Does a nanometre alumina coating on amorphous solid dispersion particles hold back drug release?

Background and motivation

An amorphous solid dispersion (ASD) gives a poorly soluble drug a solubility advantage only while the drug stays amorphous, so crystallization has to be prevented both on the shelf and during dissolution C5. On the shelf, crystallization often starts at the particle surface, where it can be orders of magnitude faster than in the interior C1. Atomic layer coating (ALC) attacks exactly that surface. In earlier work by the same groups, Al2O3 coatings of 5–40 nm kept high-drug-loading ASDs, which crystallized in under a month uncoated at highly accelerated open-dish conditions, stable for 6 months to over 2 years (Duong 2022, ref 10; Moseson 2022, ref 11) R27·C1. ALC has also been used to improve wettability, flow and dispersibility of powders and to lower their hygroscopicity C2.

A coating that blocks the surface might also block release. Studies on lactose, budesonide, acetaminophen and indomethacin particles coated with TiO2, Al2O3, ZnO or SiO2 found that thicker ALC shells delayed release (refs 21–25) C3. In the first dissolution study of an Al2O3-coated ASD, the time to maximum concentration rose with coating thickness; the maximum itself was higher for coated samples after storage, because the uncoated ones had crystallized R31·C4. That sentence in the introduction carries no reference number. A thin polymer coating (PDDA) on amorphous indomethacin slightly lowered its intrinsic dissolution rate (Wu 2007, ref 26), "similarly suggesting" that coatings may slow release C4.

Release from an ASD is itself complicated. It depends on pH, wettability, drug loading (DL) and the kinetics of hydration, phase separation and crystallization; an enteric polymer such as hydroxypropyl methylcellulose acetate succinate (HPMCAS) must be ionized before it dissolves and frees the drug; and at high DL release slows and matrix crystallization becomes more likely C5. The authors therefore set out to study release from coated ASD particles under several pH conditions and to "begin to develop a mechanistic framework" A5·A6. They chose posaconazole (PCZ) in HPMCAS because this is the composition of the marketed product Noxafil, which contains 25% DL PCZ in HPMCAS M grade made by hot melt extrusion (HME) C6.

The study has three parts. First, water uptake of coated and uncoated particles at 90% relative humidity (RH). Second, release in single-stage tests at pH 6.5 and in two-stage tests with an acid stage at pH 1.6 or 4.0 for 30 min to 4 h. Third, imaging of the coating after exposure to each medium by scanning electron microscopy with energy dispersive X-ray analysis (SEM/EDX), and of crystals by second harmonic generation (SHG) microscopy P10·P13·P14·P15·P12.

Background concepts

Atomic layer coating. ALC, also called atomic layer deposition (ALD) when done on flat substrates, is a solvent-free gas-phase method that builds a film of under 10 to 100 nm one molecular layer at a time C1. In this paper each cycle has two half-cycles at 35 °C under vacuum in a rotating powder reactor: trimethylaluminum is dosed until it has reacted with every available surface site, the excess is pumped away, and then water is dosed to react with the bound precursor and leave aluminum oxide P3·M3·M4. Repeating the cycle grows the film. Because each half-cycle saturates the surface, the film follows every bump of the particle (it is "conformal"). The growth per cycle can still vary; the authors ran slow (below 0.4 nm/cycle), medium (0.4–1.0 nm/cycle) and fast (above 1.0 nm/cycle) processes "to prepare coatings of various barrier performance" P3. The idea, developed in the Discussion, is that slower growth gives a denser film with fewer voids C60.

Thickness from oxide content. The oxide left after burning a coated sample at 800 °C in a thermogravimetric analyzer (TGA) gives the mass fraction of Al2O3. Divided by the specific surface area (SSA, from nitrogen adsorption and BET theory) and by the oxide density, this gives an average film thickness P6·P5·E1·E2. The paper describes this conversion in words only; it does not print the formula or the density used E2.

The drug and the polymer. PCZ is a weak base with experimental pKa values of 3.53 and 1.24 and a calculated log P of 4.6; its crystalline solubility is 60 µg/mL at pH 1.5, 0.33 µg/mL at pH 4.0 and 0.25 µg/mL at pH 6.5, and its amorphous solubility is 1700, 12 and 8 µg/mL at the same pH values (Table 1, values cited from Duong 2022, ref 48, and Elkhabaz 2019, ref 38) M1. The authors describe PCZ as fully ionized at pH 1.6, primarily un-ionized at pH 4.0 and fully un-ionized at pH 6.5 C7. HPMCAS behaves the other way: it has a pKa of 5.0, and the M grade dissolves only above about pH 6.0 M2. In acid, therefore, the polymer stays solid and only the drug can leave; at pH 6.5 the polymer dissolves and carries the drug out.

Non-sink and two-stage tests. In a non-sink test the dose is at or above the solubility of the amorphous drug, so the test can show whether the formulation reaches and keeps the amorphous solubility. Here the dose was 8 µg/mL in 200 mL of pH 6.5 phosphate buffer, equal to the amorphous solubility of PCZ at that pH P13·M1. The two-stage test imitates stomach then intestine: 100 mL of acid with a dose of 16 µg/mL, then 100 mL of buffer concentrate to reach pH 6.5 and a final dose of 8 µg/mL P14. Drug that dissolves in the acid stage is called leached drug.

Matrix crystallization and SHG. Matrix crystallization is crystallization of the amorphous drug inside the wetted solid after water exposure; it removes the solubility advantage before the drug has been released, and it is more likely at high DL C5. SHG microscopy (here with a SONICC instrument) detects chiral crystals such as crystalline PCZ while amorphous solid and dissolved drug stay dark; its detection limit is typically below 0.01% crystallinity, against about 1% for X-ray powder diffraction (XRPD) P12·R30. The paper reports SHG as an "integrated intensity", the summed signal of an image, not as a percentage of crystals P12.

Methods in detail

The work has no model or equation; it is a comparison of twelve samples across several tests. The flow below shows how the parts fit together.

The twelve samples

PCZ and HPMCAS (Aqoat MMP grade) were extruded at 175 °C and 200 rpm at 25 and 50% DL, cryomilled under liquid nitrogen for 3 min at 10 Hz and sieved into a small (<106 µm, "S") and a large (106–250 µm, "L") fraction P1·P2. Each fraction was kept uncoated ("NC") or coated with Al2O3; the number after "C" in a sample name is the TGA thickness estimate in nm, so PH50/S/C26 is a 50% DL small-particle sample with a 26 nm coating R1. Table A lists all twelve samples.

Table A. The samples (from Table 2; a dash means the paper leaves the cell blank). SSA was given for uncoated fractions only R1.

Two features of Table A matter later. Only one sample, PH50/S/C26, was made with slow growth, and it has by far the highest oxide content (3.68%) although its thickness is not the largest R1.i·C50. And the paper's "average" regime is not defined in the Methods, which name slow, medium and fast; average presumably means medium X7.

Characterizing the coating and the solid

Coating thickness was estimated from TGA (heating 20–30 mg to 800 °C at 10 °C/min in dry air, the residue taken as oxide) together with the BET surface area, and checked by cross-sectional transmission electron microscopy (TEM) on four 50% DL samples after focused-ion-beam lift-out P5·P6·P7. Figure 1 shows the coating as a continuous band: thickness markers read 11.2 and 12.3 nm on PH50/S/C12 and 46.3 and 39.3 nm on PH50/L/C36 R2. The authors call the TGA and TEM values "broadly consistent" and the coatings "highly conformal" C11. The solid state was checked by modulated differential scanning calorimetry (DSC), XRPD, polarized light microscopy (PLM) and SHG P8·P9·P11·P12.

Note by the AI writer (not discussed in the paper): dividing each oxide content by the SSA of the matching uncoated fraction and by the TGA thickness gives the same implied oxide density, 2.95–3.08 g/cm³, for seven coated samples, but 2.47 g/cm³ for PH25/S/C11; at the common value its thickness would be about 9 nm rather than 11. TEM also gave about 8 nm for PH50/L/C13 against a TGA estimate of 13 nm X6.

Water uptake

Water uptake was followed with a dynamic vapor sorption (DVS) balance: each sample was dried at 37 °C and 0% RH under nitrogen for 120 min, then held at 90% RH while its mass was recorded continuously; Figure 2 shows the first 120 min P10. Sample mass and replicate number are not printed X17.

Release tests

Single-stage tests used 200 mL of pH 6.5, 28.65 mM phosphate buffer (FaSSIF V1 blank) at 37 °C and 100 rpm in small-volume vessels, with in-line UV probes and a dose of 8 µg/mL PCZ, for 1–3 h P13. Two-stage tests started in 100 mL of pH 1.6 simulated gastric fluid (SGF) or pH 4.0 phosphate buffer, at a dose of 16 µg/mL, for 30 min, 1 h or 4 h, before 100 mL of buffer concentrate brought the medium to pH 6.5 and the dose to 8 µg/mL P14. Large 25% DL particles (PH25/L/NC and PH25/L/C39) went through all six acid conditions; the 50% DL samples went through 30 min at pH 1.6 A5. The strong acid is labelled pH 1.2 in Figure 5, in Figure S7 and in two places of the text; the Methods and the other figures say pH 1.6 X2. To test whether a coating itself held back release, PH50/S/C26 was cryomilled for 2 min after coating and retested in both protocols P4.

Imaging crystals and coatings

SHG images were taken of three small 50% DL powders (12 s exposure, Figure S4) and of the same samples after 3 h in pH 6.5 buffer (3 s exposure, Figure 7) P12. For SEM/EDX, particles were incubated in pH 1.6, pH 4 or pH 6.5 medium for up to 4 h, collected on carbon tape, fully dried, platinum coated and imaged at 7 kV, with element maps of C, O, Al and F; Al and O mark the coating, while C, O and F mark the ASD, F being found only in the drug P15·C31. Figures 8 and 9 show 30 and 60 min; 25% DL particles were imaged in pH 1.6 and 4, 50% DL particles in pH 1.6 and 6.5 R20·R21·R22·R23. Particles that dissolved fast could not be imaged R26.

Experimental analysis

The coating changes neither the glass transition nor the amorphous state

Every sample showed a single glass transition; the average onsets were 90.4 ± 1.2 °C at 25% DL and 72.2 ± 1.4 °C at 50% DL, with no effect of coating or particle size R3·C8. XRPD showed only amorphous halos (Figure S2), PLM showed some birefringence that the authors attribute to strain rather than crystals (Figure S3), and SHG of three 50% DL powders showed only scattered points, with integrated intensities of 1.1 (uncoated), 3.3 (12 nm) and 6.1 (26 nm) at 12 s exposure (Figure S4) R4·R5·R6·C9·C10. The coating process at 35 °C therefore did not change the solid it was applied to.

Coatings slow water uptake, and one nearly stops it

Figure 2 is the first evidence that the coating is a barrier, though not a seal. Within each DL the uncoated samples levelled off at similar water contents, about 5.1–5.4% at 25% DL and 4.2–4.5% at 50% DL, which the authors read as an "approximately equivalent" equilibrium water content; small uncoated particles reached it faster than large ones "as anticipated" R7·R8·C12. Table B gives the uptake at fixed times.

Table B. Water uptake (% mass gain) at 37 °C and 90% RH, read from Figure 2 (approximate) R7·R8.

In the large 25% DL series the effect of thickness is clean: at 30 min the uncoated particles had taken up about 4.7% water, the 11, 23 and 39 nm coatings about 4.3, 3.7 and 1.2%, and the 39 nm sample was still rising at 2 h R7.c–f. The authors call this series the clearest case of slower sorption with thicker coatings C13. Across size fractions and growth regimes thickness alone does not sort the data: the 11 nm small-particle coating, grown at the average rate, sorbed more slowly (about 2.4% at 30 min) than the 11 and 23 nm fast-grown coatings on large particles, and more slowly than uncoated large particles, although small particles sorb faster when uncoated R7.b·X8. At 50% DL, PH50/L/C36 was the slowest of the samples that sorbed, and PH50/S/C26 took up only about 0.33% in 2 h; the authors read this as "a highly conformal coating which does not allow for water penetration" ("suggesting") R8·C14. In the abstract they conclude that "coating thickness and defect density (inferred from deposition rates)" set the sorption kinetics C15.

At 25% DL the coating does not slow release

Figure 3 is the paper's central release result. In the single-stage test all six 25% DL samples reached about 90% of the 8 µg/mL dose within 10–20 min; the authors describe the profiles as "consistent" with and without coating, the small particles slightly faster C16. Read from the figure, the small particles reached about 7 µg/mL in 7.5–8.5 min and the large ones in 12.5–15.5 min, coated or not, with maxima of about 7.15–7.35 µg/mL R9. Particle size, not coating, set the order. PH25/L/C23 was the one outlier: it peaked at about 6.85 µg/mL and ended at about 6.25 µg/mL with a wide error bar R9.e.

The two-stage tests repeat this result after the particles have sat in acid. After 30 min, 1 h or 4 h at pH 1.6, both PH25/L/NC and PH25/L/C39 rose to about 7.1–7.25 µg/mL within about 20–25 min of the shift, and the profiles for the three acid times coincide (Figure 4, Figure S7) R10·R13. The same held after 30 min, 1 h or 4 h at pH 4.0, with maxima of about 7.3–7.45 µg/mL (Figure S5) R12. The authors take the unchanged release after 4 h in acid to "suggest" that the solid did not crystallize during the acid stage C18. Their conclusion is that coatings, despite slower water uptake, "do not negatively impact dissolution performance" of a simple ASD such as 25% DL PCZ/HPMCAS C59.

In acid the coating delays leaching, briefly in strong acid and longer at pH 4.0

In acid HPMCAS does not dissolve, so only drug leaches, and the amounts are small. At the low pH the leached concentration of PH25/L/NC reached about 0.68 µg/mL after 1 h; at pH 4.0 only about 0.30 µg/mL R11.a·R11.c. The authors attribute the difference to drug solubility: the 16 µg/mL acid-stage dose is far below the crystalline solubility at pH 1.5 (60 µg/mL), but far above it at pH 4.0 (0.33 µg/mL) C19·M1. Table C compares the coated and uncoated samples.

Table C. Leached PCZ (µg/mL) in acid, read from Figure 5 (approximate; the low pH is labelled pH 1.2 there) and, for 4 h, from the end of the acid stage in Figure 4c (pH 1.6) and Figure S5c (pH 4.0). Values below zero are plotted as such in the paper R11·R10.e–f·R12.e–f·X2.

The coating slowed early leaching in both media C20. In strong acid the coated sample started lower but caught up at about 40 min; the results text says the concentrations "quickly became comparable" within 30–60 min, the Discussion "after approximately 30 min" R11·C20. At pH 4.0 the coated sample stayed below the uncoated one for the whole hour and through a 4 h acid stage R11·R12·C20. Note by the AI writer (not discussed in the paper): after 4 h at pH 1.6 the coated sample had leached slightly more than the uncoated one (about 1.5 against 1.25 µg/mL), so in strong acid the coating only shifts the start of leaching R10.e–f.

The authors give two readings of why the delay lasts longer at pH 4.0, and they do not fully agree. In the Results the residual difference "suggests that the erosion of the coating is slower in pH 4.0", because Al2O3 dissolves more readily at low pH, and that the changing substrate also affects coating integrity C21. In the Discussion the difference "may" come either from faster erosion at low pH or from the higher solubility of PCZ, and the second is "more likely the dominant mechanism due to the rate of coating disruption observed, although more research would be required to confirm this supposition" C42. They add that coating erosion is "unlikely" to play a major role on the time scale of these tests C46. Coating erosion was not measured X18.

At 50% DL a thin coating gave more release, with far less crystallization

At 50% DL release at pH 6.5 was much slower than at 25% DL, "consistent with its high DL" C22. Table D collects the single-stage data and the SHG result.

Table D. PCZ concentration (µg/mL) in the single-stage test at pH 6.5, read from Figure 6a,b (approximate), and SHG integrated intensity (3 s exposure) of the particles after 3 h at pH 6.5 (Figure 7) R14·R15·R18.

For the small particles, the uncoated and the 12 nm coated samples overlapped for about 40 min; then the uncoated sample slowed while the coated one kept rising, to about 5.1 against 3.35 µg/mL at 3 h R15·C24. The text says the uncoated sample "plateaued" after about 60 min; in the figure it keeps rising slowly X19. After 3 h at pH 6.5 the SHG signal of the residual uncoated particles was 126.7, against 3.7 for the 12 nm coated and 2.7 for the 26 nm coated particles R18·C28. The authors read the slowing of the uncoated sample as matrix crystallization limiting its release, and the coating as having "delayed the onset of matrix crystallization", which "increased the supersaturation attained in one set of samples" C29·C40. That the slow-sorbing PH50/S/C26 crystallized little "could be expected"; that the thin-coated PH50/S/C12, which takes up water and releases drug, crystallized equally little is "surprising", "suggesting that the coating alters the ability of the surface to crystallize rapidly" C30.

The large 50% DL particles behaved differently. Uncoated, 13 nm and 36 nm samples overlapped for about 45 min, and from then on the coated ones ran slightly below the uncoated one, about 4.5 and 4.85 against 5.2 µg/mL at 3 h; the authors call these profiles "similar" R14·C23·X10. SHG was not measured on these samples.

The two-stage test at 50% DL (30 min at pH 1.6, then pH 6.5) did not repeat the gain. The small uncoated and 12 nm samples leached alike in acid (about 1.9 and 1.8 µg/mL at about 28 min), and after the shift the coated sample stayed slightly below (about 4.05 against 4.35 µg/mL at 90 min) R17. With the large particles, the 36 nm coated sample leached more than the uncoated one in acid (about 2.9 against 2.0 µg/mL) and stayed higher after the shift (about 1.65 against 1.15 µg/mL at 90 min) R16. The authors describe the pH-shift results as following "similar release performance trends", with similar acid-stage leaching C26·X9. They note that leaching was larger at 50% than at 25% DL, "in line with expectations" as more drug sits at the surface, and that at 50% DL leaching was "less clearly altered" by the coating, "likely" because matrix crystallization competes C27·C43.

The slow-deposited coating blocks release until it is broken

PH50/S/C26 released only about 0.7 µg/mL in 3 h at pH 6.5 and about 0.85 µg/mL in the pH-shift test R15·R17. The authors weighed two explanations C49·C50. The first, that the powder had crystallized before testing, they rejected: it was X-ray amorphous, its SHG signal was low, and cryomilling restored release C49. Read from Figure S6, after 2 min of cryomilling the powder reached about 5.4 µg/mL at 1 h and 6.55 µg/mL at 3 h in the single-stage test, and in the pH-shift test it leached about 5.8 µg/mL in 30 min of acid and reached about 6.1 µg/mL 1 h after the shift R19. The authors credit both "the lack of intact coating and reduced particle size", which the experiment does not separate C25. Note by the AI writer (not discussed in the paper): the powder SHG of PH50/S/C26 (6.1) was the highest of the three powders measured at 12 s exposure, though still low; the cryomilling result carries the rejection of the first explanation more than the SHG does R6·X11.

The second explanation, that the coating itself prevented release, they accepted. PH50/S/C26 had the highest Al2O3 content (3.68%) though not the thickest coating, took up almost no water, and in SEM/EDX after 60 min at pH 6.5 its coating was smooth and intact while the uncoated particles had turned porous (Figure 9i,j) R1.i·R8·R24·C50. "The inference is then that the coating on this ASD system had a reduced level of defects" C50. Because this sample was made by slow deposition, the authors conclude that their observations "point toward faster coating regimes and thinner coatings best suiting immediate release formulations such as ASDs" C51.

What happens to the coating in each medium

The SEM/EDX images (Figures 8, 9) are the basis of the proposed mechanism C58. In strong acid, uncoated particles stayed visually intact, which the authors attribute to the low solubility of HPMCAS C32. On PH25/L/C39 and PH50/L/C36, cracks and buckling were visible after 30 min and the coating had flaked off in places; at 60 min large areas had lost their Al signal R20·R22·C32. At pH 4, where neither drug nor polymer dissolves well, PH25/L/C39 showed cracks after 30 min and cracks with buckling after 60 min, with "minimal flaking" R21. This the authors read as the key sign: "water penetrated the coating, and the resulting volume expansion due to swelling of the underlying matrix led to cracks in the coating" ("suggests") C34. The text cites Figure 9e,g for these pH 4 images, but they are Figure 8e–h X4.

At pH 6.5, uncoated 50% DL particles turned porous within 30 min; the authors describe "continuous drug-containing domains", formed "presumably" as the dissolving polymer leaves amorphous drug behind R23·C33. On PH50/L/C36 the coating had wide cracks and missing areas at 30 min, and at 60 min the exposed parts had the same porous texture as uncoated particles; the coating "did not alter the behavior of the underlying particle" R23·C33. Thin coatings disappeared too quickly to be imaged, so most images show thick coatings; the faster loss of thin coatings is reported as "data not shown" R25·X15.

Discussion

The proposed mechanism

The authors start from a hypothesis of Vogel et al. (ref 56): release from an ALC-coated material depends on the coating thickness, on voids and defects in the coating, and on the solubility of the coating in the medium C36. Water diffuses in through voids in the shell and dissolved molecules leave through the same voids; a thicker coating has fewer through-voids and a longer path C37. Al2O3 has a solubility minimum at pH 6–7 and becomes more soluble at low and high pH, so erosion "may" be faster in acid R29·C37. The ASD adds a third element: water that gets in hydrates the solid, the solid swells, and the swelling breaks the rigid film C37·C44. The authors write that the disruption is "mostly likely related to the normal stress acting upon the coating due to increase in volume of the substrate", or simply to the extra water below the surface; the equilibrium water content of about 5% at 90% RH "would be expected to be even higher" when the particle is immersed C44·R8. Wrinkling of a rigid coating over a softer amorphous material has been reported before and attributed to mismatched mechanical properties (ref 57) C45. The chain, as the authors draw it in Figure 10, is shown below.

Two features of the data support this picture. Cracking appears even at pH 4, where neither drug nor polymer dissolves well, so dissolution of the substrate is not needed to break the coating R21·C34. And the one coating that kept water out also stayed intact and blocked release, while breaking it by milling restored release R8·R24·R19. Because the coating cracked so quickly, the authors think that coating erosion is "unlikely" to matter for release on this time scale C46. In this ASD, drug and polymer have opposite pH-dependent solubility; at pH 6.5 "the high overall solubility of the polymer drives the disruption of the coating, and drug release is unhindered" C39. In acid, the drug leaches by its own solubility and the coating gives "an initial delay in hydration" before it is disrupted C41. More generally, the authors write that for amorphous materials "the release mechanisms are further complicated by the solubility of the underlying substrate" and by polymer ionization, phase separation and matrix crystallization C38.

Coating design

Thicker coatings persisted longer on the particles, "presumably due to the balance of hydration kinetics, particle swelling, and coating cracking/flaking" C47. The water uptake data fit a barrier whose performance depends on thickness and defect density, as reported for ALD alumina films on organic devices (Klumbies, ref 58) C15. The authors use this to explain why earlier water sorption studies of ALC-coated powders disagreed: some found lower sorption with thickness (Duong 2022, ref 10; Hellrup 2017, ref 19), others no change for 5–25 nm coatings (Moseson 2022, ref 11) R28. A sorption test aims at the equilibrium water content; thin coatings reach it quickly, thick but porous coatings reach it slowly, so with too short a test the water content "could be underestimated", and highly conformal, low-defect coatings, such as slow deposition "may" give, limit water uptake altogether C48.

From this the authors derive their design rule. "The coatings must be sufficiently permeable to water to facilitate rapid release"; together with the earlier finding that coatings of 5–25 nm inhibit surface crystallization of a broad range of ASDs, this "suggests that the optimal coating qualities consist of a relatively thin coating with a sufficient level of defects" C54. The Conclusion puts a number on it: a relatively thin coating, below 20 nm C54. They also note that "there is not a clear line connecting coating thickness and release performance", and that experiments on process controls and deposition rates "may shed more light" C53. Water sorption kinetics are highlighted "as a potential predictive tool" for coating quality and its effect on dissolution, and the authors call for critical quality attributes and characterization methods for the technique C52. For the ASDs tested here, provided the coating is rapidly disrupted by water uptake and swelling, coated systems release like uncoated ones, and "we hypothesize the same would be true for other immediate release formulations" C55. Conversely, "reduced water permeability may be a suitable controlled release strategy", as explored by Hellrup et al. (ref 25) C56.

Note by the AI writer (not discussed in the paper): in these data, growth regime and oxide content separate the blocking coating from the rest better than thickness does. The fast- or average-grown coatings of 23–39 nm released like or close to uncoated particles, while the slow-grown 26 nm coating blocked release R1·R9·R14·R15. A coating specification built on these results would therefore need a limit on water permeability or deposition rate, not only a thickness range. A short DVS run at 90% RH flagged the blocked sample clearly (about 0.2% uptake at 30 min against about 1.2–4.3% for the other coated samples), but with one blocked sample there is no cut-off yet R7·R8.

Limitations

Limitations stated by the authors. The authors write that coating properties affect release but "there is not a clear line connecting coating thickness and release performance" C53. They leave open whether coating erosion or drug solubility governs the acid-stage delay ("more research would be required to confirm this supposition") C42. They state that, "given the complexities of ASD formulations and their dissolution behavior, combined with a wide variety of polymers with different dissolution mechanisms, the interplay between coating properties and drug release clearly needs additional investigation" C57. Their extension to other immediate-release formulations is labelled a hypothesis C55.

The AI writer's own judgement (1): what was measured and what was inferred. The release comparison at 25% DL is direct and robust: six samples, single-stage and two-stage tests, two acid pH values and acid times up to 4 h all give the same answer R9·R10·R12·R13. The mechanism is less direct. Defect density was never measured; it is inferred from the deposition regime and the oxide content X7. Swelling and stress were not measured either, and the SEM images show particles that were taken out of the medium, fully dried and platinum coated, so shrinkage on drying could also crack or buckle a rigid film X13. The intact coating of PH50/S/C26 is documented by one particle at 60 min, although the text speaks of "prolonged periods" X14. The 50% DL crystallization benefit rests on one coated sample against its control, measured once by SHG at 3 h; the onset of crystallization was not followed, the pH-shift test did not show the release gain, and the large coated particles released slightly less than uncoated ones X9·X10. Replicate numbers and the meaning of the error bars are not printed, and the slow-sorbing coated samples had not reached equilibrium within the 2 h of water uptake shown, so equal equilibrium water within a DL is shown only for the fast samples X17.

The AI writer's own judgement (2): the evidence for the design rule. The only sample that blocked release was also the only slow-grown sample, the one with the most oxide, and a small-particle sample; the experiment cannot tell which of these mattered X7. The recommended upper limit of 20 nm is not a threshold found in the release data, since fast- or average-grown coatings of 23, 36 and 39 nm did not block release X16. The SHG comparison between powders and incubated particles is also limited: the exposures differ (12 s and 3 s), each value comes from one image, and the blocked sample's powder gave the highest of the three powder signals X12·X11.

Inconsistencies in the source. Section 3.3.1 is headed "Release Performance 50% DL ASDs" but reports the 25% DL data X1. The strong acid is pH 1.6 in the Methods and most figures but pH 1.2 in Figure 5, Figure S7 and two places of the text; the leached concentrations at 1 h agree between Figure 5 and Figure 4b, which suggests one medium X2. The pH 4.0 two-stage figure, cited as Figure S5, is printed in the SI with the label "Figure S6", and the cryomilling figure carries the same label X3. The pH 4 SEM images are cited as Figure 9e,g but appear in Figure 8e–h X4. The Discussion calls Al3+ the dominant solution species at pH 7 and, one sentence later, soluble Al(OH)3 the majority species at neutral pH X5. And the Results and the Discussion give different causes for the longer leaching delay at pH 4.0 X18. None of these changes the main release comparison.

Conclusion

The paper asks whether an Al2O3 atomic layer coating, already shown to keep ASD particles from crystallizing at the surface, costs release. For hot-melt-extruded PCZ/HPMCAS particles the answer is no, as long as water can get through the coating. Coatings of about 11–39 nm slowed water uptake, but coated 25% DL particles released about 90% of the dose within 10–20 min at pH 6.5, as fast as uncoated particles, also after up to 4 h in acid R7·R9·R10·R12·C59. In acid the coating delayed leaching only briefly in strong acid and for longer at pH 4.0 R11·C20. At 50% DL a 12 nm coating gave more release in the single-stage test, together with far less crystallization in the wetted particles R15·R18·C40. Only the slow-deposited coating with near-zero water uptake blocked release, and milling it restored release R8·R15·R19·C50. SEM/EDX shows the coatings cracking, buckling and flaking as the particles hydrate, even at pH 4 where little dissolves R20·R21·R22·R23.

From these observations the authors propose that water penetrates the coating through defects, the particle swells, and the cracked coating lets drug and polymer out; they recommend thin coatings (below 20 nm) with enough defects, and water sorption kinetics as a test of coating quality C58·C54·C52. Defect density, swelling and erosion were not measured, and the blocked-release case rests on one sample, so the design rule is a working hypothesis for further process studies rather than a measured threshold X7·X13·X16.