How a nanometre oxide coating on ASD particles changes release from copovidone powders and tablets

Background and motivation

An amorphous solid dispersion (ASD) has to meet two requirements: it must stay amorphous on the shelf, and in the gut it must generate and maintain supersaturation (Introduction; Moseson and Taylor 2023, ref 3) C60. Atomic layer coating (ALC) addresses the first. In ALC a ceramic film, typically aluminium oxide or zinc oxide and typically less than 10–100 nm thick, is built up on each particle by repeated gas-phase reactions, one atomic layer per cycle. On high drug loading ASD powders, coatings of 5–40 nm inhibited surface crystallization under accelerated temperature and humidity, and they also improved wettability, dispersibility and flowability while reducing agglomeration and hygroscopicity (Van Duong 2022, ref 12; Moseson 2022, ref 13) C1.

The question is what the coating does to the second requirement, release. Earlier work found that TiO2, Al2O3, ZnO and SiO2 coatings delay and slow release from powders, more so the thicker they are. In the group's own study on hypromellose acetate succinate (HPMCAS) ASDs, Al2O3 thickness and defect density changed water sorption, but release was unaffected under fast-release conditions and even improved under slow-release conditions, because the coating prevented crystallization of the matrix (Moseson 2023, ref 20) C2. Two gaps remained. First, nobody had studied coated ASD particles pressed into tablets; the one tablet study, on metformin HCl, found no effect of the coating (Swaminathan 2023, ref 21). Second, coating work on ASDs had covered only HPMCAS C3. Copovidone (PVPVA, Kollidon VA64), the polymer in about half of marketed ASDs, dissolves through extensive gelation, so the authors expected a coating to matter ("likely to be important") C4·C6.

The release of PVPVA ASDs is framed by the limit of congruency (LoC). Below a critical drug loading (DL), drug and polymer leave together at the polymer's fast rate (polymer-controlled, congruent release); above it, release becomes drug-controlled, slow and possibly incomplete, "typically" because the drug has phase separated into a continuous amorphous layer that blocks further hydration (refs 22, 26, 27, 30, 31) C5. For ritonavir (RTV) and PVPVA, congruent release from neat ASD tablets has been seen up to 25% DL at 37 °C (Indulkar 2019, ref 27) R20. The authors chose this pair because its phase behaviour and release are well characterized, because it does not crystallize quickly in storage or dissolution (so fewer variables), and because it is in marketed products (Norvir, Kaletra, Viekira Pak, Paxlovid) C6. Ritonavir is a weak base (pKa 1.8 and 2.6) and is far more soluble in 0.1 N HCl (400 µg/mL) than at pH 6.8 (about 1 µg/mL crystalline, about 27 µg/mL amorphous) (Table 1) M1.

The paper therefore asks how Al2O3 coatings of different thickness and conformality on 10 and 20% DL RTV–PVPVA ASDs, both below the LoC, change release, first from powders and then from tablets, and whether formulation can recover any loss. Confocal fluorescence microscopy (CFM) with two dyes is used to watch hydration, gelation and phase separation directly C7·A1–A6.

Background concepts

The coating and how its thickness is estimated. The coating is Al2O3 deposited by ALC at either a normal rate or a slow rate (below 0.4 nm per cycle); a slower process is taken from earlier work to give a better barrier, so the slow coating was meant to be "highly conformal", that is, continuous with few cracks or pinholes P4·C8. Thickness is not measured directly. The residue after burning a coated sample to 800 °C in TGA is taken as the oxide mass fraction, and this is converted to an estimated thickness using the specific surface area (SSA) measured by nitrogen adsorption (BET) P9·P10·E1. The paper prints no formula. Note by the AI writer (not discussed in the paper): all nine thicknesses in Table 2 are reproduced, within 0.1 nm, by dividing the oxide mass fraction by an oxide density of 3.0 g/cm³ and by the SSA of the uncoated powder of the same size fraction, for example 0.0375 / (3.0 × 0.56 m²/g) = 22.3 nm for RP20/S/C22. The thicknesses are therefore uniform-shell estimates, not measured film thicknesses (X19). X19

Defects and conformality. Because Al2O3 does not dissolve on the time scale of a release test (its solubility rises only moderately at low pH, above 10^-2 M), water can reach the ASD only through gaps in the coating C26·R22. The paper uses "defects" for these gaps and "conformality" for their absence. Neither defect density nor conformality is measured; both are inferred from where and whether particles take up water X14.

The gel layer and the drug-rich phase. A PVPVA tablet first absorbs water and swells into a gel layer of entangled chains. As water content rises the chains disentangle and leave at the gel–solution interface, and a steady state between water penetration and disentanglement sets the gel thickness. In a drug–PVPVA ASD the wet gel usually also splits, because of a miscibility gap, into a drug-rich phase and a polymer/water-rich phase. Below the LoC the drug-rich phase forms nanodroplets that are carried out with the disentangling polymer; above it the drug-rich regions take up more of the gel, connect over time and hinder further hydration (Discussion; refs 22–24, 27, 35, 39, 41, 45–52) C61. The whole paper turns on which of these two morphologies forms.

How the dyes report phase and water. Trace Nile red (0.01%) partitions into the drug-rich phase and shows red; Alexa Fluor 488 (0.001%) marks the polymer/water-rich phase and shows green; where both are present the image is orange. Nile red is quenched by water while Alexa Fluor 488 becomes brighter, so the advance of water appears as a change from orange to green C9·P12. The dyes were dissolved only into a rotary-evaporated 20% DL ASD (the RP20/F samples), not into the hot-melt extrudates P3.

The release test. Every test used 50 mg of ASD (powder or tablet) in 200 mL (10% DL) or 400 mL (20% DL) at 37 °C and 100 rpm, so that the ritonavir concentration can reach at most 25 µg/mL whatever the DL or coating, close to the amorphous solubility at pH 6.8; concentration was followed in-line by UV for 1 h P11. At pH 6.8 this is a non-sink test: the curve shows how fast the drug can approach its solubility limit. In 0.1 N HCl, where ritonavir dissolves at 400 µg/mL, the same dose is far below solubility M1. Only ritonavir was measured, not PVPVA, so congruency itself is not tested here X18. The text calls 25 µg/mL "equivalent to the amorphous solubility", while Table 1 lists about 27 µg/mL (X4). X4

Superdisintegrants. Crospovidone (xPVP) draws water in by wicking and expands back toward its pre-compression shape (shape recovery); croscarmellose sodium (CCS) and sodium starch glycolate (SSG) act mainly by wicking and swelling in all directions (refs 40, 59, 60) C46. With a gelling polymer such as PVPVA a tablet that really disintegrates is hard to make; it usually needs a low ASD loading, high disintegrant levels and often kosmotropic salts (refs 5, 61, 62) C52.

Methods in detail

The study is a sequence of experiments on one drug–polymer pair. Each stage was designed in response to the previous one: powders behaved almost like uncoated material, tablets did not, imaging was used to see why, and formulation was used to fix it.

Samples and coatings

Ritonavir and PVPVA were hot-melt extruded at 145 °C and 200 rpm at 10 and 20% DL, cryo-milled for 3 min at 10 Hz under liquid nitrogen and sieved into a small (<106 µm, "S") and a large (106–250 µm, "L") fraction P1·P2. For imaging, a 20% DL ASD containing 0.01% Nile red and 0.001% Alexa Fluor 488 was prepared by rotary evaporation from ethanol, dried and cryo-milled (<250 µm, "F") P3. Each fraction was coated at a normal rate to a thin (below 15 nm) and a thick (above 20 nm) target; the dye ASD received one slow coating P4. Sample names combine DL, fraction and coating, so RP20/S/C22 is the 20% DL small fraction with an estimated 22 nm coating R1. By XRPD and modulated DSC all samples were amorphous, and Tg did not change with coating or particle size; these data are in the SI (Table S1, Figure S1), which was not available here R2·C10.

Table A. The fourteen samples (from Table 2). BET area was measured on the uncoated powder; oxide content and thickness are given for the thin and thick coating of each fraction R1.

Two features of Table A matter later. Small particles carry two to three times more oxide per gram than large ones for a similar thickness (1.95–5.30% against 0.92–2.58%), because they have more surface per gram R1. And the dye ASD that is used for every image is a different material from the extrudates whose tablets were tested: it was made by rotary evaporation and coated slowly to be conformal P3·P4·X6.

Release testing and tablets

Release used the non-sink set-up described above, with 50 mg of ASD at 37 °C and 100 rpm for 1 h, in pH 6.8, 50 mM phosphate buffer or in 0.1 N HCl P11. Neat tablets of 50 mg were compressed on a Gamlen GTP-1 press with 6 mm tooling at 140 MPa, the weight being corrected for the oxide content so that every tablet held the same drug dose; tablets were also made at other compression forces, and solid fraction and tensile strength were measured for RP20/S/NC and RP20/S/C22 tablets (SI Figure S6, Table S2) P5·R11. For the formulation work RP20/S/C22 was chosen as the coated material because coating had slowed its tablet release most, with RP20/S/NC as the uncoated control C44. The tablet compositions are in Table 3:

Table B. Tablets with disintegrants and formulated tablets (from Table 3) P5.

Card check: the "30% ASD" row does not add up. Its masses (25, 3, 3, 28.5 and 0.5 mg in 60 mg) correspond to 41.7, 5.0, 5.0, 47.5 and 0.8% w/w, not to the printed 30, 6, 6, 57 and 1%. The footnote (two tablets per 50 mg dose) fits 25 mg of ASD per tablet; the Figure 8 caption repeats 6% xPVP and 6% CCS. Which is right cannot be decided from the paper (X1). X1

Imaging

For particle imaging, a little powder was placed on a glass slide and wetted with a drop of pH 6.8 buffer or 0.1 N HCl and followed for up to 3 h at room temperature; the coated dye powder was also imaged after sonic sieving, which the authors used to create coating defects P12. For tablet imaging, 100 mg compacts were pressed in a Wood apparatus die at 1500 psi for 1 min, glued to a glass dish with UV-cured resin, and the edge of the tablet was imaged at 37 °C as pH 6.8 buffer entered, with and without 17% w/w xPVP P13. The printed tablet diameter is "80 mm", which cannot be right for a Wood apparatus die (8 mm is likely); these compacts also differ in size and compaction from the 6 mm, 140 MPa release tablets (X5). X5 The paper does not name which ASD was pressed for Figure 6; since only the RP20/F samples contain the dyes, the coated images presumably show the slowly coated RP20/F/C21 X6.

Experimental analysis

Powders: the coating delays only the start

At pH 6.8 the uncoated powders released almost at once: the large fractions were at about 23 µg/mL (about 93% of the dose) after 2.5 min, the small fractions at about 20–21 µg/mL, levelling at about 24.3 µg/mL (10% DL) and about 21.8 µg/mL (20% DL) (Figure 2) R4.a·R4.d·R4.g·R4.j. The authors summarize this as complete release, about 90% dissolved, within 10 min at 10% DL and 20 min at 20% DL C13. Coated powders started more slowly, most clearly in the small fractions, but by 60 min seven of the eight coated powders were at about 23.5–24.2 µg/mL, at or above their uncoated controls (Table C) R4. Thin and thick coatings did not rank consistently C21.

Table C. Powder release at pH 6.8, RTV in µg/mL (25 µg/mL = whole dose), read from Figure 2 (approximate) R4.

The exception is RP10/S/C27, which levelled off at about 21.2 µg/mL, about 85% of the dose. The text does not mention it, and it sits uneasily with the conclusion that coated powders release like uncoated ones whenever the coating has defects R4.f·X8. The authors attribute the slight delay of the coated powders to slower hydration and to the time needed to disrupt the coating, and conclude that "as long as defects were present" release was complete and rapid C17.

Water uptake, local wetting and the conformal sample

At 37 °C and 90% RH, coated and uncoated 20% DL powders of the same size fraction took up water at similar rates and ended at about 19–21% after 3 h; smaller particles took up water faster at first (Figure 1a, b) R3.a–f·C11. The one clear exception was the slowly coated dye sample: RP20/F/C21 took up water more slowly (about 12% against about 16.5% at 30 min) and stopped at about 15% from about 70 min, against about 20% for RP20/F/NC; the text gives 15% against 20% (Figure 1c) R3.g–h.

CFM shows what the averages hide (Figure 3). Within 5–15 s of adding buffer, an uncoated dye particle developed an intense green rim around its whole perimeter, the surrounding buffer turned green as PVPVA dissolved, and the particle lost its shape, leaving small red ritonavir-rich domains R7.a. The authors note that at room temperature 20% DL is already beyond the LoC (citing Krummnow 2022, ref 33), which is why drug-rich domains remain behind C15·R21. Coated particles did not swell visibly and kept their angular outlines; green appeared only at specific points on their edges, which the authors read as water entering "presumably at cracks or defects in the coating", and some particles were still dry after 10 min R7.b·C16.

Release told the same story. RP20/F/NC reached about 90% within 20 min, but RP20/F/C21 levelled off at 80% after 30 min; in acid, where both ritonavir and Al2O3 are more soluble, RP20/F/NC was faster (90% in about 15 min) and RP20/F/C21 again stopped at 80% (SI Figure S2) R5·R6. Coated dye particles held in acid for 3 h showed hardly any further hydration after the first fraction had wetted, and the coating stayed intact (SI Figure S3); after sonic sieving the particles hydrated much faster and fewer stayed intact (SI Figure S4) R8·R9. The authors conclude, with hedges that vary from "suggesting" to "presumably" and "assumingly", that a fraction of this sample carries a coating with no defects that keeps water out entirely, and that sieving creates defects C12·C18.

Note by the AI writer (not discussed in the paper): the strongest evidence for this conclusion (the 80% plateaus, the 3 h and the sieving images) is in the SI, and several SI panel references conflict (Figure S2c is cited both for powders in acid and for tablets at pH 6.8) (X16). X16

Neat tablets at pH 6.8: the coating cuts release

Pressed into tablets, everything slowed, as expected from the smaller exposed surface: uncoated tablets reached about 90% of the dose by 20 min (10% DL) and 30 min (20% DL) (Figure 4) R10.a·R10.d·R10.g·R10.j·C19. Tablets of coated particles released "more slowly and incompletely within 60 minutes" in the authors' words, and the loss was "most notable" at 20% DL and for the small fraction C20.

Table D. Neat tablet release at 60 min, pH 6.8, RTV in µg/mL with the percentage of the dose, read from Figure 4 (approximate) R10.

The losses are large for the 10% DL small fraction and for all 20% DL tablets, worst for RP20/S/C22 at about 35% of the dose. They are absent for the 10% DL large fraction: both coated tablets reached about 23.5–24.1 µg/mL by 30 min, so the general statement about incomplete release does not hold for them R10·X7. Large-particle coated tablets released faster than small-particle ones; the authors relate this to the larger coating weight fraction of the small particles and their "lower tendency of coating fracture", which "likely underpins this observation" C22·R1. Thickness again gave no consistent trend: the thicker coating was worse only in the 20% DL small fraction C21·R10. Tablets made at other compression forces did not change the picture; coated-particle tablets had lower solid fraction and tensile strength without any release advantage, although, as the authors note later, a lower compression force "moderately increased" release from coated tablets (SI Figure S6, Table S2) R11·C23·C45.

Neat tablets in 0.1 N HCl: no coating effect

In 0.1 N HCl the six 20% DL tablets followed nearly one curve: about 9.5–12.8 µg/mL at 10 min, about 20.8–22.5 µg/mL at 30 min and about 21.7–22.7 µg/mL at 60 min (Figure 5) R12. The authors describe release from coated particles as "within measurement error, nearly identical" to that from uncoated particles, and later as over 90% in less than 30 min C25. Card check: the plotted values at 30 min correspond to about 83–90% of the dose, and the 60 min plateaus to about 87–91% (X9). X9 Uncoated tablets released the same at both pH values, which the authors expected because 20% DL is in the polymer-controlled regime (SI Figure S5) R13·C24. Since some conformally coated particles were still dry after 3 h in acid, the authors argue that the poor release at pH 6.8 is not a matter of the coating keeping particles from the medium; it is "a result of slowed phase separation kinetics from the coating paired with reduced drug solubility" C26·C27. Note by the AI writer: this wording conflicts with the later explanation, in which the coating slows hydration and thereby allows more phase separation (X21). X21

The tablet edge: nanodroplets against connected drug-rich regions

Figure 6 is the central image of the paper. In the uncoated tablet (Figure 6a), water advanced as a green gel layer while the dry tablet stayed orange; at 1 min the Nile red channel showed small discrete red droplets throughout the gel, and these moved into solution as the polymer disentangled and dissolved, "as expected" for a system below the LoC at 37 °C R14.a·C28. Large red blobs in the 2 and 3 min frames are, according to the authors, drug that descended from above and was trapped by the mounting resin C29. In the tablet of coated particles (Figure 6b), particle outlines were already visible in the dry compact; within 2 min green and red networks formed along particle boundaries, and by 4 min connected red regions remained at the interface after the polymer had left R14.b. The authors describe "larger, connected regions of the drug-rich phase" that "remain after the polymer dissolves" and "appear to form near particle boundaries"; the outlines could be followed from the dry tablet into the hydrated gel (SI Figure S7) C30·R16. All coated particles in the tablet appeared to be wetted, "likely due to defects created during tablet compression" C18. From this the authors draw their main inference: "These observations suggest" that the lower rate and extent of release from tablets of coated particles comes from the different phase-separation morphology C31.

With 17% w/w xPVP (Figure 6c, d), uncoated tablets behaved as before, with nanodroplets carried into solution, and the solvent front (Nile red quenching) advanced faster next to the dark disintegrant particles. In tablets of coated particles the disintegrant "appears" to disrupt the connected regions: the inset shows smaller droplets, and fewer drug-rich regions remained at the interface after polymer release R15·C49. The discussion describes this experiment as "20% w/w xPVP"; the results and the Figure 6 caption give 17% (X2). X2

Tablets with one superdisintegrant

Adding a single superdisintegrant changed the coated tablets from a curve that levels off to an almost linear rise (Figure 7) C47. At 60 min, against about 8.8 µg/mL for the neat RP20/S/C22 tablet, coated tablets reached about 16.6 µg/mL with 9% xPVP, about 22.1 with 17% xPVP, about 22.3–22.6 with CCS and about 21.8–22.0 with SSG (Table E) R17. All stayed slower than uncoated tablets with the same disintegrant C47. In the uncoated tablets xPVP left the release curve unchanged, but CCS and SSG slowed early release (about 7.7–10.6 against 14.5 µg/mL at about 10.5 min) R17.a·R17.c–d·R17.g–h·R17.k–l. The authors suggest this "may be a result of the gel formation mechanism" of these disintegrants and "speculate" that they raise the viscosity of parts of the PVPVA gel C50. Because of the high ASD content, these tablets still eroded through PVPVA gelation rather than disintegrating C51.

Table E. Tablets of RP20/S/NC (uncoated) and RP20/S/C22 (coated) with one superdisintegrant, RTV in µg/mL at pH 6.8, read from Figure 7 (approximate) R17.

The authors attribute the gain mainly to faster water ingress: when hydration is accelerated, "less time is available for phase separation", and xPVP avoids aggregation of drug-rich regions C48.

Formulated tablets

Finally the coated particles were diluted in a full formulation with 6% xPVP, 6% CCS (two disintegrants with different mechanisms), lactose and 1% MgSt, at 30% or 50% ASD (Figure 8) P5·C46. At 30% ASD the coated formulation was the fastest of all tablets: about 20.2 µg/mL at about 8 min and about 22.1 µg/mL (about 88%) at about 10.5 min, against about 14.9 and 17.9 µg/mL for the same formulation of uncoated particles and about 14.5 µg/mL for the neat uncoated tablet at about 10.5 min R18.a–b·R17.a. It released nearly as fast as the coated powder itself (SI Figure S8) R19. At 50% ASD release was slower and nearly linear, with about 22.4 µg/mL (about 90%) for uncoated and about 20.5 µg/mL (about 82%) for coated particles at 60 min, the latter still rising R18.c–d. Even the formulated tablets eroded rather than disintegrated R18.

The authors' wording varies across the paper. The abstract states that this formulation "achieved ~90% release in less than 10 minutes, matching the uncoated ASD particles of the same formulation", and the Conclusion that the release "matched that of the uncoated system"; the Results state that the coated formulation released faster than the uncoated one, "suggesting that the coating further enables the ASD to release as discrete particles, rather than releasing through a gel of aggregated particles" C54·C55. For the 50% ASD tablets, the Results give "nearly 90%" within 60 min and the Discussion "released completely" C53. Card check: the coated 50% ASD tablet was at about 82% at 60 min, and "matching" understates how much faster the coated 30% ASD formulation was in the first 20 min; the Discussion also cites these data as Figure 7a and 7b, while they are in Figure 8 (X10, X11, X2). X10·X11·X2 The authors credit the improvement to the superdisintegrants' hydration mechanism, to dilution "so that large drug-rich regions cannot form", and to better access to the particle surfaces C56.

Discussion

The authors' explanation: a race between water and phase separation

The Discussion assembles the findings into one chain. Al2O3 hardly dissolves, so water reaches the ASD only through coating defects C26·C16. In a powder each particle is surrounded by medium, defects are found quickly, and release nearly matches that of uncoated powder; here, defects matter and thickness between about 10 and 30 nm does not, which refines the hypothesis of Vogel et al. (ref 54) that thickness, defect density and coating solubility together control release from ALC films C17·C21·C40·C41. A neat tablet exposes far less surface and erodes through a PVPVA gel. Compression probably adds defects, but the insoluble coating fragments "likely" stay in the gel until it erodes and lengthen the path for water going in and polymer coming out C19·C32. Slower hydration then shifts the balance: phase separation and ripening of the domains have time to proceed, larger interconnected drug-rich regions form along particle boundaries next to the coating, and "it is apparent" that drug near the undissolved coating becomes trapped C32·C33.

The authors describe the result as an "effective LoC" pushed below the loadings studied: "This slowing of hydration kinetics reduces the effective LoC to a value below that of the systems studied herein." Because the coating prevents a connected, continuous gel layer, no continuous drug-rich layer forms at the gel–solution interface; the polymer still dissolves, but the drug-rich regions are not swept out with it C34. These regions are amorphous, so ritonavir can still dissolve up to its amorphous solubility, but amorphous ritonavir dissolves much more slowly than the 20% DL ASD (Yang 2023, ref 22), so supersaturation builds slowly C35·R24. In acid, where ritonavir is far more soluble and dissolves faster, the stranded drug dissolves anyway, which is why coated and uncoated tablets release alike in 0.1 N HCl C35·C25. At 10% DL the drug is less concentrated in the polymer, phase separation and morphology evolution are slower, and the coating hurts less, although it still reduces release from the small-particle tablets C37·C38. Disintegrants and dilution work on the other side of the race: they speed water ingress and keep large drug-rich regions from forming C48·C56.

Where the explanation applies

The data come from one weakly basic drug, one neutral gelling polymer and one coating material, with release followed for 60 min under non-sink conditions at pH 6.8 and in 0.1 N HCl. The ASDs were below the LoC of the uncoated system and did not crystallize. The authors generalize in two directions. First, "coating of PVPVA-based ASDs of weakly basic drugs may have little effect on release in vivo due to their higher solubility in the acidic conditions found in the stomach" C36. Second, the data "support" their earlier conclusion that thin, defective coatings give the best balance between crystallization inhibition and release C39. They also suggest that PVPVA ASDs "may be less sensitive to coating thickness" than HPMCAS ASDs because they swell and sorb more than three times as much water, which "likely" cracks the coating, and that the larger particles used here fracture more easily C42·R23.

What it suggests for formulation

The authors' own recommendation is that "an optimized formulation can be used to combat the effect of the coating on release from PVPVA-based ASDs, while providing a stability advantage against crystallization", with the 30% ASD tablet as a "proof-of-concept" that coated particles can be used in disintegrating tablet formulations; they add that tablet solid fraction and compression settings, explored only briefly, "may yet serve to achieve a fast-releasing system" C58·C57·C45.

Note by the AI writer (not discussed in the paper): three practical points follow from the data. Powder release and water uptake were poor predictors here: they barely changed, while tablets of the same particles lost up to two thirds of their release, so a coated ASD should be judged in the dosage form it will be used in R4·R10. Because the effect behaves like a lower effective LoC, coated ASDs whose DL is close to the LoC (here 20% against 25%) are most at risk, while the 10% DL large-particle tablets were unaffected R10·R20. And because the effect disappears in acid, a 0.1 N HCl release test alone would hide it for a weakly basic drug; the pH 6.8 test is the discriminating one R12.

Limitations

Limitations stated by the authors. The authors acknowledge that the formulation work was limited by the availability of coated particles, that variables such as the tablet solid fraction "were not evaluated to a full extent and could have a significant effect on water penetration and disintegration properties", that the particle-hydration experiments were performed in quiescent conditions, and that additional development would be needed for an erodible-style tablet C45·C57·P12. They also flag the resin artefact in Figure 6a C29.

The AI writer's own judgement (1): the imaged material and the mechanism. The mechanism rests on CFM, and the CFM material is not the material whose release defines the problem. The dyes were only in the rotary-evaporated RP20/F ASD, whose coating was deposited slowly to be conformal, while the large tablet losses were measured on hot-melt-extruded particles with normal-rate coatings such as RP20/S/C22; the CFM tablets were also larger and pressed differently (X6, X5) X6·X5. Only ritonavir was measured, so congruency and the "effective LoC" shift are inferred from images and from the cited framework, not shown by polymer release or by a DL series of coated ASDs (X18) X18. Defect density, which the authors call "the most important characteristic", was never measured (X14) X14·C41. The hydration rate in tablets, the variable at the centre of the explanation, was not quantified either.

The AI writer's own judgement (2): numbers stated more strongly than the plots. Several summary statements go beyond the figures: incomplete release is stated for all coated tablets although the 10% DL large-particle tablets released completely (X7); "over 90% release in less than 30 min" in acid compares with about 83–90% in Figure 5 (X9); the coated 50% ASD tablet "released completely" in the Discussion but was at about 82% at 60 min (X10); and the coated 30% ASD formulation "matched" the uncoated one although it was clearly faster early on (X11) X7·X9·X10·X11. The composition of the 30% ASD tablet in Table 3 is internally inconsistent (X1), the number of replicates and the meaning of the error bars are not stated (X20), and some curves, such as the RP20/L coated tablets, have wide error bars X1·X20·R10.h–i. Finally, the conclusion's "stability advantage against crystallization" was not tested on these coated ritonavir ASDs; it rests on earlier work (X13) X13. None of these points changes the main contrast between powders and tablets at pH 6.8, which is large and direct.

Conclusion

The paper shows, for 10 and 20% DL ritonavir–PVPVA ASDs, that an Al2O3 nanocoating of about 9–28 nm with defects hardly affects powder release, that a slowly deposited conformal coating can keep part of a powder dry, and that the same coated particles in a neat tablet lose much of their release at pH 6.8 but not in 0.1 N HCl R4·R5·R10·R12. Tablet-edge imaging shows nanodroplets in uncoated tablets and connected drug-rich regions along particle boundaries in tablets of coated particles, which the authors read as slowed hydration letting phase separation outpace dissolution ("These observations suggest") R14·C31. Superdisintegrants, and a formulated tablet with 30% ASD, xPVP, CCS and lactose, restore fast release R17·R18·C54.