Speed, stability and what is left behind: how EPO and HPMC shape indomethacin release from amorphous dispersions

Background and motivation

Amorphous solid dispersions (ASDs) raise the apparent solubility of poorly soluble drugs by presenting the drug in a high-energy, non-crystalline form, usually dispersed in a polymer. The supersaturated solution they generate is not stable: it moves toward equilibrium by nucleation and crystal growth, and the higher the supersaturation, the stronger this tendency (Brouwers et al., 2009). Polymers in the ASD therefore have two jobs, which may conflict: they set how fast the drug is released, and they may inhibit crystallization once it is in solution (Corrigan, 1985; Janssens and Van den Mooter, 2009). C7·C8

The authors' own earlier work on celecoxib had shown that one polymer need not do both jobs. An ASD with the ionizable polyacrylic acid (PAA, pKa 4.28) dissolved much faster at pH 6.8 than one with a cellulose polymer, but HPMC was the better crystallization inhibitor, and a ternary dispersion containing both polymers outperformed either binary one (Xie and Taylor, 2016). One small inconsistency: the introduction names the comparison polymer "hydroxypropyl cellulose (HPMC)", so the name and the abbreviation disagree. R30·C9·X2

Indomethacin (IND) is a weakly acidic non-steroidal anti-inflammatory drug (pKa 4.42) whose solubility rises from 1.5 µg/mL at pH 1.2 to 105 µg/mL at pH 7.4 (Sun and Lee, 2015). It has several polymorphs; the stable γ form and the slightly less stable α form are the most common (Andronis and Zografi, 2000). Two earlier observations framed the study. Sarode et al. (2013) found that Eudragit EPO, a cationic methacrylate copolymer that dissolves in acid, greatly speeded the dissolution of IND dispersions in 0.1 N HCl, but the resulting supersaturation burst was followed shortly by a rapid fall in concentration from crystallization, which "suggests that EPO may be an ineffective crystallization inhibitor". In the same work, HPMC (Pharmacoat 603) and PVPVA dispersions gave no meaningful improvement over crystalline drug. Alonzo et al. (2010), on the other hand, found HPMC better than PVP at preventing IND crystallization from supersaturated solution, and HPMC's effectiveness "appears to be quite general" across poorly soluble drugs, perhaps because it is amphiphilic and can interact with both polar and non-polar parts of a drug molecule (Warren et al., 2010; Ilevbare et al., 2012, 2013). M1·R28·C10·R29·C11

Put together, these suggest a hypothesis the authors state in the introduction: a ternary IND dispersion with both HPMC and EPO is "likely" to dissolve fast and to sustain the resulting supersaturation, particularly at low pH, so that the drug could be released quickly in the stomach. C12

A second line of background concerns what happens at very high supersaturation. If crystallization is inhibited, a supersaturated solution of a lipophilic drug can undergo liquid–liquid phase separation (LLPS) once the concentration exceeds the amorphous solubility. The excess drug then forms a colloidal, drug-rich droplet phase in metastable equilibrium with a water-rich phase that holds the drug at its amorphous solubility (Taylor and Zhang, 2016; Murdande et al., 2010; Raina et al., 2014). Because the activity of the dissolved drug no longer rises above this point, membrane flux does not increase further (Jackson et al., 2016; Raina et al., 2014), but the droplets "are considered" beneficial as a reservoir that replenishes the water-rich phase while drug is absorbed (Indulkar et al., 2016). C13·C14

Against this background the paper sets out to (i) measure how HPMC and EPO, alone and together, change the crystalline and amorphous solubility of IND; (ii) compare the polymers as solution crystallization inhibitors; (iii) use these phase boundaries to interpret non-sink, two-stage dissolution of IND dispersions, covering release rate, LLPS or crystallization, and the redissolution of colloidal versus crystalline material after a pH increase; and (iv) check physical stability at 40 °C/75% relative humidity (RH) over 100 days. The two-stage test, with a pH 2.2 stage followed by a shift using McIlvaine buffer, is meant to mimic the pH change from stomach to small intestine (Cristofoletti and Dressman, 2017; Fagerberg et al., 2010; Tres et al., 2016). C15

Background concepts

Two solubilities, two boundaries. A drug in water has a crystalline solubility, the equilibrium concentration over crystals, and a higher amorphous solubility, the concentration in equilibrium with the amorphous (liquid-like) drug. Between the two, the solution is supersaturated with respect to crystals but still a single phase; it can stay there for some time if nucleation and growth are slow or inhibited. Above the amorphous solubility, no further increase in dissolved drug is possible: the excess separates, either as a drug-rich liquid (LLPS) when crystallization is blocked, or as crystals. At pH 2.2, IND is un-ionized and both solubilities are low; the paper measures 1.61 µg/mL for γ and 2.8 µg/mL for α crystals, and an amorphous solubility of 36 µg/mL in the presence of 100 µg/mL HPMC. R1.a·R3.a

Polymers can move both boundaries. Solubilizing additives such as micellar surfactants raise both the crystalline and the amorphous solubility; typical ASD polymers usually have little effect on crystalline solubility, with some exceptions (Loftsson et al., 1996; Usui et al., 1997). The theoretical ratio of amorphous to crystalline solubility can be estimated from the free energy difference between the two forms with the Hoffman equation, modified by Murdande et al. (2010) to include ionization and water sorption by the amorphous form; the paper names these relations but does not print them or use them in a calculation. C53·C54·E2·E3

Detecting LLPS optically. The method used here, described in Taylor and Zhang (2016) and Raina et al. (2014), adds a concentrated drug solution slowly to buffer while monitoring two wavelengths with an in-situ UV probe. At 319 nm IND absorbs, so absorbance rises in proportion to dissolved drug. At 450 nm IND does not absorb, so any signal there is extinction from light scattering by particles or droplets. The amorphous solubility is taken as the concentration at which the 450 nm extinction starts to rise. The 319 nm signal then distinguishes the two possible new phases: if crystals form, dissolved drug is removed and absorbance falls sharply; if liquid droplets form, absorbance stays high. Two further checks are used: dynamic light scattering (DLS) sizes the new particles, and the drug's own fluorescence, weak in dilute solution, is "anticipated" to increase when drug molecules sit close together in a droplet. C16·C17

Indomethacin polymorphs and second harmonic generation. Second harmonic generation (SHG) is a non-linear optical effect in which a material doubles the frequency of incoming light; it occurs only in non-centrosymmetric crystals. In γ IND the carboxylic acid groups form centrosymmetric dimers (Chen et al., 2002), so γ crystals are SHG-inactive; amorphous material is isotropic and also inactive; in α IND the molecules are arranged with a twofold screw axis and the crystals are SHG-active. An SHG microscope (here a SONICC instrument, for second order nonlinear imaging of chiral crystals) can therefore detect α crystals in a suspension without filtering or drying it, steps during which a metastable form might transform. α IND is also known to crystallize preferentially from highly supersaturated solutions (Towler and Taylor, 2007). C18·C19·C45

The two polymers. Eudragit EPO (molecular weight approximately 47,000) carries dimethylaminoethyl groups; it is described as soluble up to pH 5, is highly soluble at pH 2.2 by ionization (Sun and Lee, 2015), and dissolves only very slowly at pH 5. It is also known to form a salt with IND through strong ionic interaction between the polymer's cationic nitrogen and the drug's carboxylic acid (Liu et al., 2012). HPMC (Pharmacoat 606, 6 mPa·s, molecular weight 35,600) is neutral and forms a gel on hydration. M3·M4·C49

Polymer-controlled release. When an ASD's release rate rises as its drug loading falls, the dissolution of the polymer rather than the drug is usually taken as rate limiting (Craig, 2002; Simonelli et al., 1969). The paper uses this reading for the HPMC dispersions. C31

Methods in detail

The study has four parts: phase boundaries with and without polymer, two-stage dissolution of dispersions and crystalline controls, identification of the precipitate that forms from the EPO dispersion, and an accelerated storage test. The logic is to measure the boundaries first and then read every dissolution curve against them.

Materials and dispersions

γ IND was purchased (Letco Medical); the α form was made by adding water to a saturated ethanol solution of IND and filtering (Kaneniwa et al., 1985). Dispersions were prepared by dissolving drug and polymer in ethanol:dichloromethane 50:50 (v/v), removing the solvent by rotary evaporation and drying overnight under vacuum at room temperature. The compositions listed in the methods are IND:EPO 3:7, IND:HPMC 5:5, 3:7 and 1:9, and IND:EPO:HPMC 3:5:2 (w/w). The EPO and ternary dispersions were ground and sieved to 106–250 µm; the HPMC dispersions were cryo-milled instead. All were amorphous by powder X-ray diffraction (PXRD) and were stored over phosphorus pentoxide. The crystalline controls were sieved to the same 106–250 µm fraction. Buffers were McIlvaine buffers made from 0.1 M citric acid and 0.2 M disodium phosphate: 49:1 (v/v) for pH 2.2, 49:51 for pH 5 and 24.5:75.5 for pH 6.8. M1·M2·M5·P1

Several samples used later are not in this list: IND:EPO 5:5 and 1:9 dispersions (mentioned with "results not shown"), and the 60% drug-loading dispersions of the storage test (IND:HPMC 6:4, IND:EPO 6:4 and IND:EPO:HPMC 6:2.9:1.1). Their preparation is presumably the same, but this is not stated. Note that the 60% ternary has an EPO:HPMC ratio of 2.9:1.1, not the 5:2 of the dissolution ternary. X5

Contribution 1: the two phase boundaries, with each polymer

Crystalline solubility. Excess γ or α crystals were shaken for 48 h at 37 °C in 20 mL pH 2.2 buffer with predissolved HPMC (50, 150 or 1000 µg/mL), EPO (50, 150 or 1000 µg/mL) or 150 µg/mL EPO plus 50 µg/mL HPMC, filtered through 0.45 µm PTFE and assayed by HPLC (C8 column; mobile phase 50% acetonitrile and 50% of a 0.5% acid printed as "tetrafuroic acid", probably a misprint; 254 nm) in triplicate. The authors note that these polymer concentrations approximately correspond to those reached after complete dissolution of the ternary dispersion in the acid stage. At pH 5, only α was measured, with 100 µg/mL HPMC, EPO or both; EPO was first dissolved at pH 2.2 and the buffer then adjusted to pH 5, because EPO dissolves very slowly at pH 5. P2·P3·C21·X13

Amorphous solubility. A 10 mg/mL ethanol stock of IND was pumped at 25 µL/min for 16 min into 50 mL of buffer (pH 2.2 or pH 5) containing predissolved polymer, stirred at 300 rpm in a jacketed beaker at 37 °C, which brings the nominal concentration to 80 µg/mL. Absorbance at 319 nm and extinction at 450 nm were recorded every 15 s with a 1 cm fiber-optic probe. Each polymer combination was run twice, and the replicates agreed within 2 µg/mL. For the fluorescence and DLS checks, IND (0–80 µg/mL) was added from the ethanol stock to buffer with 100 µg/mL HPMC; emission spectra were collected from 254 to 600 nm with excitation at 254 nm, and the same samples were sized by DLS. P4·P5·P6

Contribution 2: two-stage dissolution read against the boundaries

Each test used a dose containing 4 mg IND: 13.3 mg of the 30% dispersions (ternary, IND:EPO 3:7 and IND:HPMC 3:7), 40 mg of IND:HPMC 1:9, 8 mg of IND:HPMC 5:5, or 4 mg of α or γ crystals, the α crystals also with 9.3 mg EPO predissolved (the EPO content of a 13.3 mg IND:EPO 3:7 dose). P7

In scenario 1, the dose went into 50 mL pH 2.2 buffer (37 °C, 300 rpm), giving at most 80 µg/mL; after 1 h, 50 mL of 0.2 M disodium phosphate raised the pH to 5 and the volume to 100 mL, giving at most 40 µg/mL. In scenario 2, the acid stage used 25 mL (at most 160 µg/mL) and 75 mL of phosphate raised the pH to 6.8 (again at most 40 µg/mL). The two shifts reproduce the McIlvaine recipes exactly: 49 parts citric acid with 51 parts phosphate for pH 5, and 24.5 with 75.5 for pH 6.8. Concentration was followed for 8 h by UV scans every minute, as the 319 nm absorbance minus the 450 nm extinction, converted with calibration standards in pH 6.8 buffer. Selected samples were filtered and assayed by HPLC at 1 h (before the shift), 3 h and 8 h. For droplet detection, separate runs without UV monitoring were sampled (about 2 mL, filtered through 0.45 µm PTFE) and analysed by DLS and fluorescence. A further single-stage experiment dissolved 40 mg IND:HPMC 1:9 in 50 mL pH 2.2 buffer with 0, 150, 500 or 1000 µg/mL EPO predissolved, without a pH change. P7·P8·P9·P10·P11·E4

The number of replicate dissolution runs behind the error bars, and the replicate number and error type of the HPLC values, are not stated. X6

Contribution 3: identifying the precipitate

The IND:EPO dispersion turned turbid at pH 2.2. To find out what it formed, the suspension after 1 h dissolution of 13.3 mg IND:EPO 3:7 in 50 mL pH 2.2 buffer was examined in three ways. Aliquots of 0.5 mL were imaged by SHG on a glass slide (field of view 2 mm × 2 mm, scanned in about 40 µm steps), next to γ and α references. The filtered precipitate was measured by Raman spectroscopy (785 nm) against γ and α references and by PXRD (Figure S8). For scanning electron microscopy (SEM), the filtered crystals were washed and resuspended in 0.1 N HCl, air-dried on a cover glass and gold–palladium coated. P12·P13·P14·P15

Contribution 4: physical stability

Dispersions at 60% and 30% drug loading were stored in open vials at 40 °C/75% RH (saturated NaCl) and examined by PXRD (5–40° 2θ) at intervals up to 100 days. The 60% results are in Figure 12; the 30% results are in Figure S9. P16

Experimental analysis

The boundaries: HPMC leaves them alone, EPO raises them

Table A collects the crystalline solubilities of Table 1. HPMC, up to 1000 µg/mL, left the solubility of both polymorphs essentially unchanged (γ 1.35–1.61, α 2.47–2.9 µg/mL). EPO raised it with concentration, to 6.2 µg/mL for γ and 8.9 µg/mL for α at 1000 µg/mL. With 150 µg/mL EPO, adding 50 µg/mL HPMC gave no further increase. At pH 5, where IND is partly ionized, α solubility was 15 µg/mL without polymer, 17 µg/mL with HPMC, 27 µg/mL with EPO and 26 µg/mL with both (100 µg/mL each, Table 2). R1·R2·C20·C22

Table A. Crystalline solubility of IND polymorphs at pH 2.2, 37 °C, n = 3, mean ± SD (from Table 1). R1

The amorphous solubility was measured first with HPMC alone (Figure 2). With 100 µg/mL HPMC the 319 nm absorbance rose linearly with added drug and lost linearity at 36 µg/mL; the 450 nm extinction was negligible up to 36 µg/mL and then rose linearly (to about 0.31 at 80 µg/mL). The authors read this as a new scattering phase above 36 µg/mL, a threshold "nearly identical" to the theoretical amorphous solubility of IND at 37 °C calculated by Alonzo et al. (2010). DLS detected a droplet phase of 70–100 nm (polydispersity index, PDI, below 0.1) at or above 40 µg/mL; below that, scattering was too weak to analyse. IND's fluorescence between 325 and 475 nm was negligible below 40 µg/mL and grew with concentration above it, from about 88 to about 445 intensity units at the band maximum near 378 nm between 40 and 80 µg/mL (Figure 3). With 1000 µg/mL HPMC the amorphous solubility was the same, 35 µg/mL. R3.a–b·R4·R6·R7·C23·C25

Figure 2 (right) also shows how long HPMC can hold this level. After the 16 min addition the signal stayed on a plateau for about 1.5 h; then the absorbance fell, from about 2.5 to about 0.5 by 240 min, and the extinction changed, which the authors read as crystallization. R5·C24

With 100 µg/mL EPO alone (Figure 4) no amorphous solubility could be measured. During the addition the extinction rose and the absorbance fell at the same time, and many very small crystals were visible by eye. The absorbance stayed far above what a solution at the crystalline solubility would give, which the authors take as "suggesting" that the crystals are very small and absorb light, as reported for nanocrystals by Alonzo et al. (2011). A small source detail: this paragraph speaks of a "methanol" stock, whereas the method uses ethanol. R8·C26·X3

With HPMC present to prevent crystallization, EPO raised the amorphous solubility (Table B). The authors attribute this "presumably" to solubilization by EPO, consistent with the general observation that amorphous solubility rises when crystalline solubility rises (Lu et al., 2017; Raina et al., 2015b). For the 100/100 µg/mL mixture the text gives 51 µg/mL and Table 3 gives 50 µg/mL; the replicates agree within 2 µg/mL, so the difference is within the stated spread. At pH 5, no LLPS was detected up to 80 µg/mL by any of the three methods and no crystallization occurred (Figures S2–S5), as expected if ionization raises the amorphous solubility. R3·R9·C27·C28·X1

Table B. IND amorphous solubility at pH 2.2 with predissolved polymers, n = 2, replicates within 2 µg/mL (from Table 3). R3

The acid stage: three different behaviours

Figure 5 shows the pH 2.2 → 5 test for all eight formulations; Table C gives readings from it. Above the amorphous solubility, the UV signal includes light scattering, so these are apparent concentrations. R10

Table C. UV apparent IND concentration (µg/mL) in the pH 2.2 → 5 test; pH 2.2 up to 60 min (50 mL, at most 80 µg/mL), pH 5 afterwards (100 mL, at most 40 µg/mL); read from Figure 5, approximate. R10

IND:EPO 3:7. Because EPO is "highly soluble due to ionization" at pH 2.2, the dispersion dissolved rapidly and completely, reaching about 77 µg/mL within minutes, well above any amorphous solubility measured here. It then crystallized, "after about 15~20 min" in the results ("about 20 mins" in the methods), into a turbid suspension of many tiny crystals. In the plotted run the fall from about 76 to about 35 µg/mL already happens between about 10 and 20 min. The UV reading at 1 h, about 29.5 µg/mL, is not a solution concentration: the filtered supernatant held 8.8 ± 0.6 µg/mL by HPLC, which the authors describe as close to the α solubility in the presence of a high EPO concentration (Table A: 8.9 µg/mL with 1000 µg/mL EPO). Note by the AI writer: 13.3 mg of IND:EPO 3:7 in 50 mL gives about 186 µg/mL EPO, where Table A suggests an α solubility nearer 3.4 µg/mL (the 150 µg/mL row), so the 1 h supernatant may still have been supersaturated with respect to α, or held particles that passed the filter. IND:EPO dispersions at 5:5 and 1:9 behaved very similarly (data not shown). The profile matches the earlier reports of Sarode et al. (2013) and Sun and Lee (2015). R11·R12·C29·C30·C60·X4·X7·X20

IND:HPMC. Release from the HPMC dispersions was slow and became faster as drug loading fell (at 60 min about 18.5, 24.5 and 39.5 µg/mL for 50, 30 and 10% drug loading), which the authors call "indicative of a polymer controlled dissolution mechanism". The 30% dispersion dissolved "substantially slower" than the EPO and ternary dispersions at the same loading, supersaturation rose gradually and no crystallization was observed. The 10% dispersion was biphasic, fast to about 30.5 µg/mL in 10 min and then much slower; its concentration "appeared to just reach" the amorphous solubility by 60 min. The 30% and 50% dispersions stayed well below it before the pH shift. Filtered 60 min samples agreed: DLS found droplets of about 65–80 nm (PDI about 0.24 ± 0.03) only in the 10% sample, and only that sample showed a (weak) change in fluorescence. R10.c–e·R13·R14·C31·C32·C33

Ternary IND:EPO:HPMC 3:5:2. The ternary showed both benefits, which the authors attribute to EPO (fast release) and HPMC (sustained concentration). Complete release would give "about 150 µg/mL EPO and 50 µg/mL HPMC" and hence, from Table B, an amorphous solubility "around 56 µg/mL", against a nominal 80 µg/mL. The extinction rose sharply together with the absorbance (Figure S6), the signature of LLPS. The UV estimate went above the maximum possible 80 µg/mL (about 93.5 µg/mL at 60 min), which the authors explain by the change in extinction coefficient when nanodroplets are present (compare the slope change in Figure 2). Separate runs sampled at 30 min and 1 h held nanospecies of about 75 nm (PDI about 0.1), and fluorescence spectra indicated drug-rich species (Figure S7). Without a pH change, the ternary crystallized after 3 h. The authors suggest that this profile "might lead to rapid therapeutic onset" and less irritation in the lumen, because solid drug particles would not touch the membrane. R15·R16·C34·C35·C36·C37

Note by the AI writer (not discussed in the paper): from the dose and the 3:5:2 ratio, complete release of 13.3 mg into 50 mL gives about 133 µg/mL EPO and 53 µg/mL HPMC, somewhat less EPO than the "about 150" quoted, so the 56 µg/mL taken from the 150/50 row of Table 3 is, if anything, a slight overestimate of the ternary's amorphous solubility in the acid stage. X11

Crystalline controls. α dissolved faster than γ, as expected from its slightly higher solubility, and predissolved EPO increased the α dissolution rate, "presumably" because it raises the crystalline solubility. R17·C38

EPO added to a low-loading HPMC dispersion. Figure 6 tests whether dissolved EPO alone can lift the 10% HPMC dispersion. With 0, 150, 500 and 1000 µg/mL EPO predissolved, the concentration at 20 min was about 31.5, 45.5, 57 and 65.5 µg/mL, and the maxima about 41, 53, 69 and 77.5 µg/mL; the authors report that the concentration at which the biphasic pattern set in increased with EPO concentration. Note by the AI writer: at 0, 150 and 500 µg/mL EPO the curves also fall after about 115–160 min (to about 31, 32.5 and 50.5 µg/mL at 300 min), which the text does not discuss; only the 1000 µg/mL curve stays at about 77 µg/mL. The same 1:9 dispersion without EPO reads about 39.5 µg/mL at 60 min in Figure 5 but about 35.5 µg/mL in Figure 6. R18·C39·X19·X12

After the shift to pH 5: droplets, crystals and undissolved solid

Raising the pH to 5 changes the picture because IND (pKa 4.42) becomes partly ionized and both its crystalline solubility (α: 15 µg/mL, Table 2) and its amorphous solubility (above 80 µg/mL, no LLPS) rise. Table D lists the HPLC values the paper reports. R2·R9

Table D. IND concentration at pH 5 by HPLC of filtered samples; complete dissolution gives 40 µg/mL (from the text). The replicate number and the meaning of ± are not stated. R17·R19·R20·R21

Ternary. The droplet phase "disappeared rapidly", seen as a sharp fall in extinction (Figure S6), "consistent with an increase in IND amorphous solubility due to ionization" (Indulkar et al., 2015). In Figure 5 the apparent concentration falls from about 93.5 to about 43 µg/mL within about 2 min of the shift and settles at about 39–40 µg/mL; by HPLC it was 39.1 µg/mL at 3 h, close to the 40 µg/mL of complete dissolution and slightly supersaturated with respect to α in the presence of both polymers (26 µg/mL). The authors conclude: "These observations thus demonstrate" that the droplet phase formed above the amorphous solubility serves as a drug reservoir, readily available after dilution to higher pH. R20·C40·C41·C42

HPMC dispersions. The 30% HPMC dispersion also achieved complete release (39.3 µg/mL), but it needed about 90 min after the shift to dissolve its remaining solid. The 10% dispersion needed only about 10 min, but its low drug loading "might render this system impractical for high dose formulations". R21·C43

EPO dispersion. The crystals precipitated at pH 2.2 dissolved rapidly after the shift, "albeit incompletely": 32.1 µg/mL after 3 h, against 40 µg/mL for complete dissolution. The authors call it surprising that this crystalline precipitate redissolved at a rate comparable to the amorphous IND:HPMC 3:7 dispersion and substantially faster than the reference α crystals, with (24.6 µg/mL at 8 h) or without (17 µg/mL at 8 h) predissolved EPO. In Figure 5 the precipitate rises from about 21 to about 32 µg/mL within about 20–30 min of the shift and then levels at about 33 µg/mL; the 30% HPMC dispersion starts lower but reaches about 40 µg/mL by about 150 min. "Comparable" therefore applies to the first minutes only. R19·R10.b·R10.d·C4·X8

What the precipitate is

The SHG images (Figures 8 and 9) settle the polymorph without filtration: the γ reference is dark, the α reference bright, and the suspension from the EPO dispersion shows many bright spots, "indicating that the precipitated material was indeed the metastable α form". In suspension the Raman signal was too weak to identify the form; after filtration the spectrum matched α (Figure 11: bands near about 1688, 1679, 1649, 1610 and 1579 cm⁻¹, and no γ band near 1698 cm⁻¹), and PXRD of the same sample agreed (Figure S8). The SEM images (Figure 10) show butterfly-shaped particles "mostly less than 10 µm" with fine needles protruding from the wings; the discussion gives a "uniform particle size of 5~10 µm". R23·R24·R25·C46

Why does this crystalline precipitate dissolve so fast? The authors first consider solubility: EPO raises IND solubility and so, by the Noyes–Whitney equation, should raise the dissolution rate. But the α crystals with matched predissolved EPO dissolved much more slowly than the precipitate, so "this effect alone cannot explain" it. Their explanation is size: the rapid dissolution "is likely explained by the very small size of the crystals forming the precipitated aggregates"; the discussion adds the EPO solubilization and suggests that the precipitate is "likely" to be optimally wetted because it formed directly in the aqueous solution. C47·C48·E1

Note by the AI writer: the size explanation is plausible but untested here. The reference α crystals were 106–250 µm, roughly 10 to 50 times larger than the precipitate, and no small α crystals of matched size were tested; wetting was not measured. X10

A shift to pH 6.8 hides the differences

In scenario 2 (Figure 7), the acid stage was more concentrated (at most 160 µg/mL). The ternary read about 166 µg/mL, the EPO dispersion peaked at about 152 µg/mL and fell to about 55–61 µg/mL, and the 30% HPMC dispersion rose to about 29 µg/mL by 60 min. After the shift to pH 6.8, "little discrimination could be seen": all three 30% dispersions reached the target 40 µg/mL within 10 min, α crystals within about 20 min, and only γ crystals were slow, needing more than 4 h. One hour after the shift, the five formulations ranged from about 3 to 39.5 µg/mL at pH 5 but from about 27.5 to 41.5 µg/mL at pH 6.8 (read from Figures 5 and 7). R22·R10·C44

Physical stability

At 60% drug loading and 40 °C/75% RH (Figure 12), the IND:HPMC 6:4 dispersion showed sharp PXRD peaks after 5 days (near about 8.5, 12, 14.6, 18.6 and 22.1° 2θ) and many stronger ones after 100 days. The IND:EPO 6:4 dispersion showed only faint features after 5 days and weak peaks after 100 days. The authors read this as much slower crystal growth with EPO, as they expected from the EPO–IND ionic interaction that HPMC lacks, although both binaries crystallized within 5 days. The ternary IND:EPO:HPMC 6:2.9:1.1 kept an amorphous halo for more than 100 days, outperforming both binaries. The authors call this "unexpected", since they had anticipated similar or slightly worse stability than the EPO binary when part of the EPO is replaced by HPMC, and they did not explore the mechanism. At 30% drug loading the ternary was also more stable than the HPMC binary (Figure S9). R26·R27·C49·C50·C51·C52

Discussion

One profile, three processes

The authors' central message is that an ASD's concentration–time profile is not set by one property. It is "a complex interplay of drug release rate, precipitation kinetics and outcome, and precipitate redissolution rate, whereby each of these processes is highly dependent on the polymer(s)", and in the conclusions also on drug loading and pH. The three dispersions illustrate each process separately. C6

A third, slower path is the HPMC binary: release is limited by the gelling polymer, "probably through a matrix-regulated diffusion mechanism", so the 30% dispersion never reaches the amorphous solubility in the acid stage and needs about 90 min after the shift to dissolve completely. The authors stress that HPMC is "an excellent crystallization inhibitor in both the solid matrix and solution phase", yet "formulation of an ASD per se does not necessarily ensure rapid drug release"; drug loading and polymer selection are "critical formulation attributes". C64·C65·R21

Solubilization versus inhibition

The authors separate two contributions to the ternary's higher concentrations: inhibition of crystallization by HPMC and "some extent of solubilization by EPO". EPO's effect on both solubilities is "presumably" due to drug–polymer interactions in solution; in a supersaturated solution containing both polymers, they write, the thermodynamic activity of IND "was lowered by the solubilization effect of EPO", so LLPS began at a higher concentration. They argue that such mechanisms need to be deconvoluted as formulations become more complex, because they "may have different implications in vivo". The activity itself was not measured; the evidence is the shift in both solubility boundaries. C55·C56·C57

Crystalline precipitate is not the end of the story

For the EPO binary, the paper offers what the authors call "one of the few observations describing the redissolution characteristics of a crystalline precipitate". It has been "widely speculated" that a drug that precipitates in crystalline form in vivo will redissolve readily; "This widely held supposition is supported by the observations of this in vitro study." The precipitate was the more soluble metastable α form, small (5–10 µm), dissolving in the presence of EPO released from the ASD, and probably well wetted. They add that redissolution "may occur even more rapidly in vivo" because of bile salt micelles in the small intestine. Overall, the binary EPO ASD still improved dissolution "despite the occurrence of crystallization at acidic pH". C61·C62·C63·C69

What a dissolution test can hide

The pH 6.8 scenario parallels Terebetski et al. (2014), who found for ibuprofen sodium that a polymeric inhibitor extended supersaturation and gave a higher Cmax and earlier Tmax in vivo, although the time to complete release after a shift to pH 6.8 did not distinguish the formulations. The IND results after the shift to 6.8 were similarly undiscriminating, whereas the shift to pH 5 separated them, with the ternary best and all dispersions better than the crystalline powders. Hence the recommendation that for ionizable drugs "it may be informative to include an intermediate pH", or a food effect, to improve the discriminative power of in vitro tests; the conclusions state more firmly that the comparison "demonstrates the importance" of this. R31·C66·C67

Note by the AI writer (not discussed in the paper): for a weak acid like IND, the pH after the shift sets how much the solubility rises, and so how much any remaining solid can redissolve. At pH 6.8 even γ crystals eventually reach 40 µg/mL; at pH 5 the α solubility is only 15 µg/mL, so differences between formulations are no longer masked by a large solubility excess. The intermediate pH thus works as a stress test of the form in which the drug leaves the acid stage.

Limitations

Limits stated by the authors

• The mechanism behind the superior physical stability of the ternary dispersion "was not explored in depth". C52
• In vivo fluids "are much more complex" than these simple buffers; in the small intestine, bile salts and phospholipids "may impact both the solubility and crystallization tendency of the system". C68
• The 10% HPMC dispersion, though fast after the shift, has a drug loading that "might render this system impractical for high dose formulations". C43

The AI writer's own judgement

• Apparent versus dissolved concentrations. In-situ UV includes scattering by crystals and droplets. For the EPO dispersion at 1 h, UV gave about 29.5 µg/mL against 8.8 µg/mL by HPLC; the ternary read above its nominal maximum (about 93.5 against 80 µg/mL; about 166 against 160 µg/mL at pH 6.8). Acid-stage curves above the amorphous solubility should be read as apparent values. X7
• "Comparable rate" holds only at the start. The crystalline precipitate rose faster than the HPMC 3:7 dispersion in the first minutes after the shift but stopped at about 80% dissolved (32.1 of 40 µg/mL), while the HPMC dispersion reached complete release. No statistical test supports "significantly outperformed". X8
• The explanations are untested. Crystal size (and wetting) as the cause of fast redissolution lacks a size-matched α control; EPO solubilization and lowered activity rest on solubility shifts only, since the amorphous solubility with EPO alone could not be measured. X10·C27·C56
• "Reservoir" here means redissolution. The droplets' disappearance after the shift shows that the excess drug is available once the amorphous solubility rises, not that the droplets sustain membrane flux, which was not measured. The extinction trace that shows the disappearance is in Figure S6, not available for this reading. The droplets' amorphous nature is inferred from optical methods; SHG, Raman and PXRD were applied only to the EPO precipitate. X9·X18
• Replicates and sample lists. The number of dissolution runs and the replicate basis of the HPLC values are not stated. The same 1:9 HPMC dispersion differs by about 4 µg/mL at 60 min between Figures 5 and 6. Several dispersions (IND:EPO 5:5 and 1:9, all 60% samples) are not in the preparation list, and the stability ternary has a different EPO:HPMC ratio from the dissolution ternary. X6·X12·X5
• Small source inconsistencies. The ternary's LLPS onset at 100/100 µg/mL is 50 µg/mL in Table 3 and 51 µg/mL in the text; the onset of EPO-ASD crystallization is given as 15–20 min, while the plotted fall begins near 10 min; one paragraph refers to a methanol stock although the method uses ethanol. None changes the conclusions. X1·X4·X3
• Scope. One drug, two polymers and one ternary ratio for dissolution; simple buffers without bile components; no absorption data. X5

Conclusion

Three measured facts carry the paper. First, at pH 2.2 HPMC holds IND supersaturated up to its amorphous solubility (36 µg/mL) without crystallization for about 1.5 h, while EPO raises both the crystalline and the amorphous solubility but cannot stop crystallization. Second, in a pH 2.2 → 5 test the ternary EPO–HPMC dispersion releases fast above the amorphous solubility into about 75 nm nanodroplets that redissolve within minutes of the pH rise (39.1 of 40 µg/mL), the HPMC dispersion releases slowly but completely, and the EPO dispersion crystallizes into small α crystals that redissolve quickly but only to 32.1 µg/mL, still well ahead of reference crystals. Third, a shift to pH 6.8 hides these differences. The authors' explanations, EPO solubilization and the small size of the precipitated crystals, are reasonable but were not tested directly, and the in vivo meaning of the nanodroplet reservoir was not examined. R3·R5·R8·R10·R19·R20·R22·C6·C48·C56

For formulation, the paper argues for pairing a fast-dissolving polymer with a crystallization inhibitor, for testing ionizable drugs at an intermediate pH, and for characterizing any precipitate, amorphous or crystalline, and its size, because these decide how fast the drug comes back into solution. C59·C67·C70