Where do the nanodroplets of a dissolving amorphous solid dispersion come from? An isotope-labelling test with nifedipine

Background and motivation

An amorphous solid dispersion is a poorly water-soluble drug dispersed at the molecular level in a polymer. Its value for oral delivery rests on two things: the amorphous drug is more soluble than the crystal, and the small drug particles that form when the ASD dissolves give a large surface area. The authors call nanoparticle formation during dissolution "paramount" for in vivo performance (Introduction). C1 Unlike micelles or complexes, which raise the total amount of dissolved drug but not the free drug concentration, an amorphous form raises the free drug concentration above the crystal solubility, and passive membrane flux grows linearly with free drug concentration, also into the supersaturated range. C2

That supersaturation has a ceiling. Once the drug concentration passes the amorphous solubility, the excess drug separates out as water-saturated, drug-rich nanodroplets while the bulk stays at the amorphous solubility. This is liquid–liquid phase separation (LLPS), or glass–liquid phase separation (GLPS) when the separated drug-rich phase is a glass (Raina 2014, Ilevbare 2013, Mosquera-Giraldo 2015). C3 An earlier membrane-transport study from the same group showed that the nanodroplets do not raise flux further, but hold it at the maximum until they are used up, because they are small, amorphous and pre-wetted and can quickly replace absorbed drug. The authors call this the nanodroplet reservoir effect and conclude that it is "desirable" for ASDs to form nanodroplets when they dissolve (Indulkar 2016; Taylor and Zhang 2016). C4·C5 Nanodroplets have indeed been seen on dissolution of several ASDs, more often at low drug loading and depending on the polymer (Alonzo 2011; Jackson 2015; Harmon 2016). C6

The open question was where the nanodroplets come from, and the paper sets out two mutually exclusive answers (Figure 1). C7

• Mechanism I (proposed in the group's earlier papers, Ilevbare 2013, Alonzo 2011, Jackson 2015): drug and polymer leave the ASD together, congruently or nearly so; in the stagnant boundary layer or the bulk they part, because the polymer prefers water; where the drug concentration exceeds the amorphous solubility, the excess precipitates as amorphous nanodroplets. C7
• Mechanism II (Harmon et al. 2016, with anacetrapib–PVPVA): water entering the ASD causes amorphous–amorphous phase separation (AAPS) inside the hydrated matrix; nano-sized drug-rich domains form in a continuous polymer phase; the polymer dissolves faster than the drug, the domains are released as nanodroplets, and some of them dissolve until the solution reaches the amorphous solubility. At higher drug loading the domains would agglomerate ("hydrophobic capture") into a scaffold that is not released. C7·C8

Both routes end in the same state: nanodroplets in equilibrium with a solution at the amorphous solubility. Looking at the end state therefore cannot tell them apart. Harmon et al. had suggested that mechanism II "may be" a general mechanism, particularly for ASDs made with PVPVA, because poorly water-soluble drugs have little affinity for this polymer. C8 The authors state that, from observations in their own laboratory, they "believe" mechanism I occurs and is "perhaps" the more common one. C9 Their goal was to build a test that distinguishes the two routes directly. They borrowed an idea from reaction chemistry, where isotope labels are tracked from reactant to product to reveal a pathway (Maitra 1997; Mosi 1995). C10 The model drug was nifedipine, a neutral, poorly soluble compound with a log P of 3.17, formulated with HPMC-E5 and with PVPVA, the polymer for which mechanism II had been proposed as general. M1·C9

Background concepts

Amorphous solubility and LLPS. The amorphous solubility is the highest concentration of dissolved drug that can coexist with a drug-rich amorphous phase; above it, the excess drug separates as a second, drug-rich liquid (or glassy) phase. In practice it is measured as the onset of light scattering when a concentrated drug solution in an organic solvent is added slowly to water (the solvent-shift or "UV extinction" method). The paper relates the gain in solubility of the amorphous form to the free energy difference between the forms (Discussion, Eq. 3, after Hancock 2000): E3·C28

In plain terms: the larger the free energy gap between amorphous and crystalline drug, the larger the ratio of their solubilities. Here $\Delta G$ is the free energy difference between the two forms, $R$ the gas constant, $T$ the temperature, and $C_{amorph}$ and $C_{xtal}$ the amorphous and crystalline solubility; the paper does not state the direction in which $\Delta G$ is taken, and it gives no crystalline solubility for nifedipine. E3

Congruent release and the carrier effect. Release is congruent when drug and polymer leave the ASD in the same proportion as in the solid. Older dissolution studies (Simonelli et al. on sulfathiazole–PVP) found three regimes: crystalline drug-controlled release at very high drug loading, amorphous drug-controlled release at intermediate loading, and polymer-controlled release, the fastest, at very low loading, where the polymer was "postulated" to carry the drug out. R14·C29 In the polymer-controlled regime the drug's own solubility should not matter to the release rate. Mechanism I is, in the authors' words, the carrier-controlled dissolution of this historical view followed by nanodroplet formation. C31

Drug-controlled dissolution: the Noyes–Whitney equation. If, instead, a drug-rich layer forms on the dissolving particle, for example through AAPS, the drug controls its own release, and its rate follows a Noyes–Whitney law (Eq. 4): C33·E4

In plain terms: the drug dissolves at a rate proportional to the exposed area $A$, its diffusion coefficient $D$ divided by the boundary-layer thickness $h$, and the gap between the solubility of the dissolving phase $C_s$ (here the amorphous solubility) and the drug concentration in the bulk $C_b$. E4 In a medium already saturated at the amorphous solubility (Eq. 5), E5

In plain terms: the gap is zero, so drug-controlled dissolution of amorphous drug should stop. The paper uses this as a test: an ASD that still dissolves in such a medium cannot be dissolving under drug control. E5·C35

Amorphous–amorphous phase separation. AAPS is the splitting of a single amorphous mixture into a drug-rich and a polymer-rich amorphous phase. Water absorbed by an ASD can trigger it, because water favours the polymer; whether it occurs depends on the balance of drug, polymer and water interactions, and so on the drug, the polymer and the composition (Rumondor 2010, 2011; Marsac 2010). C38

Isotope labelling read by NMR. Deuterated nifedipine (D-NFD) carries deuterium in place of the four hydrogen atoms on its nitrophenyl ring; everything else, including the two methyl groups on the dihydropyridine ring, is unchanged (Figure 2). In a proton NMR spectrum both forms show the methyl peak at 2.29 ppm (6 protons), but only ordinary, proteated nifedipine (H-NFD) shows the aromatic peaks between 7 and 8 ppm (4 protons) (Figure 5). M2·P9 Two practical points make the test possible. First, drug inside nanodroplets gives no signal in the solution NMR spectrum, so the measured ratio is that of the dissolved drug only. Second, the paper assumes that precipitation "does not differentiate" H-NFD from D-NFD, so the two forms share one amorphous solubility and separate in proportion. C11·C12 "Isotope scrambling" in the paper means that the two forms end up evenly distributed between solution and droplets.

Glass transition and exchange. Molecules in a drug-rich droplet can exchange with molecules in the surrounding solution. Above its glass transition temperature ($T_g$) the droplet is a mobile liquid and exchange is fast; below it the droplet is a glass and exchange slows. Water lowers $T_g$ (plasticisation), so the relevant value is the $T_g$ of water-saturated drug, the "wet" $T_g$. C18·M1

Methods in detail

The study has one decisive experiment, the isotope-scrambling dissolution, and three supporting ones that make it interpretable: the amorphous solubility at the test temperature, a control that shows drug does not move quickly between droplets and solution at that temperature, and dissolution profiles that show when the ASDs are fully dissolved and whether they form droplets.

Materials and ASDs. H-NFD came from Sigma Aldrich and D-NFD from C/D/N Isotopes; HPMC-E5 Premium LV from Dow and PVPVA (Kollidon VA 64) from BASF. M1·M2·M3·M4 ASDs of H-NFD with each polymer were made at 10% and 20% drug loading (DL) by rotary evaporation from a 20% w/w solution, in methanol for PVPVA and in dichloromethane:methanol 1:1 for HPMC-E5, with a 50 °C water bath; the powder was dried under vacuum, milled and sieved to 150–250 µm. P1 All aqueous work used D2O, so that the water itself gives no proton signal in the NMR spectra. M5

Contribution 1: the isotope-scrambling dissolution test

The problem. Both mechanisms end with nanodroplets in a solution at the amorphous solubility; the test has to tell from the end state which route the ASD drug took. C7

What they did. A D2O medium was saturated with D-NFD at its amorphous solubility at 5 °C, 110 µg/mL, by adding a concentrated D-NFD solution in deuterated methanol. An H-NFD ASD was then added and allowed to dissolve completely under non-sink conditions, and the dissolved drug was analysed by proton NMR immediately after complete dissolution, judged from UV profiles. P8 Table 1 of the paper sets four doses for each ASD. P8

Table A. Doses of the isotope test (from Table 1 of the paper; the 100% expectation for mechanism II is the dashed line in Figure 9). P8·C11

Why it works. Under mechanism I the ASD drug passes through the dissolved state; once the total drug exceeds the amorphous solubility, drug separates as droplets without regard to isotope, so the dissolved drug takes the overall isotope ratio. With 110 µg/mL D-NFD and, for example, 110 µg/mL H-NFD, the solution should hold 50% D-NFD. Under mechanism II the ASD drug leaves as droplets; because the medium is already at the amorphous solubility, those droplets have no driving force to dissolve ("step 3 of mechanism II is prohibited"), and the solution should remain 100% D-NFD with no H-NFD signal (Figure 3). C11·C12 Figure 3 adds a subtlety for mechanism I: the first H-NFD to dissolve displaces D-NFD from a saturated solution, so the first droplets are mainly D-NFD, and only at complete dissolution do both forms distribute evenly. C11

Quantification. For the HPMC-E5 ASDs the methyl integral at 2.29 ppm was set to 6 protons and the H-NFD fraction was read from the aromatic integral (Eq. 1): P9·E1

In plain terms: because the methyl signal counts all dissolved nifedipine and the aromatic signal counts only H-NFD, the aromatic integral (on a scale where the methyl signal is 6) divided by its 4 protons gives the H-NFD share, and the rest is D-NFD. E1 PVPVA peaks overlap the methyl region, so for the PVPVA ASDs the H-NFD concentration was measured against maleic acid as an external standard (two –CH protons; Larive 1997), with the spectra acquired under the same conditions (Eq. 2): P9·E2

In plain terms: the H-NFD concentration is the standard's concentration scaled by the ratio of the per-proton signals of drug and standard; NF, the "normalisation factor", is the reciprocal of the number of protons behind each integrated peak. The D-NFD concentration was then taken as 110 µg/mL minus the H-NFD concentration. E2·P9 As printed, Eq. 2 contains the "no. of protons" where the text defines the peak integrals (integral_H-NFD, integral_{maleic acid}), which do not appear in the expression, and the conversion from molar to mass units is not stated; the integrals were most likely intended. X12 Neither the integrals nor the maleic acid concentration are reported, so no value can be recomputed. E1·E2

Contribution 2: a control for exchange between droplets and solution

The problem. Scrambling could also appear under mechanism II if H-NFD in released droplets exchanged quickly with D-NFD in solution. The test is only decisive if droplet–solution exchange is slower than ASD dissolution plus NMR analysis, and temperature controls that rate. C13

What they did. Two solutions were made by pipetting CD3OD stock solutions (10 mg/mL) into D2O containing 1 mg/mL HPMC-E5: system A held D-NFD at 1× the amorphous solubility (one phase), system B held H-NFD at 3× (two phases: H-NFD droplets in a solution at the amorphous solubility). P5 On mixing, the concentration of each species is halved, and the system holds 2× the amorphous solubility in total. Right after mixing (t₀) the solution contains 0.5× D-NFD and 0.5× H-NFD, so 50% D-NFD, and the droplets contain only H-NFD; at full exchange, both phases hold 0.25× D-NFD and 0.75× H-NFD, so 25% D-NFD (Figure 4). P5 The mixture was transferred to an NMR tube and measured at 5, 10, 15, 20 and 25 °C, and over about an hour at 5 °C, with a 30 s delay, 0.819 s acquisition, 90° pulses and 16 scans per time point. P5·A4·A6

The glass transition behind the choice of temperature. The dry $T_g$ of quench-cooled amorphous nifedipine was 42 °C by DSC, in line with earlier reports. Its water uptake, measured at 25 °C up to 95% relative humidity and extrapolated to 100%, was estimated at 2.5%. Using the observation (for indomethacin, Andronis 1997) that $T_g$ falls by about 10 °C for each 1% of water, the authors estimated the wet $T_g$ at about 17 °C (42 − 10 × 2.5 = 17 °C). M1·P3·P4·E6·C16 This is a stated relation, not an equation in the paper, and the 17 °C value is an estimate rather than a measurement. E6·X8

Contribution 3: dissolution into a medium already saturated with drug

The problem. The isotope test needs to know when each ASD has fully dissolved and whether it forms droplets. The same experiment also tests whether the drug or the polymer controls dissolution, through Eqs. 4 and 5. P7·C35

What they did. At 5 °C, 20 mL of D2O was stirred at 300 rpm and the absorbance at 360 nm was followed for 60 min with a fibre-optic UV dip probe. In set one the medium was plain D2O and enough ASD was added to release 220 µg/mL of drug, twice the amorphous solubility. In set two the medium was first brought to 110 µg/mL with a methanolic H-NFD aliquot, and enough ASD was added to release a further 110 µg/mL, again 220 µg/mL in total. P7 Note that set one therefore received twice as much ASD as set two. X10

Supporting measurements

Amorphous solubility. A 10 mg/mL methanol solution of H-NFD was pumped at 30 µL/min into 15 mL of D2O containing 1 mg/mL HPMC-E5 at 5, 10, 15, 20 and 25 °C, and scattering at 450 nm, a wavelength the drug does not absorb, was followed; the concentration at which scattering began to rise linearly was taken as the amorphous solubility (n = 3). The HPMC-E5 was added to prevent crystallisation, and its concentration was chosen as the maximum polymer concentration "that could be reached upon ASD dissolution". P2·C14

Nanodroplet size. ASDs were dissolved at 5 °C in 10 mL of D2O to twice the amorphous solubility, and nanoparticle tracking analysis (NTA; Nanosight LM10, 532 nm laser, ×20) followed the Brownian motion of the scattering centres, converted to size by the Stokes–Einstein equation in the instrument software (n = 3). P6·E7

Experimental analysis

The amorphous solubility rises with temperature

The amorphous solubility of nifedipine in D2O with 1 mg/mL HPMC-E5 rose from 110 µg/mL at 5 °C to 156 µg/mL at 25 °C (Table 2), and the onset concentration in D2O was close to that in H2O, whose values are not given. R1·R2·C15 The 5 °C value, 110 µg/mL, sets every concentration of the isotope test. P8

Table B. Amorphous solubility of nifedipine in D2O containing 1 mg/mL HPMC-E5, mean (SD), n = 3 (from Table 2). R1

Two qualifications follow from how this number was obtained. It was measured for H-NFD only, while the isotope medium was made by dissolving D-NFD to the same 110 µg/mL; the identical solubility of the two forms is assumed. X7·C12 And it was measured with 1 mg/mL HPMC-E5, described as the highest polymer level reached on ASD dissolution, whereas the NMR samples contained, by simple arithmetic from Table 1, between about 0.14 and 6.93 mg/mL of polymer, including PVPVA, for which no amorphous solubility was measured. X6·C14

Below about 15 °C, drug barely moves between droplets and solution

In the exchange control, the share of D-NFD in solution fell from the 50% expected right after mixing toward the 25% of full exchange only at the higher temperatures (Figure 6A). R3

Table C. Exchange control: %D-NFD in solution after mixing D-NFD solution (1× amorphous solubility) with H-NFD droplets (3×); 50% means no exchange, 25% full exchange (read from Figure 6A, approximate; single bars without error bars). R3·P5

The authors call the temperature effect "evident": at 25 °C the ratio approaches equilibrium within the roughly 15 min the experiment takes, including mixing, transfer and data collection. C17 They attribute the fast exchange to the droplets being above their estimated wet $T_g$ of about 17 °C, where molecular mobility "presumably" speeds equilibration, while at 15 °C and below the droplets are glassy, which "appears to slow down" equilibration. C18·M1 The step between 15 and 20 °C fits that picture, but no temperature between them was tested and the 17 °C value is itself an estimate. X8

Held at 5 °C (Figure 6B), the solution stayed between about 39.5 and 47% D-NFD for the whole hour: 29 points, recorded about every 2 min from about 5 to 61 min, start at about 41.2% and end at about 44.6%, with no drift toward 25%. R4 The authors conclude that complete exchange "does not occur for at least 60 min" at 5 °C, and chose this temperature for the isotope test. C20·C19

The first 5 °C value already lies about 9 points below 50%. The authors explain this "partially" by H-NFD on the droplet surfaces, which can equilibrate fast: from the droplet surface area, the area of one molecule and the number of droplets (none of these inputs are given), about 2.5% of the H-NFD sits on the surface. C21·R5 Note by the AI writer (not discussed in the paper): even if all of that surface H-NFD swapped with D-NFD in solution, the solution would fall only to about 46–47.5% D-NFD, so most of the early 9-point drop remains unexplained. X3 The times are also uncertain: the stated acquisition (16 scans with a 30 s delay) takes about 8 min per spectrum, yet Figure 6B shows a point every 2 min, and the caption says the experiment began about 1 min after mixing. X4

All four ASDs form nanodroplets of about 200 nm

NTA detected scattering centres for every ASD dissolved at 5 °C (Figure 7A), with a narrow size distribution and a mean size of about 200 nm (Figure 7B). R6·R7·C22

Table D. Nanodroplet size by NTA, n = 3 (read from Figure 7B, approximate). R7

The UV experiments point the same way: the absorbance passed the value for the amorphous solubility, the extinction at a non-absorbing wavelength rose once the amorphous solubility was exceeded, and the solutions turned turbid without visible particles. "Hence", the authors write, all formulations underwent LLPS, in plain and in drug-saturated medium. R10·C23

The ASDs keep dissolving in water already saturated with drug

PVPVA ASDs dissolved in less than 5 min and HPMC-E5 ASDs more slowly, with complete release at about 20 min (Figure 8). C23·R8·R9 The key point is that release did not stop in the medium pre-saturated at 110 µg/mL: the apparent concentration carried on to about 214–218 µg/mL by 60 min, close to the 220 µg/mL of full release. R8·R9

Table E. Apparent drug concentration (µg/mL) at 5 °C; "saturated" = D2O pre-saturated with 110 µg/mL H-NFD (read from Figure 8, approximate; 20 min not transcribed for PVPVA). The last column divides the drug released from the ASD by the ASD dose (220 µg/mL in plain water, 110 µg/mL in saturated medium); this is the AI writer's arithmetic, not a value in the paper. R8·R9·P7·X10

The authors read Figure 8 as showing that pre-dissolved drug "does not impact dissolution" ("This suggests") and that the release rates are "similar" in both media. C24·C35 Combined with Eqs. 4 and 5, which say that drug-controlled dissolution of amorphous drug would stop in a saturated medium, they conclude that the profiles "clearly indicate" polymer-controlled dissolution; in the conclusions, polymer-controlled dissolution "is confirmed". C35·E4·E5 Note by the AI writer: because the plain-water runs received twice the ASD of the saturated-medium runs, equal rises in concentration mean different fractions released. On the fraction basis of Table E, release in the saturated medium was faster for the 10% DL ASDs and similar (HPMC-E5) or slower in the first minute (PVPVA, about 51% versus 72% at 1 min) for the 20% DL ASDs, but it continued in every case, which is what the argument needs. X10 Whether release was truly complete when NMR was run is less clear for the HPMC-E5 ASDs, whose curves still rose after 20 min. X11

The decisive result: the label is diluted as mechanism I predicts

After the H-NFD ASDs dissolved in D2O saturated with D-NFD, the share of D-NFD in solution fell with the H-NFD dose and sat close to the mechanism I values at every dose, far below the 100% of mechanism II (Figure 9). R11·P8

Table F. %D-NFD in solution after complete ASD dissolution at 5 °C; mechanism II predicts 100% at every dose (measured values and red lines read from Figure 9, approximate; single values without error bars; Table 1 values from the paper). R11·P8·R12

At the highest dose the PVPVA ASDs gave no data, because the PVPVA peaks at that polymer level hid the drug's aromatic signal. R12 The authors find the data in "excellent agreement" with the predictions for mechanism I, and conclude that the results "strongly suggest" that the nanodroplets form first by dissolution of the drug in the bulk medium or stagnant diffusion layer, followed by LLPS above the amorphous solubility, for HPMC-E5 and PVPVA alike. C25·C26 Because exchange is slow at 5 °C, they read the H-NFD found in solution as dissolved ASD drug that replaced D-NFD, not as drug leaking out of droplets. C27

Two features of Figure 9 deserve care. First, the measurements are single bars without error bars or a stated number of replicates, and they scatter by up to about 6 points on either side of the Table 1 values, for example about 69.4% against 75% for the 20% DL HPMC-E5 ASD at 36 µg/mL. X2 Second, the red expectation lines in the figure are drawn at about 75.0, 57.9, 29.1 and 14.4%, whereas Table 1 gives 75, 50, 25 and 12.5% for the same doses. The Table 1 values equal 100 × 110/(110 + dose), as full mixing of 110 µg/mL D-NFD with the H-NFD dose requires; the paper gives no reason for the line positions, and the figure caption calls them "red solid bars" while the text calls them "horizontal red lines". X1 Neither issue changes the main reading: every bar lies 20 or more points below the 100% that mechanism II predicts. X2·R11

Discussion

How the authors put the findings together

The Discussion frames the result historically. Since the 1960s, faster dissolution from solid dispersions has been attributed to the polymer acting as a carrier, and Simonelli's sulfathiazole–PVP study showed the fastest, polymer-controlled release at the lowest drug loadings. R14·C29·C31 A β-carotene–PVP dispersion was reported in 1965 to give a colloid of about 100 nm on dispersal in water, and the authors judge that these particles "most likely" have the same origin as the nanodroplets described here. R13·C30 Their own picture follows two branches, depending on what happens at the surface of the dissolving particle: C31·C33

For the nifedipine ASDs, the isotope test and the saturated-medium dissolution both place the system on the left branch: the results provide "strong evidence for a polymer mediated dissolution process leading to LLPS and subsequent nanodroplet formation", and complete isotope scrambling shows, in the authors' words, that the H-NFD ASD dissolves completely to molecularly dissolved drug before droplets form. C34·C26 The authors also address an alternative route to scrambling, namely D-NFD diffusing from the solution into the solid ASD and mixing with H-NFD there. They call it "theoretically plausible" but "unlikely in practice": since the 5 °C exchange data "suggest" that D-NFD does not mix with glassy H-NFD droplets, D-NFD should not diffuse into the more viscous ASD either, and the scrambling "can be attributed solely" to mechanism I. C36 This is an inference from the droplet control; uptake of D-NFD into the hydrating ASD was not measured. X9

Mechanism II is not ruled out in general

The authors see "no evidence of mechanism II" for the nifedipine ASDs, with either polymer, and so "cannot concur" that mechanism II, as observed for anacetrapib–PVPVA, is likely a general mechanism for all PVPVA dispersions. C37 They do not deny that it can happen. AAPS depends on drug–polymer miscibility in the presence of water, which varies with the drug, the polymer and the composition, and drug–PVPVA ASDs that dissolve rapidly while staying miscible in water "are expected" to form nanospecies by mechanism I. C38 The occurrence of LLPS on dissolution also depends on polymer type and polymer:drug ratio, and "most likely" on the strength of drug–polymer interactions, which affects how fast the polymer is released. C32 The miscibility of the hydrated nifedipine ASDs, on which this distinction rests, was not measured in the paper. C38

What it suggests for formulation

The authors' practical message is general: understanding the origin of nanodroplets is needed to design ASDs with the required performance, and the insights are said to enable "development of optimally performing ASD formulations"; since nanodroplets act as a reservoir that sustains maximum flux, it is "desirable" that an ASD forms them. C40·C5·C4

Note by the AI writer (not discussed in the paper): for these nifedipine ASDs, nanodroplets are a consequence of fast, congruent, polymer-driven release followed by supersaturation above the amorphous solubility. The polymer changed the release time (under 5 min for PVPVA against about 20 min for HPMC-E5) and slightly the droplet size (about 190–198 nm against 217–222 nm), but not the route. R7·R8·R9·R11 A formulator who wants nanodroplets should therefore protect the conditions of the left branch above: a polymer that dissolves fast and keeps carrying the drug, and a drug loading low enough that no drug-rich layer forms. The isotope test itself is a reusable diagnostic for the route, but it relies on droplets that stay glassy and exchange slowly during the experiment, which for nifedipine meant running it at 5 °C rather than at body temperature.

Limitations

Stated or implied by the authors. The authors chose 5 °C precisely because exchange between droplets and solution is fast above the droplets' estimated $T_g$, so the test as designed works only below it. C13·C19 They partly explain the early offset of the 5 °C control by surface exchange ("partially explained"). C21 At the highest dose the PVPVA signal hid the drug peaks, so the PVPVA ASDs have data at three doses only. R12 They limit their conclusion to "the nifedipine systems" and expect that ASDs which remain miscible in water follow mechanism I, leaving room for mechanism II elsewhere. C26·C38

The AI writer's own judgement.

• The decisive data are single values without error bars or stated replicates, scattered by up to about 6 points around the expectation. X2
• The expectation lines drawn in Figure 9 differ from Table 1 at 110, 330 and 770 µg/mL without explanation; Table 1 is the one that follows from simple mixing arithmetic. X1
• The slow-exchange premise was shown with droplets made from a solvent stock in a medium containing 1 mg/mL HPMC-E5, not with droplets released from the ASDs and not with PVPVA present; and about 9 points of exchange in the first minutes at 5 °C are not accounted for. X5·X3
• The amorphous solubility was measured for H-NFD with 1 mg/mL HPMC-E5 only. D-NFD is assumed to behave identically, and the test samples contained up to about 7 mg/mL polymer, including PVPVA. If the actual polymer level changed the saturation point, the 100% expectation for mechanism II would also be affected. X6·X7
• The droplet $T_g$ of about 17 °C is estimated from an extrapolated sorption value and a rule of thumb taken from indomethacin. X8
• The time scale of the exchange data is uncertain (point spacing shorter than the stated acquisition time), and the exact time of the NMR runs after ASD addition is not given. X4·X11
• The comparison of release in plain and saturated media used different ASD amounts. X10
• Only one drug, two loadings (10 and 20%), two polymers and one temperature were tested, so the result says nothing yet about higher loadings, where release slows, or about 37 °C. P1·P8

Two small presentation issues do not affect the conclusions: Figure 2 labels H-NFD "Protonated Nifedipine" although it is the neutral, non-deuterated drug, and the Figure 9 axis reads 37 where Table 1 gives 36 µg/mL. X13·X1

Conclusion

The paper turns a debate about where ASD nanodroplets come from into a direct measurement. By dissolving ordinary-nifedipine ASDs into water already saturated with deuterated nifedipine, at a temperature where droplets and solution hardly exchange, it shows that the ASD drug mixes with the labelled drug in solution about as much as full mixing predicts at all four doses, and never leaves the solution near the 100% labelled drug that release of intact domains would require. R11·P8·R4 Together with continued release into a medium already at the amorphous solubility, this supports, in the authors' words, mechanism I "for the nifedipine systems": the polymer carries the drug into solution, and the nanodroplets form by liquid–liquid phase separation once the amorphous solubility is exceeded. C26·C35 Mechanism II is not excluded for other drugs, particularly those whose ASDs phase separate as they take up water. C37·C38