Does drug–polymer hydrogen bonding help an amorphous solid dispersion release? Indomethacin and its methyl ester in PVPVA

Background and motivation

Amorphous solid dispersions (ASDs), in which a poorly soluble drug is dispersed at the molecular level in a polymer, are among the most used ways to raise solubility. They carry two well-known risks: the drug can crystallize, in the matrix or from the supersaturated solution, and the dispersion can split into a drug-rich and a polymer-rich amorphous phase (amorphous–amorphous phase separation, AAPS). The introduction calls AAPS a "likely precursor" to matrix crystallization that "may also impact" dissolution (Mosquera-Giraldo 2018; Purohit 2015; Knopp 2018; Purohit 2017; Saboo 2019). C1 For storage stability, specific drug–polymer interactions are the usual defence: drug–polymer hydrogen bonding is "generally considered advantageous" to physical stability, and dispersions with stronger hydrogen bonding "appear to be more resistant" to water-induced AAPS under high humidity (Matsumoto 1999; Rumondor 2009; Rumondor 2011; Kothari 2015). C5

Release is the less understood side. For dispersions in poly(vinylpyrrolidone-co-vinyl acetate) (PVPVA, copovidone), several studies found that at low drug loading (DL) drug and polymer dissolve at the same rate, so release is controlled by the polymer; this is called congruent release (Simonelli 1969; Corrigan 1985; Chen 2016). Above a loading that differs between drug–polymer pairs, drug release drops abruptly and becomes drug-controlled and incongruent (Saboo 2019; Indulkar 2019). C2 The highest loading that still releases congruently is the limit of congruency (LoC), and the sudden fall above it is described as "falling off the cliff" (Indulkar 2019). C2·C40 The threshold matters for two reasons. Polymer-controlled release typically drives the drug above its amorphous solubility, so drug-rich nano-sized aggregates form by liquid–liquid phase separation (LLPS), and these "are thought to" enhance absorption; and a higher workable loading means smaller tablets and a lower pill burden (Ilevbare 2013; Raina 2014; Wilson 2018; Stewart 2017). C3

What sets the LoC was not known. In the group's previous PVPVA study, two drugs with substantially different logP had the same LoC, a result the authors call "somewhat counterintuitive" (Saboo 2019). C4 This paper turns to drug–polymer hydrogen bonding. The authors chose a minimal pair: indomethacin (IND), whose carboxylic acid hydrogen bonds to the pyrrolidone carbonyl of PVPVA (Taylor 1997; Yuan 2015), and indomethacin methyl ester (INDester), in which the acid OH is replaced by a methoxy group, so that no specific interaction with the polymer is possible (Figure 1). C6·M1·M2 The "primary hypothesis under evaluation" was that hydrogen bonding would favour release: IND dispersions would resist AAPS during dissolution and release congruently up to a higher loading than INDester dispersions. C7 Two further factors entered as the work developed: INDester crystallized on the dissolving surface, and the authors changed the dissolution temperature to probe the role of the glass transition temperature (Tg) relative to the experiment. A3·A4

Background concepts

Congruent release and the normalized rate. The paper measures release from a compressed tablet of fixed exposed area rotating in the medium, so that drug and polymer can be compared directly. Each component's release rate is normalized by the exposed area and by its fraction in the dispersion (eq 1):

Here $R$ is the surface-normalized release rate (mg/min/cm²), $k$ is the slope of the linear regression of dissolved concentration against time (µg/mL.min), $V$ is the volume of medium (100 mL), $S$ is the exposed tablet area (0.5 cm²) and $x$ is the mass fraction of the component in the dispersion. E1·P5 In plain terms: the rate is expressed per unit area and per unit of that component in the tablet, so if drug and polymer leave together, their normalized rates are equal, whatever the loading. Congruent release therefore shows as two equal bars, and the LoC is the highest tested loading at which they stay equal. Loadings were tested in 5% steps, so "LoC ~10%" means that 10% was congruent and 15% was not. C14·P4

Amorphous solubility, supersaturation and LLPS. The amorphous solubility is the highest concentration the amorphous drug can sustain in a medium; above it the drug separates into a drug-rich liquid or glassy phase, here detected as nano-sized drug-rich aggregates by nanoparticle tracking analysis (NTA) and as turbidity. The crystalline solubility is lower. The ratio of the two is the maximum supersaturation an amorphous form can give against the crystal. R1·P2·P10

Glass transition, wet Tg and mobility. Water plasticizes amorphous materials. The paper uses the "wet Tg", the Tg of the material saturated with water. Above its Tg a material is a mobile supercooled liquid; below it, a glass. For neat IND the wet Tg was estimated rather than measured: the water content of water-saturated amorphous IND was extrapolated from a sorption isotherm at 37 °C, and Tg was lowered by 10 °C for each 1% of water sorbed, an empirical rule taken from earlier work on indomethacin (Andronis 1997). E6·P22·M1 For the 70% DL IND–PVPVA dispersion, which the authors use as a model of the drug-enriched surface, the wet Tg was measured by DSC after equilibration at 100% relative humidity (RH). P23·R30

Polymer dissolution. The discussion draws on polymer science: a polymer dissolves by solvent uptake and chain disentanglement; above Tg the plasticized polymer forms a gel layer from which chains disentangle and diffuse away, while below Tg the gel layer thins or disappears and dissolution switches from disentanglement to "cracking", in which small blocks of polymer break off (Miller-Chou 2003; Raptis 1969). C54

Gordon–Taylor equation. To turn a local Tg into a local composition, the paper uses the Gordon–Taylor equation (eq 3):

Here $T_{g,\mathrm{m}\mathrm{i}\mathrm{x}}$ is the Tg of a homogeneous mixture, $w_1$ and $T_{g1}$ the weight fraction and Tg of IND, $w_2$ and $T_{g2}$ those of PVPVA, and $k$ a fitted constant. E3 In plain terms: the Tg of a miscible mixture lies between those of its components, weighted by composition and by the constant $k$; once $k$ is fitted on mixtures of known composition, a measured Tg can be inverted to give the drug fraction.

Crystallization tendency. The drugs were classified by a DSC protocol (Baird 2010): Class 1 crystallizes on cooling the melt at 20 °C/min, Class 2 stays amorphous on cooling but crystallizes on reheating at 10 °C/min, and Class 3 stays amorphous through both. IND is Class III, INDester Class II, so INDester is the faster crystallizer. P20·M1·M2

Surface techniques and their depth. Three methods estimate the drug content of a partially dissolved surface, each sampling a different depth. Attenuated total reflectance infrared (ATR-FTIR) probes a depth given by (SI eq S2):

Here $d_p$ is the penetration depth, $\lambda$ the wavelength, $n_1$ the refractive index of the ATR crystal (diamond, 2.4), $n_2$ that of the sample (taken as 1.2–1.5) and $\theta$ the angle of incidence (45°). E5·P24 In plain terms: longer wavelengths and a higher sample refractive index give deeper sampling; at the bands used here the paper lists 1.22–2.17 µm. R32 Note by the AI writer: the two values for $n_2$ = 1.2 (2.11 and 2.17 µm) cannot be reproduced from the printed equation, which gives about 0.81–0.83 µm, while the $n_2$ = 1.5 values (1.22 and 1.25 µm) reproduce; by the equation the ATR depth is about 0.8–1.3 µm. X4 Energy-dispersive X-ray analysis in the SEM (EDX) samples 600–800 nm by Monte Carlo simulation, and AFM-based nanoscale thermal analysis (nanoTA), in which a heated tip sinks into the surface at its softening point, senses about 50 nm. R33·R34

Methods in detail

The study combines release measurements with an unusually broad set of surface and solution analyses. It can be read as three contributions: a matched release comparison of the two drugs, an orthogonal measurement of what the dissolving surface becomes, and a temperature series that places the dissolution temperature above or below the wet Tg. The flow below summarizes the work.

Contribution 1: a matched release comparison of a hydrogen-bonding and a non-bonding drug

Problem. To isolate the effect of hydrogen bonding, the two drugs must differ in that one feature and be tested so that drug and polymer release can be compared at each loading. A1·C6

What they did. IND was bought and INDester was synthesized from IND with benzenesulfonic acid in methanol at about 50 °C; NMR showed the methyl ester signals at δ 3.70 ppm (1H) and 52.11 ppm (13C) and a purity above 95%. M1·M2 Dispersions of 5, 10, 15, 20 and 25 wt% drug in PVPVA (Kollidon VA 64) were made by rotary evaporation from dichloromethane:methanol 1:1 (IND) or tetrahydrofuran (INDester), dried, cryo-milled, sieved to 106–250 µm and checked to be amorphous by XRD and polarized light microscopy (PLM). P4·M3 The two drugs were dissolved in different media: IND in 100 mM phosphate buffer at pH 2.1, where it is unionized (pKa 4.5), with 200 µg/mL of the crystallization inhibitor HPMC-606 pre-dissolved; INDester in 100 mM phosphate buffer at pH 6.8, to avoid ester hydrolysis, with 1000 µg/mL HPMC-606 because it crystallizes faster. M5·M6·M4·M1·C8 In these media the crystalline and amorphous solubilities were 1.3 and 22.1 µg/mL for IND and 0.6 and 3.9 µg/mL for INDester (n = 3), and control experiments showed that neither HPMC-606 nor PVPVA up to 1 mg/mL changed the equilibrium solubility. R1·R2·C10

For release, 100 mg of dispersion was compressed at 1500 psi in an 8 mm die, leaving 0.5 cm² exposed, and rotated at 100 rpm in 100 mL of medium at 37 °C, non-sink with respect to the amorphous solubility. P5 Samples of 2 mL were taken every minute for 5 min and replaced with fresh buffer; drug was assayed by reverse-phase HPLC and PVPVA by size-exclusion HPLC, and rates were calculated by eq 1 from the slopes (n = 3). P5·P3 A second series ran for 30 min, and after 30 min the medium was examined by NTA and photographed for turbidity. P6·P10

Why it works. The rotating disk keeps the exposed area constant and removes particle size and shape effects, and normalizing by each component's fraction makes congruent release visible as equal rates. P5·E1 Measuring the medium after 30 min shows whether a congruent dispersion actually reaches the supersaturated, nanodroplet-forming state that motivates the work. C15·C16

How it differs from earlier work. The group's previous study compared drugs that differed in logP; here the pair differs by a single functional group that controls hydrogen bonding to the polymer. C4·C6 Note by the AI writer (not discussed in the paper): the methyl ester also changes much else, which matters for interpreting the result; see Limitations. X6

Contribution 2: three orthogonal measurements of the dissolving surface

Problem. If incongruent release arises because polymer leaves the surface faster than drug, the surface should become drug-rich. Measuring the composition of a thin, wet and changing surface layer is difficult, so the authors used three methods with different principles and depths. C43·C44

ATR-FTIR. In amorphous IND the carbonyl region shows bands at 1734, 1705 and 1678 cm⁻¹ (free acid, cyclic-dimer acid and amide carbonyl, as assigned by Taylor and Zografi 1997) and an IND-specific aromatic band at 1591 cm⁻¹; PVPVA shows 1730 (vinyl acetate) and 1668 cm⁻¹ (vinyl pyrrolidone carbonyl). R13 In IND–PVPVA dispersions of 10–70% DL the dimer band at 1705 cm⁻¹ disappears and a band at 1634 cm⁻¹, assigned to hydrogen-bonded pyrrolidone carbonyl, appears and grows with loading, "indicating" more IND–PVPVA hydrogen bonds; the 1591 cm⁻¹ band grows continuously with loading and serves as a marker of drug content. R14·C20 Tablets were taken out after dissolution, blown dry, vacuum-dried overnight and measured; the surface composition of the 25% DL tablet was estimated by matching its spectrum to those of dispersions of known loading (Figure S8). P11·P12

SEM-EDX. IND (C19H16ClNO4) carries one chlorine and one nitrogen, while PVPVA has one nitrogen per copolymer unit and no chlorine. The authors therefore relate the Cl/N atomic ratio to the mole fraction of drug (eq 2):

Here $n_{\mathrm{d}\mathrm{r}\mathrm{u}\mathrm{g}}$ and $n_{\mathrm{p}\mathrm{o}\mathrm{l}\mathrm{y}\mathrm{m}\mathrm{e}\mathrm{r}}$ are the moles of drug and of polymer units in the sampled volume. E2·C25 In plain terms: every drug molecule adds one Cl and one N, every polymer unit one N, so the percent Cl/N ratio follows the mole percent of drug, which is then converted to weight percent with the drug molecular weight and that of an idealized polymer unit. A calibration on tablets of 10, 30, 50, 70 and 90 wt% IND was linear (Table S1, Figure S11), and the 25% DL tablet surface was measured before and after 5, 10 and 20 min of dissolution (n = 3). R20·P14 Note by the AI writer: the conversion factor is not stated. With IND 357.8 g/mol and the printed PVPVA unit C10H15NO3 (197.2 g/mol), 10 wt% would be 5.8 mol% rather than the printed 6.1 mol%; all the printed mole fractions reproduce with about 209 g/mol per polymer unit, and so do the weight percentages of Table 4. X3 In Table S1 the rows labelled "IND alone" and "PVPVA alone" carry each other's values, and the SI text lists standards of 10–50% that differ from the table's 10–90%. X2

AFM-nanoTA on films. Films about 100–200 nm thick of 25% DL dispersion were spin-coated on ZnS in a dry glovebox (about 18% RH) so that they started miscible, immersed in about 5 mL of buffer for 2 s to 20 min, blown dry and vacuum-dried. P15·P16 AFM gave 5 × 5 µm height images, and nanoTA gave the local Tg of the remnants. A Gordon–Taylor calibration on homogeneous films (IND 53 °C, PVPVA 130 °C, five loadings between; k = 0.91) converted each local Tg into a drug content. P17·P18·R27·E3 The authors note that these nanoTA values lie above calorimetric Tg values, "likely" because of the technique. C30 For the INDester film, the AFM cantilever's Lorentz contact resonance (LCR) frequency, higher on the bare stiff substrate than on the polymer film, was followed to find when the film had fully dissolved. P19·C29

Why it works. The three methods measure different things (vibrational bands, element ratios, local softening) over depths from about 50 nm to about 1 µm, so agreement between them is stronger evidence than any one alone. R32·R33·R34·C44

Contribution 3: dissolution temperature against the wet Tg

Problem. Removing the OH group also changes the Tg drastically: the Tg onset is 43.5 °C for IND and 1.7 °C for INDester. So Tg, and not only hydrogen bonding, might explain the different release. C50·M1·M2

What they did. IND tablets were dissolved for 5 min at 10, 17, 25 and 57 °C as well as 37 °C; at 10 °C only the 5% DL dispersion is reported. P7·R29 INDester could not be tested below its Tg (about 2 °C) because the medium would freeze. C9 A 25% DL INDester tablet was instead dissolved at 15 °C for 60 min to suppress surface crystallization. P7·R10 To decide which Tg is relevant, the authors estimated the wet Tg of neat IND (29 °C) and measured the wet Tg of a 70% DL IND–PVPVA dispersion, chosen to match the drug-enriched surface, after equilibration at 100% RH and 37 °C (constant mass by 4 h), by modulated DSC. M1·P23·R30

Why it works. Over a range of temperatures that all lie above a Tg, a change in LoC would point to temperature itself; a change only when the temperature crosses the Tg points to the glass transition. C53

Experimental analysis

Main result 1: the hydrogen-bonding drug has the lower limit of congruency. Over the first 5 minutes, neat PVPVA dissolved at about 3.4 mg/min/cm² in the IND medium and about 3.8 in the INDester medium (the text gives ~3.5), around 1700 times faster than amorphous IND alone (0.002 mg/min/cm²); amorphous INDester alone gave concentrations below the HPLC detection limit. R3 Table A lists the normalized rates. For IND, drug and polymer matched each other and the neat polymer at 5 and 10% DL, but from 15% DL the IND rate fell to the level of amorphous IND and the PVPVA rate fell to about 0.5–0.7 mg/min/cm², a fifth or less of its neat value. For INDester, drug and polymer matched at every loading up to 25%. R3·R4 No crystals were found on any tablet surface within these 5 minutes by XRD or PLM. R5 The authors read this as polymer-controlled, congruent release for IND at ≤10% and INDester up to 25%, and as drug-controlled release for IND at ≥15%, with the lower PVPVA rate "most likely" caused by a drug-rich interface hindering polymer release from below. C11·C12·C13 Their summary: IND–PVPVA dispersions have an LoC of ~10%, INDester–PVPVA dispersions release congruently up to 25%. C14

Table A. Surface-normalized release rates over the first 5 min (mg/min/cm², mean of n = 3), read from Figure 2 (approximate); the neat-IND value is from the text. IND in pH 2.1 and INDester in pH 6.8 medium, 37 °C, 100 rpm. R3

Main result 2: only congruent release supersaturates. Over 30 minutes (Figures 3 and 4) the picture extends. Table B gives the release at 30 min and the plotted amorphous-solubility level for each loading. R6·R8 IND at 5% DL reached about 51% release and crossed its amorphous-solubility level between 25 and 30 min; at 10% DL it reached about 35% and crossed the level between 15 and 20 min. At 15–25% DL less than 1% of the IND dissolved and PVPVA reached only about 6–8%, with the polymer release slowing over time. R6 NTA frames showed scattering centres only in the media from the 5 and 10% DL IND dispersions, and those media were turbid without birefringence; the aggregates measured 250–350 nm. R7·R12 The authors read the turbidity as LLPS rather than crystallization. C19 Every INDester medium exceeded the amorphous solubility of INDester (3.9 µg/mL) and showed particles. R8·R7·C16 No IND tablet crystallized within 30 minutes. R11

Table B. Percent of drug and polymer released after 30 min, with the amorphous solubility expressed as % release of the drug in the tablet, read from Figures 3 and 4 (approximate). R6·R8

Main result 3: INDester fails later, by crystallizing at the surface. INDester and PVPVA released together for the whole 30 minutes only at 5% DL. At 10, 15, 20 and 25% DL the INDester release levelled off from about 20, 7.5–10, 10 and 12.5–15 minutes while PVPVA kept dissolving, at a reduced rate. R8 XRD showed no crystals on any INDester tablet at 5 min but Bragg peaks at 30 min for 10–25% DL and none for 5% DL (Figure S3A); PLM and SEM of the 25% DL surface showed no crystals at 5 min, scattered needles at 10 min and a crystal-covered surface by 20 min (Figure 8, Figure S3B); and the band of crystalline INDester at 1607 cm⁻¹ appeared in the surface IR spectrum at 20 min but not at 5 min (Figure S10). R9·R22·R18 INDester showed no sign of hydrogen bonding to PVPVA in the infrared. R18·C23 The authors attribute the loss of congruency to "drug crystallization at the compact surface". C17 The key control is the 15 °C experiment: the 25% DL INDester tablet released drug and polymer together for 60 minutes, reaching about 84% PVPVA and 88% INDester, with all points above the amorphous solubility from 5 minutes. R10 This "lends support" to surface crystallization rather than to the alternative that a loss of congruency left too little polymer at the interface to inhibit crystallization. C18 Note by the AI writer: the text gives the crystallizing range once as "drug loadings >5%" (results) and once as ">10%" (discussion); Figure 12A ("10–25% DL"), the 10% DL XRD pattern and the levelling-off at 10% DL all support the first reading. X1

Main result 4: the IND surface becomes drug-rich and rough. After 5 minutes of dissolution the 10% DL IND tablet gave the same IR spectrum as before, while the 15% DL tablet showed stronger 1634 and 1591 cm⁻¹ bands; for the 25% DL tablet both bands grew from 5 to 10 to 20 minutes and the dimer band at 1705 cm⁻¹ never appeared. R15·R16 The authors read this as drug enrichment of the surface, with the drug still hydrogen bonded to PVPVA and some polymer still present ("it is reasonable to assume"). C21·C22 Table C gives the surface drug content of the 25% DL dispersion by the three methods. The fresh tablet held 25 ± 2 wt% drug by EDX and the fresh film 23 ± 7 wt% by nanoTA; after 5 minutes the estimates were 40–50%, 41 ± 3 and 44 ± 5 wt%, and after 20 minutes about 60%, 72 ± 9 and 60 ± 5 wt%. R17·R21·R28 The local Tg of the film fell from 111 ± 6 °C to 95 ± 4, 88 ± 6 and 82 ± 3 °C at 5, 10 and 20 minutes. R28 The authors call the agreement between methods with different principles and depths "compelling evidence" of surface drug enrichment. C44

Table C. Surface composition of the IND:PVPVA 25:75 dispersion during dissolution (tablets) or buffer immersion (films), from Tables 3, 4 and 5 and their summary in Table 6 (mean ± SD, n = 3; FTIR values are best matches to standards of known loading). R17·R21·R28

The topography follows the composition. Fresh tablets of both drugs were smooth; after 5 minutes the 10% DL IND tablet was still smooth, while the 15 and 25% DL tablets were rough and porous (Figure 7). R19 On the 25% DL IND film, AFM showed roughness from 2 seconds, with height features growing from about 15 nm at 2 s to about 150 nm at 20 min (Figure 9), although the film looked unchanged by eye. R24·R23 The authors explain both by faster polymer release leaving an uneven drug-rich phase ("presumably"; "can be explained"). C24·C28 The INDester film behaved differently: it stayed smooth within ±12.5 nm, produced a cloudy plume around it in the buffer, and its LCR resonance rose from 119.1 kHz to the bare-substrate value of about 121.5 kHz by 2 minutes, which the authors read as complete dissolution of the film. R25·R23·R26·C27·C29

Main result 5: the IND limit falls below the wet Tg of the drug-rich layer. Table D collects the temperature series. The IND LoC was 10% DL at 57, 37 and 25 °C, 5% at 17 °C, and below 5% at 10 °C, where even the 5% DL dispersion released incongruently; no partially dissolved tablet had crystallized. R29 The water-saturated 70% DL IND–PVPVA dispersion took up about 6% water and had a Tg onset of 18 °C (midpoint about 23 °C) against 63 °C dry (Figure S16), about 45 °C lower. R30·C34 The estimated wet Tg of neat IND is 29 °C. M1 Using the 70% DL value as the relevant one, the authors place the drug-rich phase in the supercooled-liquid state at 25, 37 and 57 °C, "slightly below" the Tg onset at 17 °C, and in the glassy state at 10 °C. C32 The LoC thus stayed constant over a span of more than 30 °C above the wet Tg and changed only when the temperature crossed it. C53

Table D. Limit of congruency of IND–PVPVA dispersions against dissolution temperature (Figure 11) and the authors' placement of the drug-rich phase relative to its wet Tg (Section on dissolution at 10–57 °C; Figure S16). R29·R30·C32

Comparison of the two failure modes. Although INDester crystallized, it released far more drug than IND at the same loading and time: at 15–25% DL after 30 minutes about 8.4–12.8% of INDester had dissolved against about 0.1–0.6% of IND (Table B). R6·R8 The authors conclude that here crystallization was "less detrimental" than AAPS, which "highlights the extremely deleterious effects of drug surface enrichment". C46 After the INDester surface crystallized, the solution stayed on a plateau above both the amorphous and the crystalline solubility, with no measurable desupersaturation; the authors infer that crystallization started at the hydrated tablet surface rather than from solution, and that these surface crystals could not grow enough to consume the supersaturation. R8·R1·C36·C37

Discussion

Two routes to failure. The authors' discussion separates the two drugs. At low loading both release congruently and "appear" polymer-controlled, at least over short times; at higher loading or longer times polymer leaves faster than drug, but for different reasons. C35 The chain they propose can be drawn as follows.

For INDester the authors call the behaviour "relatively straightforward to explain". The drug first releases with the polymer at all loadings up to 25%; then crystals nucleate and grow on the solid–liquid interface, and drug release depends on the balance between dissolution from the remaining amorphous surface and surface crystallization. Once the surface is covered, the solution sits on a plateau above both solubilities, so there is no driving force for further dissolution, and the crystal layer impedes polymer release. C36 Since the solution did not desupersaturate, crystallization was "not triggered by the evolving supersaturation"; rather "the polymer is unable to prevent crystallization at the compact interface following hydration". C37 The susceptibility "may arise, in part" from the low Tg of INDester, and the competition between rapid polymer-controlled release and matrix crystallization "appears to show a dependency on drug loading". C38·C39

For IND the authors follow the competition picture of earlier work (Indulkar 2019; Purohit 2017; Saboo 2019): AAPS in the hydrating matrix "would be expected" to release the polymer faster and deplete it at the interface; drug release is then governed by the drug-rich phase, which dissolves about three orders of magnitude slower than the polymer; by mass balance the surface enriches in drug and forms a barrier through which polymer must diffuse, which "likely explains" why the polymer release slows with time (Figure 3). C40·C41 They weigh an alternative, loss of polymer from the surface without AAPS, and favour phase separation because the surface roughened over about 500 nm, which simple polymer loss "would not be anticipated" to produce. C42 Figure 12B summarizes this: no change at low loading, increasing surface drug enrichment with time at high loading. C43

Hydrogen bonding. The study was designed to show that hydrogen bonding helps. The release data over short times and at low temperature "clearly demonstrate that our hypothesis appears to be incorrect": the IND dispersions have a much lower LoC than the INDester dispersions, and the hydrogen bonds present in the dry IND–PVPVA dispersion "did not have a positive impact on the LoC". C48 The conclusion section words it as hydrogen bonding that "appears to be somewhat detrimental to dissolution performance". C59 The authors immediately add that a general trend needs additional compounds with different hydrogen-bonding abilities. C49 Note by the AI writer (not discussed in the paper): the infrared data show that IND stays hydrogen bonded to PVPVA in the drug-enriched surface (the 1634 cm⁻¹ band grows, the dimer band stays absent), so hydrogen bonding did not keep the polymer at the interface. R16·C22 The paper does not explain why the hydrogen-bonded pair would separate at the surface more readily than the non-bonded one; it reports the outcome.

Glass transition. Because the methyl ester also lowers the Tg drastically, the authors ask whether Tg rather than hydrogen bonding explains the difference. C50 The IND temperature series "suggests that there may be some impact of the drug Tg on the LoC, whereby drugs in the glassy state may achieve lower limits of congruency, relative to those in the supercooled liquid state". C51 Because the drug-rich surface still contains considerable polymer (about 30%, from Table 6), "plausibly, a Tg effect on the LoC might be better linked to the wet Tg of the drug-rich ASD surface layer" (about 70% drug, 18 °C) than to the wet Tg of neat drug. C52 The authors connect this to polymer dissolution: the ease of chain disentanglement is "expected to depend" on the water absorbed to form the gel layer, which falls as loading rises and "may explain" polymer-controlled release at low loading, and on the Tg of the hydrated matrix relative to the temperature; lower temperature impedes chain mobility and disentanglement, which "may explain" the lower LoC. C55 A literature value fits this: ledipasvir–PVPVA, with a dry Tg of about 160 °C, has an LoC of about 5% (Que 2019). C56·R35 An alternative, that a lower temperature only changes the relative rates of surface phase separation and dissolution, is "not supported" by the low-temperature INDester run, and "the role of Tg on release properties, if any, needs further investigation". C57

Table E. LoC of PVPVA dispersions cited in the discussion (Bhardwaj 2018; Indulkar 2019; Saboo 2019; Que 2019, with Tg values from refs 6, 40, 41, 42), with this paper's two drugs; donor status as stated by the authors ("all with the exception of itraconazole"). R35·C58·M1·M2·R3

The authors note that most of these drugs carry hydrogen-bond donors and have dry Tg values that "most likely" leave them supercooled liquids at 37 °C in water, except ledipasvir. C58 Note by the AI writer (not discussed in the paper): in Table E the donor-free itraconazole does not stand out, so the literature set neither supports nor contradicts a hydrogen-bonding trend; the low-Tg, donor-free INDester is the only high value, and its LoC is confounded with surface crystallization over longer times.

What it suggests for formulation. The authors' practical message is that "enrichment of the surface with amorphous drug needs to be avoided through appropriate formulation design", because the neat amorphous drug dissolves so much more slowly than the polymer, and that the drug releases at essentially the neat polymer rate when release is congruent. C47 They expect the findings to "aid in the development of optimized ASD formulations" and plan to test other drug–polymer pairs. C64 Notes by the AI writer (not stated in the paper): a drug–polymer hydrogen bond chosen for storage stability does not guarantee release in PVPVA; a 5-minute congruency test can pass a formulation that fails after 10–20 minutes by surface crystallization, as INDester did; HPMC-606 in the medium at 1000 µg/mL did not stop crystals forming on the tablet surface, so inhibition inside the hydrated surface layer is what counts; and a dispersion that releases well at 37 °C may not do so in cooler conditions if its drug-rich layer sits near its wet Tg. R3·R8·R9·M6·R29

Limitations

Limitations stated by the authors. They state that "additional compounds with different hydrogen bonding abilities need to be studied" to know whether a general trend exists, and that a greater diversity of compounds, including higher-Tg drugs, should be evaluated. C49 They acknowledge that replacing the hydroxyl by a methoxy group also changes the melting point and Tg substantially, so the Tg difference may have contributed. C50 They could not test INDester below its Tg because the medium would freeze. C9 They describe the role of Tg on release as needing further investigation. C57 And they note that nanoTA Tg values are higher than calorimetric ones, "likely" because of the technique. C30

The AI writer's own judgement (1): what the comparison can and cannot show. The release comparison itself is clean: triplicate rates, consistent 5-minute and 30-minute data, and NTA and turbidity that agree with them. R3·R6·R7·R8 But the general claim about hydrogen bonding rests on a single pair that differs in more than one property: Tg onset 43.5 against 1.7 °C, melting point 160.7 against 91.4 °C, crystallization class III against II, and amorphous solubility 22.1 against 3.9 µg/mL. X6·M1·M2·R1 The two drugs were also dissolved in different media, pH 2.1 with 200 µg/mL HPMC-606 against pH 6.8 with 1000 µg/mL, and neat PVPVA itself dissolved at about 3.4 and about 3.8 mg/min/cm² in them. X5·R3 The finding is therefore best read as "for this pair, under these conditions, the hydrogen-bonding drug had the lower LoC". The table-of-contents graphic lists surface crystallization among the factors that lower the LoC; in the data, crystallization ended congruent release only after the first 5 to 20 minutes and did not lower the 5-minute LoC of INDester. X7·R3·R8

The AI writer's own judgement (2): how firm the mechanisms are. AAPS at the dissolving IND surface, the core of the proposed mechanism, is inferred: the paper shows surface drug enrichment and roughening, but no measurement resolves two amorphous phases at the interface, and the ~500 nm length scale used to favour phase separation over simple polymer loss was not measured (AFM height ranges reach ±150 nm, and SEM pores were not sized). X8·X14 The temperature argument rests on LoC values in 5% steps without the underlying rates, on an estimated wet Tg for neat IND, and on the measured wet Tg of a bulk 70% DL dispersion standing in for the surface layer; the statement that 17 °C lies "slightly below" this Tg depends on using its onset (18 °C) rather than its midpoint (about 23 °C). X9·X16·R30 The three surface methods sample very different depths and, for nanoTA, films rather than tablets; their estimates at 20 minutes range from about 60 to 72 wt% drug, which is agreement within the stated spreads rather than identity. X15

The AI writer's own judgement (3): reporting issues in the source. Several printed details do not reconcile. The INDester crystallization range is given as above 5% and above 10% DL in different places. X1 In SI Table S1 the labels of the pure IND and pure PVPVA rows are interchanged, and the SI text lists calibration standards (10–50%) that differ from the table (10–90%). X2 The weight-to-mole conversion behind Tables S1 and 4 is not stated and implies about 209 g/mol per polymer unit rather than the 197.2 g/mol of the printed formula; the effect on Table 4 is about 1–2 wt%, within its standard deviations. X3 The ATR depths for $n_2$ = 1.2 cannot be reproduced from eq S2. X4 The IND pKa in Table 1 is cited to a paper on ibuprofen. X13 The SEM captions state scale bars and magnifications that differ from the instrument data bars on the images, and Table 5 labels a film measurement "initial tablet". X12·X11 None of these changes the main conclusions, but they matter for anyone reusing the numbers.

Conclusion

The paper set out to show that drug–polymer hydrogen bonding raises the limit of congruency and found the opposite for its pair: hydrogen-bonded IND dispersions released congruently only up to about 10% drug loading, while the INDester dispersions did so up to at least 25% over the first five minutes. R3·C14·C48 It identified two ways release fails: incongruent release in which the polymer leaves first and leaves a slowly dissolving, drug-enriched amorphous surface, documented by three surface methods, and surface crystallization, which stopped INDester release after 10–20 minutes and could be avoided at 15 °C. C62·R17·R21·R28·R8·R10 It also showed that the IND limit drops when the dissolution temperature falls below the wet Tg of the drug-enriched composition, which the authors link "plausibly" to the mobility of the surface layer while calling the role of Tg in need of further study. R29·R30·C52·C57 Because the two drugs differ in Tg, crystallization tendency and test medium as well as in hydrogen bonding, the general rule remains to be tested on more compounds, as the authors themselves say. X6·X5·C49