How the dissolution medium changes supersaturation and crystallization of ezetimibe amorphous solid dispersions

Background and motivation

An amorphous solid dispersion (ASD) dissolves faster than the crystalline drug and can produce a supersaturated solution, one whose concentration exceeds the crystalline solubility (Introduction). Supersaturation is metastable: crystals can nucleate and grow, and the concentration falls back towards the crystalline solubility, a process called desupersaturation. How much supersaturation a formulation can reach depends on the drug's intrinsic solubility, the drug loading and the volume available. C1 For immediate-release products it is well known that the dissolution medium matters: drug solubility depends on the medium, simple aqueous buffers often predict in vivo performance poorly for poorly water-soluble drugs, and biorelevant media that mimic the composition of the gastrointestinal lumen predict better, including fed versus fasted effects for lipophilic drugs (Dressman and Reppas 2000; Klein 2010). C2

The gastrointestinal contents differ from a buffer in pH, buffer capacity, osmolality, surface tension and, above all, in solubilizing species such as bile salts and phospholipids. These can change both how much supersaturation forms and how long it lasts. Bile salts alone change the crystallization kinetics of supersaturated drugs (Chen et al. 2015; Li et al. 2016; Lu et al. 2017), and the authors raise the possibility ("It is also possible") that bile salts and the polymers in a formulation act together. C3·C4 For supersaturating formulations, however, little in vitro work had looked at supersaturation and the detailed phase behaviour in biorelevant media; the authors expect that such work "would presumably" give a better picture of the intraluminal supersaturated state. C5

A practical obstacle is turbidity. Biorelevant media can be cloudy from the start, and an ASD that exceeds its amorphous solubility makes a solution turbid with drug-rich droplets. UV concentration readings are distorted by scattering particles, and light scattering cannot tell a non-crystalline drug-rich phase from crystals (Van Eerdenbrugh et al. 2011). The authors therefore argue that "a full mechanistic understanding of the phase behavior of supersaturated solutions cannot rely solely on concentration-time profiles", and that a crystal-specific method is needed. C30

The study uses ezetimibe, a BCS class II cholesterol-lowering drug. Its phenol has a pKa of 9.72, so it is essentially un-ionized across the physiological pH range, which removes pH-dependent ionization as a variable. M1·C6 It has three chiral centres, two on the β-lactam ring and one in the side chain, and its commercial monohydrate crystallizes in the non-centrosymmetric space group P2₁2₁2₁, which makes its crystals SHG-active. M1·C8 The β-lactam ring is unstable at neutral to alkaline pH: earlier studies reported a rearrangement to an amide and a tetrahydropyran ring in tertiary amine buffers at pH 7-12 and 39 °C, and in pH 6.8 phosphate buffer at room temperature. R26

The paper set out to characterize the phase transformations that occur when 50:50 ezetimibe ASDs made with PAA, PVP or HPMC-AS dissolve in three media: 10 mM sodium phosphate buffer at pH 6.8, FeSSIF (a model of the intestinal fed state) and Ensure Plus (a model of the gastric fed state), and to test whether SHG microscopy, in the form of second order nonlinear optical imaging of chiral crystals (SONICC), can follow crystallization in these turbid media. A1·A2·C7

Background concepts

Two limits: crystalline solubility and the LLPS concentration

Two concentrations frame everything in this paper. The crystalline (equilibrium) solubility is the concentration in equilibrium with the stable crystal. The LLPS concentration is the concentration at which a supersaturated solution separates into a drug-lean solution and a drug-rich, non-crystalline colloidal phase (liquid-liquid phase separation, LLPS); it is the experimental estimate of the amorphous solubility, the highest concentration a single phase can hold. Between the two limits the solution is supersaturated but one phase. Above the LLPS concentration a drug-rich phase coexists with a solution held at maximum supersaturation, and crystals can nucleate from either phase (Ilevbare and Taylor 2013; Jackson et al. 2014; Raina et al. 2014). C38·C12

Solubilizing media raise both limits. The authors cite earlier work showing that sodium taurocholate, the only bile salt in FeSSIF, raises the concentration at which drugs phase separate, and add that in such media "the thermodynamic activity-boundaries remain the same, although the amorphous solubility increases" (Lu et al. 2017). C39·C40 In other words, a higher measured concentration in a solubilizing medium does not necessarily mean a higher thermodynamic driving force.

Induction time and polymer inhibitors

The nucleation induction time is the time a supersaturated solution stays free of detectable crystals. Here it was measured at each medium's LLPS concentration, so the solution starts at maximum supersaturation with some drug-rich phase present; the onset was read as a drop of the 233 nm absorbance, where ezetimibe absorbs, together with a change in scattering at 350 nm, where it does not. P5·C12·C13 A polymer that lengthens the induction time when predissolved is a crystallization (nucleation) inhibitor in that medium.

The three media

Table A. The three dissolution media as described in the paper (Materials, Discussion). M7·M8·M9·M10·R20

Ensure Plus splits on centrifugation into an upper oily phase, an intermediate aqueous phase and a lower pellet, and the authors measured ezetimibe in both liquid phases separately; its solubilizing power is attributed to its lipids, to possible drug binding to proteins and to casein micelles. P2·C48

Second harmonic generation microscopy

SHG is a nonlinear optical effect in which two photons of the laser (here 1064 nm) combine into one photon at twice the frequency. It can arise only in structures without a centre of symmetry. In a solution or an amorphous solid the molecules point in all directions and the signals cancel, so no coherent SHG is produced; a chiral crystal such as ezetimibe monohydrate gives a strong signal. The images therefore show crystals as bright spots on a dark background, regardless of how turbid the medium is, provided the medium itself gives no background. C8·P10·R9 From time series of such images the authors derived a nucleation rate (new particles per unit time per unit volume, by threshold particle counting), a growth rate (change in mean crystal size, estimated by Fourier analysis because the crystals overlapped too much to size one by one) and a final crystallinity (crystalline material per volume of medium in the field of view, not the crystallinity of the ASD). P11·C31

Methods in detail

The study has four layers: reference limits for each medium, solution tests of each polymer as a crystallization inhibitor, dissolution of the ASDs with crystal detection, and a quantitative analysis of the SHG images.

Reference limits: equilibrium solubility and LLPS concentration

Excess crystalline ezetimibe monohydrate was stirred at 300 rpm and 37 °C for 48 h in each medium, without polymer or with 10 µg/mL PAA, PVP or HPMC-AS (n = 6). P1 In buffer the measurement was made at pH 2.0 rather than 6.8 to avoid hydrolysis of the β-lactam ring; ezetimibe does not ionize at pH 2, and a single HPLC peak showed that it did not degrade. C9·C10·R2 Buffer and FeSSIF samples were ultracentrifuged at 35,000 rpm and 37 °C for 30 min. Ensure Plus samples followed the method of Diakidou et al.: centrifugation at 21,000 g and 8 °C for 20 min, separate collection of the oily and aqueous phases, dilution in acetonitrile and a second centrifugation to precipitate the medium's own components. P1·P2 Ezetimibe was assayed by reversed-phase HPLC with UV detection at 233 nm (C18 column, water:acetonitrile 40:60, standards 0.05-10 µg/mL, r² = 0.9998). P3

The LLPS concentration was found by adding ezetimibe with a syringe pump at 2 µg/min to buffer (pH 6.8) or FeSSIF at 37 °C and following the extinction at 350 nm with a fibre-optic UV probe; a small amount of PVP or HPMC-AS (2 µg/mL) was present to delay crystallization, and the sudden rise in extinction marks the second phase (n = 3). P4·C11 The methods section gives the polymer level as 1-2 µg/mL and does not say in which solvent the ezetimibe stock was added. X15

Polymers as inhibitors in solution

Supersaturated ezetimibe was held at 21 µg/mL in buffer and at 123 µg/mL in FeSSIF, each medium's LLPS concentration, with no polymer or with 2, 10 or 100 µg/mL (buffer) or 2, 10 or 50 µg/mL (FeSSIF) of predissolved PAA, PVP or HPMC-AS, at 37 °C and 300 rpm, with UV spectra every 30 s (n = 3). P5 In buffer the onset of crystallization was read from the fall of the 233 nm absorbance plus the 350 nm scattering; in FeSSIF only the 350 nm scattering was followed, which makes induction times comparable within a medium rather than across media. P5·X17

ASD preparation and dissolution

Each ASD was made from 500 mg ezetimibe and 500 mg polymer dissolved in methanol, dried by rotary evaporation at 45 °C and overnight in a vacuum oven, and cryo-milled; PLM and SHG showed no crystallinity before use. P6·R8 For dissolution, 15 mg ASD (7.5 mg ezetimibe) was added to 15 mL medium at 37 °C and 300 rpm. P7 This dose is far above every solubility in Table 1, so the tests ran under non-sink conditions with respect to both the crystalline and the amorphous solubility. P7·R1 Buffer and FeSSIF samples were passed through 0.45 µm glass fibre filters and assayed by HPLC; Ensure Plus samples could not be filtered and were centrifuged as above, the sampling time being taken as the start of the first centrifugation. P7·P2 Runs lasted 6 h in buffer and Ensure Plus and 8 h in FeSSIF (n = 3). P7

Note by the AI writer (not discussed in the paper): each vessel also received 7.5 mg of polymer, up to 500 µg/mL if it all dissolved, while the solubility, LLPS and induction-time references were measured with 1-100 µg/mL polymer; the reference lines drawn on the dissolution plots are therefore approximate for the ASD experiments. X6

Crystal detection

Three methods were used. Polarized light microscopy (PLM) showed birefringent crystals in buffer and FeSSIF samples. P8 PXRD of pellets centrifuged from FeSSIF samples at 1, 2 and 3 h compared the solid with the monohydrate reference. P9 SHG microscopy imaged aliquots held in a 0.381 mm thick liquid cell, with 1925 µm × 1925 µm fields and Z steps of 100 µm, using a 1064 nm femtosecond laser at 350 mW. P10 None of the three media gave an SHG signal on its own, while PLM and PXRD could not separate crystals from the components of Ensure Plus. R9·R20

SHG image kinetics

Nucleation rates were obtained by threshold-based particle counting over several fields of view, growth rates from the change in mean crystal size estimated from the first minimum of the radial power spectrum, and final crystallinity by integrating the SHG intensity at the last time point. Both rates are averages over the whole experiment. P11·C31 Table 3 prints the nucleation rate in "number of particles/minute/µm³". Note by the AI writer (not discussed in the paper): one field of view is about 1925 × 1925 × 381 µm, roughly 1.4 × 10⁹ µm³, so 15 particles per minute per µm³ would mean about 2 × 10¹⁰ new particles per minute in a single field; the volume unit or scaling is most likely misprinted, and the rates are best compared with each other within this paper. X7

Experimental analysis

The media move both limits upwards

Table B. Crystalline equilibrium solubility (µg/mL, 37 °C, 48 h, n = 6; from Table 1) and LLPS concentration (µg/mL, 37 °C, n = 3; from Table 2). R1·R3

FeSSIF raised the crystalline solubility nearly twenty-fold over buffer, and Ensure Plus raised it further, to 48 µg/mL in its aqueous and 162 µg/mL in its oily phase (Table 1). R1·C15 The authors summarize the polymer effect as "generally" minor; that holds within about 10% for most entries, but 10 µg/mL HPMC-AS raised the FeSSIF solubility from 18.6 to 26.3 µg/mL, about 41% higher, so the FeSSIF solubility line drawn for all ASDs (the no-polymer value) understates the reference for the HPMC-AS dispersion. C15·X4 The buffer value comes from pH 2 and is used as the pH 6.8 reference on the strength of the high pKa. X5

In buffer, the extinction at 350 nm stayed flat at about −0.0015 a.u. up to about 20 µg/mL and then rose almost linearly, giving an LLPS onset of 21 µg/mL (Figure 2). R4 In FeSSIF the onset moved to 123 µg/mL. R3 The authors read this increase as "consistent with" earlier findings that taurocholate raises the phase-separation concentration. C39 The supersaturation window, from crystalline solubility to LLPS concentration, is therefore about 0.97-21 µg/mL in buffer and 18.6-123 µg/mL in FeSSIF. R1·R3

Polymers in solution: the ranking flips between media

Table C. Nucleation induction times (min; buffer at 21 µg/mL, FeSSIF at 123 µg/mL; n = 3), read from Figure 4, approximate. R6·R7

Without polymer, ezetimibe at 21 µg/mL in buffer crystallized within about 1 min, and 10 or 100 µg/mL PAA did not delay it at all (Figure 3, Figure 4A). R5·R6 PVP and HPMC-AS both extended supersaturation, PVP more strongly: at 10 µg/mL the concentration held at about 21 µg/mL for about 25 min with PVP and about 8 min with HPMC-AS before falling, and at 100 µg/mL the induction times were about 71 and 50 min. R5·R6 The authors conclude that PVP was "appearing to be the most effective" inhibitor at 10 µg/mL and "generally more effective than HPMC-AS" in buffer. C16

In FeSSIF at 123 µg/mL the picture reverses (Figure 4B). Without polymer crystallization began after about 9 min; PAA at 50 µg/mL (about 7 min) and PVP at 2-50 µg/mL (about 8-10 min) did not delay it, whereas HPMC-AS gave about 17 and 73 min at 2 and 10 µg/mL and prevented crystallization for more than 6 h at 50 µg/mL. R7·C17 Figure 4 shows no bars for PAA at 2 µg/mL in buffer or at 2 and 10 µg/mL in FeSSIF; the text says PAA was ineffective "at all concentrations" without stating whether those conditions were run. X14

PAA dispersions crystallize on contact

In buffer the PAA ASD reached about 2.8 µg/mL at the 2-min sample and then stayed between about 0.7 and 1.9 µg/mL up to 60 min (Figure 5, inset). R11 Birefringent crystalline aggregates were seen by PLM at 10 min (Figure 6A), and SHG showed crystals from the first image at 6 min, growing in size and number to 120 min (Figure 7A). R12·R13 In FeSSIF the concentration sat at about 30-34 µg/mL from 2 to 60 min (Figure 8, inset), PLM at 30 min showed crystal aggregates (Figure 9A), PXRD of the pellet matched ezetimibe monohydrate at 1 and 2 h (Figure 10B), and SHG showed many large crystals already at 15 min with no further change (Figure 11A). R15·R16·R17·R18

The authors conclude that the PAA dispersions gave no supersaturation, the concentration being "more or less equivalent" to the equilibrium solubility, and that "it can be inferred that crystallization from the solid matrix was the dominant mechanism", as Alonzo et al. had seen for amorphous felodipine. C33·C34 Note by the AI writer: the plotted concentrations are about 1.6-1.8 times the FeSSIF solubility and up to about twice the buffer value, so "low supersaturation" (the wording the results section itself uses for FeSSIF) describes the data better than "none"; and the discussion cites Figure 9 for the FeSSIF profile, which is in Figure 8. X2·X3·C22

Buffer: modest supersaturation, crystals from about 30 min

In buffer neither the PVP nor the HPMC-AS dispersion reached the 21 µg/mL LLPS concentration (Figure 5). C19 The PVP ASD rose from about 6 µg/mL at 5 min to a peak of about 13 µg/mL at 25 min, then declined to about 4.9 µg/mL at 120 min and about 2.8 µg/mL at 360 min; the HPMC-AS ASD rose to about 11.5 µg/mL at 25 min, stayed near 9.5-11 µg/mL to 75 min and fell to about 5.5 µg/mL at 360 min. R10 The authors describe the two profiles as differing only slightly, with desupersaturation from about 30 min. C19

The crystal images match this timing. At 60 min PLM showed small crystalline domains on the surfaces of the dissolving PVP and HPMC-AS particles (Figure 6B, 6C, arrows). R12 SHG showed almost no crystals for PVP at 10 and 20 min, a few at 35 min and many small ones at 120 min; for HPMC-AS a few spots at 10 min and many domains from 20 min (Figure 7). R13 The authors note that the desupersaturation after about 30 min is "in good agreement" with these SHG times. C21 They also state that all three ASDs showed "a comparable degree of crystallinity after 120 minutes"; Table 3, however, gives final crystallinity of 30 ± 5% (PAA), 13 ± 2% (PVP) and 32 ± 2% (HPMC-AS) in buffer. C21·R23·X1

Why did neither ASD reach the amorphous solubility? At 25 °C, in experiments whose data are not shown, both dispersions held a plateau at the 25 °C LLPS concentration for a few hours, PVP doing better than HPMC-AS. R24 At 37 °C crystallization was much faster, and the authors consider it "likely" that solution-mediated crystallization started earlier and used up the supersaturation before the "amorphous" solubility advantage could be reached. C37·X13

FeSSIF: PVP spikes and crashes, HPMC-AS holds

Table D. Ezetimibe concentration (µg/mL) during ASD dissolution in FeSSIF at 37 °C, read from Figure 8, approximate (n = 3). Crystalline solubility 18.6 µg/mL; LLPS concentration 123 µg/mL. R14·R1·R3

In FeSSIF the PVP dispersion released very fast: about 195 µg/mL at 15 min (the text says approximately 200), well above the 123 µg/mL LLPS concentration, with a turbid solution. R14·R19·C41 By 30 min it had fallen to about 41 µg/mL and it stayed at about 32-36 µg/mL to 8 h. R14 At 30 min PLM showed many small birefringent crystals across the field (Figure 9B), SHG a dense field of fine crystals (Figure 11B), and PXRD the monohydrate pattern (Figure 10C). R16·R18·R17 Table 3 gives this condition the highest nucleation rate of all, 600 ± 200, with the lowest growth rate, 0.020 ± 0.001 µm/min, and a final crystallinity of 8 ± 4%. R23

The HPMC-AS dispersion released a little more slowly but stayed at about 140-172 µg/mL from 15 to 90 min, also above the LLPS concentration, and then fell to about 64 µg/mL at 120 min (the text says around 65 µg/mL after 2 h). R14·C22 SHG showed one crystal at 30 min and only a few up to 240 min (Figure 11C); PLM showed needle-shaped crystals at the edges of undissolved particles at 3 h (Figure 9C); PXRD showed an amorphous halo at 1 and 2 h and only a small ezetimibe peak at 3 h (Figure 10D). R18·R16·R17 Its nucleation rate was 0.04 ± 0.03 and its final crystallinity 0.04 ± 0.02%. R23 PXRD peaks at 32 and 46° 2θ in several patterns are assigned to sodium chloride precipitating from FeSSIF. R17·C25

The authors' reading has three parts. First, "These images suggest that PVP acts as a poor nucleation inhibitor for ezetimibe in FeSSIF": the high supersaturation from rapid release gives a high nucleation rate and many small crystals. C42 Second, the slower crystallization of the PVP dispersion in buffer "could be due either" to slower release in buffer or to PVP being less effective in FeSSIF; the paper does not separate the two. C43 Third, HPMC-AS "vastly outperformed" PVP and "appeared to be an effective crystallization inhibitor in the presence of the drug-rich colloidal phase of ezetimibe", with FeSSIF components not weakening it as they weakened PVP. C44·C46 Because HPMC-AS MF is reported to be only sparingly soluble at pH 5.2, the authors add that "It thus appears" more soluble in FeSSIF (pH 5) than pH alone predicts, a conjecture supported by the finding of Bevernage et al. that HPMC-AS stabilized supersaturation best for four of five drugs in media including FeSSIF. C45 After its drop, the HPMC-AS dispersion stayed at about 62-80 µg/mL, roughly three to four times the crystalline solubility, which the authors explain as "probably" crystal growth inhibition by HPMC-AS at low supersaturation, or disorder in the crystals formed. R14·C47

Ensure Plus: little crystallization, and the ranking depends on the phase

Table E. Ezetimibe concentration (µg/mL) in the aqueous and oily phases of Ensure Plus during ASD dissolution at 37 °C, read from Figure 12, approximate (n = 3). Crystalline solubility 48 µg/mL (aqueous) and 162 µg/mL (oily). R21·R1

In Ensure Plus all three dispersions supersaturated both phases (Figure 12). The PAA dispersion peaked in the aqueous phase at about 141 µg/mL at 15 min and declined slowly from 30 min towards the aqueous solubility, reaching about 54 µg/mL at 6 h, while its oily phase stayed at about 197-216 µg/mL. R21·C27 The results section calls this supersaturation "short-lived"; the discussion says that "moderate supersaturation was maintained for a longer time period than for the other media", which is true in comparison with the immediate crystallization seen in buffer and FeSSIF. C27 The PVP dispersion rose faster and higher in the aqueous phase (about 237-239 µg/mL from 180 min), while the HPMC-AS dispersion rose slowly in both phases and reached the highest oily-phase values, about 342-353 µg/mL from 180 min; the authors call the HPMC-AS behaviour a slow release into the oily phase that "seemed to reach the maximum supersaturation after 3 hours". R21·C27 They place next to it earlier findings that fat-rich media such as Ensure Plus can slow release from HPMC matrix tablets by forming a rate-controlling surface gel. C53

Crystallization was slight. SHG images showed only scattered crystals for PAA from 25 min and one large cluster at 120 min, a few crystals for PVP from 75 min, and an almost dark field for HPMC-AS up to 180 min (Figure 13); the text reports the first crystals after 10 min (PAA) and 20 min (PVP) and at 120 min (HPMC-AS), times earlier than the first images shown. R22·X10 Table 3 gives nucleation rates of 0.02-0.07 and final crystallinity of 0.08% or less for all three. R23 The authors' explanation is that the higher solubility in Ensure Plus lowers the maximum theoretical supersaturation and with it the driving force for crystallization: "This explains why the SHG images for Ensure Plus® dissolution demonstrated fewer SHG-active regions". They add that differences in absorption and laser penetration depth "might also contribute", and that digestion of the lipids "may" change the supersaturation in vivo. C49·C50·C51

Two further points concern the phase split. First, the reversal of the polymer ranking between the aqueous phase (PVP higher) and the oily phase (HPMC-AS higher) "raises questions about the role of polymers in maintaining supersaturation from a hydrophilicity/hydrophobicity perspective". C52 Note by the AI writer: in the oily phase PVP was equal or higher up to about 90 min, and HPMC-AS was clearly higher only from about 180 min. X19 Second, the authors state that the end-of-test aqueous and oily concentrations add up to roughly 500 µg/mL, the total ezetimibe added, for PVP and HPMC-AS (read from the plot: about 237 + 278 and 141 + 349 µg/mL) but not for PAA (about 54 + 207 µg/mL), consistent with crystalline precipitation for PAA. R25·R21·C54 Note by the AI writer: a sum of concentrations in two phases equals the dissolved fraction of the dose only if both are expressed per total medium volume; the phase volumes are not given, so this mass balance cannot be checked from the paper. X12

SHG kinetics across all nine conditions

Table F. Crystallization kinetics from SHG image analysis (from Table 3, n = 3). Rates are averages over each experiment; final crystallinity is crystalline material per volume of medium in the field of view, not the crystallinity of the ASD. R23·P11

Across the nine conditions the nucleation rate spans more than four orders of magnitude, from 0.02 to 600, while the growth rate stays between 0.020 and 1.0 µm/min. R23 In buffer the average nucleation rate was higher for PVP and HPMC-AS than for PAA, despite their later onset, with the crystal number still rising at 120 min. C36 The authors conclude that "it is clear that differences in the nucleation kinetics rather than growth rates dominate the overall trends in crystallinity", and more broadly that solubilizing additives "can significantly alter crystal nucleation and growth kinetics" and the inhibitory effect of the polymer. C59·C58 Note by the AI writer: across media the claim follows the numbers, but within buffer the final crystallinity (HPMC-AS ≈ PAA > PVP) does not follow the nucleation ranking (HPMC-AS > PVP > PAA). X8

Discussion

How the authors connect the findings

The authors' framework comes from earlier work on supersaturated solutions: dissolution of a supersaturating formulation can end in a crystalline or an amorphous precipitate; crystals can form in the dissolving matrix or by nucleation from solution; above the amorphous solubility, with crystallization inhibited, a drug-rich phase forms and the solution stays at maximum supersaturation, and nucleation can then start in either phase. C38 The medium sets where the two limits lie, and the polymer, if it inhibits nucleation in that medium, decides how long the system stays between them. C39·C35 Read through this framework, the dissolution results fall into a few cases.

Two further cases sit outside this split. PAA dispersions never get far from the crystalline solubility in buffer or FeSSIF; the authors infer crystallization in the solid matrix. C33·C34 In Ensure Plus all dispersions reach high absolute concentrations, but the authors argue that the higher solubility leaves a smaller supersaturation ratio and a weaker driving force, so few crystals form. C49 In buffer, PVP and HPMC-AS dispersions stay below the LLPS limit and lose supersaturation from about 30 min, which the authors attribute to early solution-mediated crystallization at 37 °C ("it is likely"). C19·C37

What the paper shows, and what it infers

The measured core is solid within this system: the solubility and LLPS limits, the induction-time rankings in buffer and FeSSIF, and the concentration-time profiles with their matching crystal images. R1·R3·R6·R7·R10·R14·R21 The contrast between the PVP and HPMC-AS dispersions in FeSSIF is large and consistent across HPLC, PLM, PXRD and SHG at the same time points. R14·R16·R17·R18 The interpretation that PVP is a poor nucleation inhibitor specifically in FeSSIF is supported by the solution test, where PVP at 2-50 µg/mL gave induction times of about 8-10 min against about 9 min without polymer. R7·C42 The broader statements are inferences. That "solubilizing additives" change nucleation and growth rests on comparing whole media that also differ in pH, salts, viscosity and lipid content; no additive was added to or removed from a medium at fixed pH. C58·X9 The idea that HPMC-AS dissolves better in FeSSIF than its pH-solubility suggests is, in the authors' own words, a conjecture. C45

Seeing crystals in turbid media

The method contribution holds up well. None of the media gave an SHG background, and SHG detected crystals in buffer, FeSSIF and Ensure Plus, including the opaque Ensure Plus in which PLM and PXRD could not be used. R9·R20·R22 In buffer and FeSSIF the first strong SHG signal coincided with the start of desupersaturation, which is the paper's practical argument for pairing concentration profiles with a crystal-specific read-out. C57·C56 The quantitative part, nucleation and growth rates from image analysis, is less mature: the rates are averages over each experiment, the FeSSIF PAA nucleation rate could not be determined, and the printed nucleation-rate unit cannot be taken at face value. P11·R23·X7

What it suggests for formulation

The authors' recommendation is general: biorelevant media are "crucial" for testing enabled formulations, and "it is important to evaluate the impact of complex biorelevant media on the dissolution performance of ASDs". C60 Note by the AI writer (not discussed in the paper): for ezetimibe a polymer screen in buffer would have favoured PVP, which in FeSSIF lost nearly all of its supersaturation within 30 min, while a short induction-time test in FeSSIF already ranked HPMC-AS first, as did the ASD dissolution in FeSSIF. R6·R7·R14 Screening crystallization inhibitors in the medium of interest therefore looks more informative than screening in buffer, at least for this drug. In lipid-rich media, reporting a single phase can reverse the apparent polymer ranking. R21

Limitations

Limitations stated by the authors. The authors note that differences in absorption coefficient and laser penetration depth between buffer or FeSSIF and Ensure Plus "might also contribute" to the low SHG counts in Ensure Plus, that the extent of supersaturation "may change" during digestion of the lipids in Ensure Plus, that information on HPMC-AS solubility in FeSSIF "is not currently available", and that information on polymer-drug interactions in the different phases of a medium such as Ensure Plus is still lacking. C50·C51·C45·C52 They also mention that the 25 °C dissolution results supporting their temperature argument are not shown. R24

The AI writer's own judgement (1): what the comparisons can and cannot separate. Buffer, FeSSIF and Ensure Plus differ in pH (6.8, 5, not printed), in salts (sodium chloride precipitated from FeSSIF), in viscosity and in lipid and protein content, so the conclusion that solubilizing additives drive the differences is not isolated from these other factors; HPMC-AS, an ionizable polymer, is itself pH-sensitive, as the authors note. X9·C45 The reference lines were measured with 1-100 µg/mL polymer, whereas each dissolution vessel held up to 500 µg/mL of polymer, and 10 µg/mL HPMC-AS alone raised the FeSSIF solubility by about 41%. X6·X4 The buffer solubility was measured at pH 2 and used at pH 6.8. X5 Induction times were read with different signals in the two media. X17 Ensure Plus samples were separated by a 20 min centrifugation at 8 °C, so their time points are approximate. X18

The AI writer's own judgement (2): statements that go beyond the data. The PAA dispersions are said to give no supersaturation, but their concentrations sat about 1.6-2 times above the crystalline solubility. X2 All three dispersions are said to reach comparable crystallinity in buffer at 120 min, but Table 3 gives 13% for PVP against 30-32% for the others. X1 Nucleation is said to dominate the crystallinity trends, which holds across media but not within buffer. X8 The Ensure Plus mass balance adds per-phase concentrations without phase volumes. X12 The discussion cites Figure 9 for the FeSSIF PAA profile (it is Figure 8), the PXRD scan range is given as 5-50° while the patterns are plotted to 60°, and the earliest Ensure Plus SHG times rest on images not shown. X3·X11·X10 None of these change the main FeSSIF result, but they matter for anyone reusing the numbers. Overall the evidence is moderate: direct and replicated (n = 3) for the profiles and induction times in one drug at one loading, and inferential for the mechanisms.

Conclusion

For 50:50 ezetimibe ASDs at 37 °C, the dissolution medium decided which polymer kept the drug supersaturated. PAA did not inhibit crystallization in any medium, and its dispersions crystallized within minutes. R6·R11·R15·C34 PVP was the better inhibitor in pH 6.8 buffer, where neither PVP nor HPMC-AS dispersions reached the LLPS concentration and both lost supersaturation from about 30 min. R6·R10 In FeSSIF the PVP dispersion spiked to about 195 µg/mL and crashed within 30 min with many small crystals, while the HPMC-AS dispersion held about 140-172 µg/mL, above the 123 µg/mL LLPS concentration, for about 90 min. R14·R3 In Ensure Plus all dispersions reached high concentrations with very little crystallization, and the polymer ranking depended on which phase was sampled. R21·R23 SHG microscopy detected crystals in every medium, including the opaque Ensure Plus, and its crystal onset matched the start of desupersaturation. C56·C57 The mechanisms the authors propose (PVP failing as a nucleation inhibitor in FeSSIF, solubilizing additives changing nucleation and growth, HPMC-AS dissolving better in FeSSIF than expected) remain inferences from whole-medium comparisons. C42·C58·C45·X9