpm in 50 mL buffer, nilvadipine was followed with an in situ UV probe, and cilnidipine by filtered samples (0.45 µm PTFE), because its suspensions scattered too much light for the probe. P19

What happens at the surface

Compacts on either side of the switch, 10% and 15% nilvadipine and 15% and 20% cilnidipine, were stopped after 10 min of dissolution. Micro-CT gave cross-sections; the 15% nilvadipine compact was also imaged at 20 and 30 min. P14 Field-emission SEM imaged the dissolving face, and a fractured 20% cilnidipine compact showed the region below it. P15 On a 15% nilvadipine compact, EDX measured the N/O ratio at four points inside a pit, four on its raised rim, and four on an undissolved compact. P16 XPS measured surface composition before and after 10 min for the two incongruent compacts (15% nilvadipine, 20% cilnidipine) and, as controls, two congruent ones (10% nilvadipine, 15% cilnidipine), with 3–5 spots per sample. P17

Inducing phase separation on purpose

To ask whether water-induced phase separation causes slow release, the authors took a congruently releasing system, a 10:90 nilvadipine compact, and held it in its die at 97% RH for 12 h before dissolution; they then imaged its surface by SEM. A matching 10:90 film was held at 97% RH for 12 h and imaged by CFM. P13·P18 How the 97% RH was generated is not stated.

Experimental analysis

Solubility: the bar the drug must clear

Crystalline solubility was 1.7 µg/mL for nilvadipine and 0.05 µg/mL for cilnidipine. R1.a The four methods gave nilvadipine amorphous solubilities of 31.9 ± 2.8 (UV extinction), about 30 (fluorescence), about 30 (mass flow) and 31.2 ± 3.3 µg/mL (ultracentrifugation), which the authors call "good agreement"; cilnidipine's was 0.6 ± 0.1 µg/mL. R1.c–f·C10 The amorphous solubility is therefore 18 times the crystalline solubility for nilvadipine and 12 times for cilnidipine. R1·E3 A higher PVPVA concentration, 1 mg/mL, left these values within the same ranges (data not shown). R2 The literature values for cilnidipine are attributed to ref 23 in the footnote of Table 1 but to ref 24 in the methods text. X1

The cliff in release rates

Table A collects the release rates. Neat PVPVA dissolved at 3.49 mg/min/cm² and neat amorphous nilvadipine at 0.002, which the authors put at about 1700 times slower; neat cilnidipine dissolved too slowly to quantify. R5.a·R5.f·R6.g·R7 Up to 10% nilvadipine and 15% cilnidipine the drug and polymer rates match, with ratios of about 1 and no significant difference in Figure 15. One step higher, the drug rate collapses while the polymer rate falls much less. R5·R6·R8

Table A. Surface-normalized release rates (mg/min/cm², normalized to each component's content; mean ± SD, n = 3), from Tables 2 and 3, with the t-test marks of Figure 15. R5·R6·R8

The release curves of Figure 2 show the same contrast directly. From the 10% nilvadipine compact, drug and PVPVA rise together to about 90–96% in 2 h; from the 15% compact PVPVA reaches about 39% while the drug reaches only about 2% (read from Figure 2, approximate). R4 The authors read equal rates as release controlled by the polymer ("This suggests"), and the 15% and 20% nilvadipine rates, "closer to the dissolution rate of the amorphous drug alone", as control by the amorphous drug ("indicating"). C11·C12 A small increase in loading, from 10 to 15 wt% nilvadipine, cut the drug release rate 60-fold. R7·C32 Note by the AI writer (not discussed in the paper): the authors call the cilnidipine fall "similar", but Table 3 gives 2.18 to 0.006 mg/min/cm², about 360-fold; and the "drug-controlled" nilvadipine rates of 0.035 and 0.013 are still about 17 and 6.5 times faster than the neat amorphous drug, a difference Figure 15 marks as significant. X8·X9·R8.c–d

Even congruently releasing dispersions released PVPVA more slowly than neat polymer: 2.35–2.76 against 3.49 mg/min/cm², marked significant in Figure 15 (p ≤ 0.05 to 0.01). The authors cite earlier observations of this (refs 2, 20) and say it is "most likely" due to the higher hydrophobicity of the formulation or to drug–polymer interactions. R8·C13 No crystals formed in any compact during the tests. R3

Log P did not move the switch

The switch lies between 10 and 15% w/w for nilvadipine and between 15 and 20% w/w for cilnidipine. The more hydrophobic drug therefore stayed congruent one step longer on a weight basis, the opposite of the second hypothesis. R5·R6·C6 On a molar basis, the paper places the switch between 6 and 9 mol% for nilvadipine and between 7 and 9 mol% for cilnidipine, "a similar drug loading". R7·C14 The authors conclude, "we clearly see that our hypothesis is incorrect for the systems evaluated". C35

Note by the AI writer (not discussed in the paper): the molar values cannot all be reproduced with one convention. With the printed formulas (C₁₉H₁₉N₃O₆, 385.4 g/mol, for nilvadipine and C₁₀H₁₅NO₃, 197.2 g/mol, for the PVPVA repeat pair), the nilvadipine values reproduce as moles of drug per mole of repeat pair (5.7% for 10:90, 9.0% for 15:85), while the cilnidipine values at 20:80 and 25:75 (9% and 12%) match mole fractions instead; an "average monomer" of half the pair would give only about 2.8% for 10:90. Under either convention the two switch ranges, about 5.4–9.0 and 6.6–10.0 mol%, still overlap, so the conclusion stands. X2·E1

To widen the comparison, the authors cite other dispersions: itraconazole (ClogP 5.4) in PVPVA dropped off above 20% loading (ref 2), ritonavir (ClogP 4.5) in PVPVA above 25% (ref 42), and sulfathiazole (ClogP −0.5) in PVP above 25% (ref 40), the last being "difficult to compare" because the polymer differs. R22 Taken together, they say, these "further support the lack of correlation between molecular hydrophobicity" and the loading at which release drops. C36 The highlights put it more strongly: the switch loading "is independent of drug logP for PVPVA ASDs". C37

Drug-rich droplets appear only with congruent release

Three techniques agree on where droplets form. In media from the lower-loading compacts the 4-Di-2-Asp emission maximum sat at about 560–570 nm, against about 593–595 nm in buffer and in media from the higher loadings (read from Figure 4, approximate); the cutoff lay between 10 and 15% for nilvadipine and between 15 and 20% for cilnidipine. R10 DLS found droplets of 237–366 nm, with polydispersity below 0.2, only at those lower loadings (Table B), and NTA showed scattering particles at exactly the same loadings and none above (Figures 5 and 6). R11·R12 The authors read the blue shift as the probe entering drug-rich aggregates and the scattering centres as "consistent with" amorphous nano-droplets. C18·C20

Table B. Droplet diameters by DLS (d₅₀, mean ± SD, n = 3; "none" means no meaningful data) from Table 4, and the 4-Di-2-Asp emission maximum read from Figure 4 (approximate; about 593 nm in buffer). R10·R11

Droplets grew with loading for cilnidipine (255, 300, 366 nm), which the authors say "could indicate" lower stability against agglomeration or an effect of loading on how droplets form, a question "under debate" (refs 29–32). R11·C19 Two reporting details differ: the abstract gives the droplets as "~200nm", and the title of Table 4 says "ASD powder dissolution" although the methods and Section 3.3 sample the medium of compact dissolution. X6·X5

Note by the AI writer (not discussed in the paper): the droplet tests were run at about 1 h, when, by Figure 2, the 10% nilvadipine compact had put roughly 59% of its 10 mg into 100 mL (about 59 µg/mL, well above 31 µg/mL) and the 15% compact roughly 1% of its 15 mg (about 1.5 µg/mL). The absence of droplets above the switch thus follows from how little drug was released, so the "clear link" between LLPS and congruency may largely reflect release extent; the authors' conclusions say as much when they note that above the threshold the amorphous solubility is not exceeded. X14·R4·C47·C48

The dissolving surface turns porous and drug-rich

Micro-CT cross-sections of compacts stopped after 10 min differ sharply on either side of the switch (Figure 7). The 10% nilvadipine and 15% cilnidipine compacts were dense and uniform; the 15% nilvadipine and 20% cilnidipine compacts had a porous layer at the face exposed to the medium, formed within those 10 min. For 15% nilvadipine, porosity grew further at 20 and 30 min and a network of channels developed in the matrix (Figure S3, not supplied). R13 The authors read this as the polymer dissolving faster than the drug, leaving pores and drug at the surface ("consistent with"). C21

SEM agreed (Figure 8). Below the switch the dissolving faces were smooth; above it they were rough and pitted, with holes up to about 20 µm across and raised rims for 15% nilvadipine and many circular holes of about 0.1–1 µm for 20% cilnidipine (sizes read from the scale bars, approximate). A fractured 20% cilnidipine compact showed that the holes reach below the surface (Figure 9). R14·R15 In the abstract the pits "most likely arise" from faster polymer than drug release. C22

EDX and XPS supply composition (Table C). On a 15% nilvadipine compact, the raised rim around a pit about 100 µm across had an N/O ratio of about 0.50, the value of the pure drug, against about 0.35 inside the pit and about 0.31 on the undissolved compact; rim and pit interior differ at p < 0.0001, pit interior and undissolved surface do not. R16·E2 The authors conclude that the protruding regions are drug-rich, "presumably" left behind by faster polymer dissolution, and that dissolution is non-uniform across the surface. C23 XPS gave a surface drug fraction of 32 ± 1% on the 15% nilvadipine compact after 10 min, up from 2 ± 0%, and 60 ± 4% on the 20% cilnidipine compact, up from 20 ± 1% (both p < 0.0001); nitrogen and oxygen peak fits agreed (data not shown), and the congruent 10% nilvadipine and 15% cilnidipine compacts did not change (Table S1, not supplied). R18 The authors take this as "further support" for the supposition that incongruent release forms a drug-rich interface. C25

Table C. Surface composition of incongruently releasing compacts: XPS from Table 5 (mean ± SD, n ≥ 3) and EDX N/O ratios read from Figure 10c (approximate, n = 4). Pure nilvadipine has N/O 0.5, PVPVA about 0.3. R16·R18

The starting surface of the 15% nilvadipine compact held only 2% drug by XPS, against 15% nominal; the authors suggest this "could be" polymer enrichment at the outer 10 nm of the ASD particles, as seen before in other systems (refs 2, 34, 35). C26 Note by the AI writer (not discussed in the paper): the EDX value on the same undissolved compact, about 0.31, is also close to PVPVA and below the roughly 0.36 expected for a uniform 15:85 mixture from the printed formulas, so the post-dissolution drug fraction is best compared with the nominal loading (32 against 15%) as well as with the starting surface. X13

Humidity makes a congruent dispersion incongruent

The humidity test asks whether phase separation in the wet matrix can by itself cause incongruent release. In the first 15 min, the unexposed 10% nilvadipine compact released about 15% of both drug and PVPVA; after 12 h at 97% RH the same type of compact released about 0.3% of its drug and about 2.6% of its PVPVA (read from Figure 3, approximate). R9 Between 15 and 20 min the surface layer appeared to peel off, "presumably" exposing fresh material, and drug release then became much faster (Figure S1, not supplied). R9·C16 The authors call the reduction "dramatic" and note that the polymer also slowed. C17

SEM showed a wrinkled, rough surface on the exposed compact (Figure 11), "most likely due to water-induced AAPS at the exposed surface"; after 15 min of dissolution the features were gone, which the authors read as AAPS confined to the surface rather than the bulk. R17·C24 In the matching film, prodan and R6G were evenly distributed before exposure; afterwards drug-rich domains, roughly 5–15 µm across, appeared as filled circles or as rings with polymer-rich material inside, within a continuous polymer-rich phase (Figure 12). R19 The authors say this "substantiates" the pattern seen on dissolving compacts: pits surrounded by a protruding drug-rich phase. C27

Powder dissolution shows the same cliff

Powder dissolution, run to check that the ranking was not an artefact of the compact hydrodynamics, gave the same break points. Nilvadipine powders at 5% and 10% reached about 91 and 89 µg/mL within the first hour and turned the medium turbid with LLPS (confirmed by DLS, NTA and fluorescence, data not shown); powders at 15% and 20% peaked at about 15 and 13.5 µg/mL near 2 h, below the amorphous solubility, in a clear medium (read from Figure 13, approximate). R20·C28 Cilnidipine powders at 5–15% released essentially all their drug within minutes, while at 20% the concentration reached about 1.33 µg/mL and at 25% about 0.65 µg/mL after 4 h (read from Figure 14). R21·C29 No crystals were seen in the nilvadipine media during the first 2 h. R20

Note by the AI writer (not discussed in the paper): two features of these plots go unexplained. After about 200 min the low-loading nilvadipine concentrations fall to about 12 and 22 µg/mL by 300 min, below the amorphous solubility, although crystals were looked for only during the first 2 h. And the 20% cilnidipine concentration exceeds the plotted amorphous solubility (about 0.63 µg/mL) for most of the 4 h test, whereas the conclusions state that above the threshold "the amorphous solubility is not exceeded". X7·X4

Discussion

The authors' mechanism: a race at the wet surface

The discussion assembles the findings into one picture. Because the polymer leaves the incongruent compacts faster than the drug, the interface "must" become enriched in drug, which explains the porous surface given how little drug the compact holds. C38 The pits themselves "can likely be attributed" to rapid phase separation at the interface into drug-rich and polymer-rich areas, followed by dissolution of the polymer-rich regions. C39 The confocal images of humid films supply the analogy: the drug-rich "donuts" there are "remarkably similar" to the drug-rich rims around pits in Figure 10, and the polymer-rich interior of each donut "likely dissolves faster", leaving the pits seen in Figure 8. C40 The humidity test is "further evidence": phase separating the surface turned a congruent system incongruent. C41

From this the authors "invoke" a mechanism of competing rates, which the abstract states as a result ("is a result of competitive kinetics between phase separation and the release rate of ASD components") and the conclusions as a suggestion ("We suggest that there is a competition between phase separation in the hydrated matrix and dissolution"). C42 The chain splits into two outcomes:

For nilvadipine the authors add that the amorphous drug alone releases more than three orders of magnitude more slowly than the polymer, "explaining the precipitous drop in release rate" when the mechanism changes. C43·R7 On this picture, what sets the switch is which rate wins at a given loading; drug hydrophobicity was the candidate they tested and rejected for these systems, and they leave the deciding properties open: a variety of drug and polymer properties is "likely" to play a role, and the physicochemical causes "are not well understood". C35·C34·C45

Note by the AI writer (not discussed in the paper): every direct image of phase separation in this paper comes from a 10:90 film or compact held at 97% RH for 12 h, a congruent composition under water vapour. The pits, rims and drug-rich surfaces come from 15:85 and 20:80 compacts after 10 min in buffer. No compact was imaged for phase separation while dissolving, and no rate of phase separation was measured to compare with the dissolution rate. The competition is therefore an inference that fits the observations, not a measured race. X10

A contrasting report

Chen and co-workers (ref 43) found that water-induced phase separation had no effect on the dissolution rate of an ASD. The authors offer "two possible explanations": that ASD had 40% drug loading and would already release in the drug-controlled regime, where phase separation during the test is fast anyway, so pre-exposure to moisture adds nothing; and the intrinsic dissolution conditions differed, which may change the relative rates of drug and polymer release, so "a direct comparison between the two studies may not be valid". C44·R22

What it means for formulation

The authors draw a direct recommendation: "if rapid release is required, drug loadings below this 'drop-off' concentration, need to be selected for optimal performance". C33 They also note the trade-off that frames the paper: making a drug amorphous should improve its dissolution far less than formulating it with a hydrophilic polymer whose release controls the drug's, but this "appears to be possible only at relatively low drug loadings", which "may not be desirable" for a commercial product because of the large dosage form. C46 Understanding what sets the switch "is likely to be important for formulation design". C45 The congruent regime is the one that produces drug-rich nanodroplets, which earlier work suggests act as a reservoir that keeps membrane flux at its maximum and may shuttle drug across the aqueous boundary layer (refs 5, 8, 36). C31

Note by the AI writer (not discussed in the paper): three practical points follow from the data. The cliff is steep: for both drugs a single 5% w/w step took drug release from the polymer's rate to near the neat drug's, so a loading chosen close to the switch leaves little margin. R5·R6 Measuring polymer release from the same samples, here by size-exclusion chromatography, shows directly which regime a formulation is in, because the ratio of polymer to drug rate jumps from about 1 to 20 or more. R5·R6 And because log P did not predict the switch for these two analogues, the maximum congruent loading still has to be found by testing each drug–polymer pair. C35

Limitations

Limits the authors state. The authors call their own two-regime picture of earlier work "likely" oversimplified and do not claim to know which physicochemical factors set the switch; they write that "other factors need to be explored" and that "further studies are clearly needed to probe dissolution mechanisms from ASDs". C4·C34·C45·C49 They note that the sulfathiazole comparison is hard to interpret because the polymer differs, and that their comparison with Chen et al. may not be valid. C36·C44 Several supporting data are described as "not shown": solubility at high polymer concentration, XPS nitrogen and oxygen fits, SEM of the humid compact after 15 min, and the LLPS checks of the nilvadipine powder media. R2·R17·R18·R20

The AI writer's own judgement. The measurements are solid where they are direct: drug and polymer rates from the same samples with t-tests, three agreeing droplet methods, and four surface techniques that all show a drug-rich, porous interface above the switch. R5·R6·R8·R10·R11·R12·R13·R14·R16·R18 The weaker parts are interpretive and of scope.

• The mechanism is inferred from humidity-conditioned 10:90 films and compacts; phase separation was not observed during dissolution of the compositions that actually release incongruently. X10
• The log P conclusion rests on two analogues whose similar hydrogen bonding with PVPVA was assumed, not measured, with loadings tested in 5% w/w steps, and on cited systems one of which used PVP; the highlight that the switch "is independent of drug logP for PVPVA ASDs" goes beyond this. X11·C7·C37
• The molar loadings that underpin "a similar drug loading" for both drugs cannot all be reproduced with one stated convention, although the overlap survives either convention. X2
• The 20% cilnidipine powder exceeded the amorphous solubility, against the statement that it is not exceeded above the threshold; and the low-loading nilvadipine powder concentrations fell after 2 h without comment. X4·X7
• "Drug-controlled" release is still 6.5–17 times faster than the neat amorphous drug for nilvadipine, and the cilnidipine drop is about 360-fold rather than "similar" to the 60-fold nilvadipine drop. X9·X8
• Smaller reporting issues: the source of the literature cilnidipine solubility (ref 23 or 24), the title of Table 4 (powder versus compact dissolution), the droplet size in the abstract (~200 nm versus 237–366 nm), and the dashed amorphous-solubility line in Figure 2b–d, which sits at the µg/mL value on a % release axis. X1·X5·X6·X3
• The supplemental information (Figures S1–S5, Table S1) was not available for this reading, so the later release of the humid compact, the NTA sizes, the 20–30 min micro-CT and the XPS of congruent compacts are known only from the text. R9·R12·R13·R18

Conclusion

For nilvadipine and cilnidipine in PVPVA, release is fast and congruent, and the drug forms drug-rich nanodroplets, only up to 10% and 15% w/w loading; one step higher, the polymer still dissolves but the drug stays behind, droplets no longer form, and the dissolving surface becomes porous and drug-enriched. R5·R6·R11·R13·R18 The switch sits at similar molar loadings for drugs whose log P differs by 2.5 units, so drug hydrophobicity did not set it for these two analogues. R7·C35 Exposing a congruent compact to high humidity, which phase-separates its surface, made its release incongruent, and the authors suggest that the switch reflects phase separation of the hydrated matrix outpacing dissolution, a mechanism inferred rather than measured here. R9·C41·C42·X10 The study defines congruent release as the "sweet spot" of ASD dissolution and shows how narrow it can be.