How a polymer that does not dissolve lowers the highest concentration an amorphous drug can reach

Background and motivation

Many new drug candidates dissolve poorly in water (Section 1). Surfactants, cosolvents and cyclodextrins raise the amount of dissolved drug but often lower its apparent permeability, so absorption gains little. An amorphous solid dispersion (ASD) works differently: it raises the free drug concentration, and with it the drug's thermodynamic activity, which is the driving force for passive transport across the gut wall (Raina 2015; Miller 2012). C5 Pure amorphous drugs tend to crystallize, so a polymer is added to inhibit crystallization in the solid and in the supersaturated solution. C6 Enteric polymers such as Eudragit and hydroxypropylmethylcellulose acetate succinate (HPMCAS) are popular because they process well by hot-melt extrusion and spray drying and inhibit solution crystallization; because they do not dissolve at low pH, their ASDs are "expected" to sit in the stomach as suspensions. C6 Insoluble polymers are also used on purpose in controlled-release products and implants, where they turn release from dissolution-controlled to diffusion-controlled (Sun 2015). C7

The ceiling on supersaturation is the drug's "amorphous solubility", the highest free concentration the amorphous form can give before a drug-rich liquid phase separates. C8 It is not a true equilibrium solubility, because the amorphous solid can always crystallize, but it can be measured when crystallization is slow (Ilevbare 2013; Murdande 2010). C8 Two strands of earlier work suggested that this ceiling is not fixed. Trasi and Taylor (2015) found that a second drug, miscible with the first in the amorphous state, lowered the maximum supersaturation of each, and Huang et al. (2016) analysed drug–polymer blends thermodynamically and predicted a reduced amorphous solubility. C9 If a polymer lowers the ceiling, supersaturation, crystallization and bioavailability all change. C2

What had not been studied was how much a polymer that does not dissolve in the medium lowers the maximum achievable concentration; the authors call this "poorly explored" and point to suspension stability, the thermodynamic stabilization of the drug and in vivo performance as the reasons it matters (Section 1). Their hypothesis: the drug's solubility falls when it is mixed with a poorly soluble polymer, by dilution, and the strength of drug–polymer interactions changes how much. C10 Lopinavir (LPV) was chosen because it crystallizes slowly from water, so its concentration can be followed for hours (Li 2016). C11 Six polymers that stay undissolved at pH 3 were used: cellulose acetate (CA), ethylcellulose (EC) and Eudragit RL PO (EUD), which never dissolve in water, and the enteric cellulose acetate phthalate (CAP), HPMCAS and hydroxypropylmethylcellulose phthalate (HPMCP), which do not dissolve at low pH. Hydroxypropylmethylcellulose (HPMC), which dissolves, served as the positive control (Figure 1). C12·M2–M8

Background concepts

Amorphous solubility and the solubility advantage. The paper starts from the standard relation between the two solubilities of a drug (Eq. 2):

Here $S_A$ and $S_C$ are the solubilities of the amorphous and crystalline forms, ${\mu }_A$ and ${\mu }_C$ the chemical potentials of the drug in the amorphous and crystalline solid, $R$ the gas constant and $T$ the temperature. In plain terms: the amorphous form dissolves to a higher concentration than the crystal by a factor set by how much more free energy it carries. E2 The authors note the assumptions: the drug's activity coefficient in solution changes little between the two concentrations, and the amorphous solid must equilibrate with the solution, which crystallization, water uptake by the solid and relaxation of the glass all work against. C41 For LPV at pH 3 and 25 °C the measured values are 3.1 ± 0.1 μg/mL (crystal) and 17.4 ± 0.3 μg/mL (amorphous), a ratio of about 5.6. R1.a·R1.b·E2

The free energy difference itself can be written from melting data (Eq. 3):

Here $\Delta H_m$ is the heat of fusion, $T_m$ the melting temperature, and $C_{p,L}$ and $C_{p,X}$ the heat capacities of the liquid (amorphous) and crystalline drug. In plain terms: the excess free energy of the amorphous form comes from the heat of fusion it has not released, corrected for how the heat capacities of the two forms differ between the melting point and the experimental temperature. E3 Note by the AI writer (not discussed in the paper): the first term is printed as $\Delta H_m(T-T_m)/T_m$, which is negative below the melting point; the matching term of Eq. 6 below, $\frac{\Delta H_m}{R}(\frac{1}{T}-\frac{1}{T_m})$, has the opposite sign, which is the usual one. Eq. 3 most likely carries a sign misprint. X1 No melting data for LPV are given, so neither equation is evaluated in the paper. E3

How a polymer and water lower the drug's chemical potential. When the drug is mixed with a polymer, its chemical potential falls below that of the pure amorphous drug. Flory–Huggins theory describes this (Eq. 4):

Here $\Delta {\mu }_{drug}$ is the change in the drug's chemical potential, ${\Phi }_{drug}$ and ${\Phi }_{poly}$ are the volume fractions of drug and polymer, $r=V_{poly}/V_{drug}$ is the ratio of their molar volumes and $\chi$ is the drug–polymer interaction parameter. In plain terms: dilution by the polymer lowers the drug's chemical potential, and a negative $\chi$, meaning drug and polymer attract each other more than themselves, lowers it further. E4·C44 Water taken up by the dispersion lowers it again, through the Gibbs–Duhem equation (Eq. 5):

Here $x_{drug}$, $x_{poly}$ and $x_{water}$ are mole fractions in the hydrated dispersion and $a_{water}$ is the water activity. In plain terms: the more water the solid absorbs as the water activity rises toward 1, the lower the drug's chemical potential. E5·C44 Huang et al. combined Eqs. 2–5 into one expression for the solubility of a drug formulated in an ASD, $S_A^{ASD}$ (Eq. 6):

In plain terms: the solubility of the drug from an ASD is the amorphous solubility advantage, reduced by the polymer-mixing term and by the water term. E6 The authors set three limits on it. It applies only when drug and polymer form one phase that stays one phase after hydration, and only when the polymer is insoluble enough that the composition of the undissolved solid does not change, as for an enteric polymer at low pH. C45 It ignores the known change of $\chi$ with water (Rumondor 2010), and $\chi$ is hard to measure, so the model "is probably best applied in a qualitative manner to rationalize observed trends". C46 It is used in that way throughout: no $\chi$, heat capacity or sorption value is put into it. E6·C51

The "ideal" line. Figure 2 carries a dashed line from zero to 17.4 μg/mL at 100% drug loading, labelled "ideal behavior": the amorphous solubility scaled by the drug's weight fraction. E7 The authors describe concentrations above it as weaker-than-ideal and below it as stronger-than-ideal interaction. C20·C51 Note by the AI writer (not discussed in the paper): this line is not the $\chi =0$ case of Eq. 4. With $\chi =0$ and a polymer much larger than the drug, Eq. 4 gives an activity of ${\Phi }_{drug}\exp \left({\Phi }_{poly}\right)$, about 0.82 rather than 0.5 at half drug, before the water term is added. The position of a curve relative to the dashed line is therefore a comparison with simple dilution by weight, not a reading of the sign of $\chi$. X2

Liquid–liquid phase separation as a measure of the amorphous solubility. When a drug solution is pushed past its amorphous solubility, a drug-rich liquid phase separates as small droplets that scatter light. The concentration at which UV extinction at a non-absorbing wavelength starts to rise marks that point and has been found "in good agreement" with the amorphous solubility (Ilevbare 2013). C13

Reading the mixing state of the solid. Two tools are used. A single glass transition temperature (Tg) between those of the two components means one mixed phase; two Tgs, each near a pure component, mean two phases. C27·C30 Water sorption is the second tool: if drug and polymer do not interact, a dispersion should take up the weighted sum of what each takes up (Eq. 1):

Here $M$ is the equilibrium water content, $w$ the weight fraction and $D$ and $P$ drug and polymer. In plain terms: the expected uptake is the weight-averaged uptake of the two components. A dispersion that takes up less than this is read, after Crowley and Zografi (2002), as one in which drug–polymer contacts replace sites that would otherwise bind water. E1·C34

Methods in detail

The study has three parts: measuring the ceiling (crystalline and amorphous solubility) and how far each ASD falls below it; finding out whether drug and polymer are mixed and whether they interact; and testing how the ASDs crystallize when nothing in the medium protects them. Everything was done in 50 mM citric acid buffer at pH 3 and 25 °C, where all polymers except HPMC stay undissolved. M10·A1

Part 1: the ceiling and how far each ASD falls below it

Reference solubilities. Excess crystalline LPV was stirred at 300 rpm in 10 mL of buffer at 25 ± 1 °C for 48 h, alone or with an excess of each polymer, then ultracentrifuged (35,000 rpm, 30 min) and the supernatant analysed by HPLC at 210 nm, in triplicate. P1·P8 The amorphous solubility was measured two ways. A 5 mg/mL methanolic LPV solution was pumped at 40 μL/min into 15 mL of stirred buffer while UV spectra were taken every 10 s; the concentration at which extinction at 300 nm began to rise marked liquid–liquid phase separation. P2 Separately, a 10% drug-loading (DL) LPV–HPMC film holding 0.3 mg LPV was equilibrated with 10 mL of buffer for 48 h; HPMC dissolves and leaves the drug at its amorphous solubility. P3

Maximum achievable concentration from each ASD. Each film held 2 mg of solids, cast by rotary evaporation from 1 mL of a 2 mg/mL methanol–dichloromethane solution directly onto the inside of a 20 mL vial, so that a large surface faces the medium. P4·C14 Ten mL of buffer containing 20 μg/mL of dissolved HPMC was added; this small amount of HPMC blocks crystallization from the supersaturated solution and was taken not to change the solubility. P5·C15 Samples were taken at 3 h, taken as the gastric transit time of an enteric ASD, and at 48 h, an empirical time frame for solubility; they were filtered (0.45 μm), the 48 h samples were also ultracentrifuged before HPLC, and each point is a triplicate. P5·C16 Drug loadings were 15, 33, 50, 70 and 85%; the 100% point of every curve is the Table 1 amorphous solubility, set by convention rather than measured on an LPV film. R2·R3 Release was also followed to 7 days, but those data are not shown. R4 A physical mixture of amorphous LPV and HPMCAS was tested the same way at 48 h; its preparation is not described. P14·X16

Note by the AI writer (not discussed in the paper): with 2 mg of solids in 10 mL, the films hold more drug than the amorphous solubility at every loading (0.3 mg LPV at 15% DL is 30 μg/mL if all of it dissolved), so each curve measures a ceiling rather than complete dissolution. The same arithmetic matters for the composition of the residue: an 85% DL film holds only 0.3 mg of polymer, so if the polymer dissolved at the 25 μg/mL upper end of the authors' estimate, most of it would leave and the residue would approach about 97% DL. The authors' estimate of a change of ± 2% in DL holds only near the 3 μg/mL end. P4·R5·X3

Part 2: are drug and polymer mixed, and do they interact?

Glass transitions. Dispersions of 200 mg were made by rotary evaporation, dried overnight under vacuum and scanned by modulated DSC at 5 °C/min (± 1 °C every 60 s). The second heating was used, except for EUD systems, where the pure-EUD Tg disappeared on reheating and the first heating was used. P9 Powders were also held at 97% RH (saturated potassium sulfate) and 25 °C for 48 h and rescanned in sealed pans. P10

Infrared spectra. Films were spin-coated on KRS-5 at below 18% RH, to avoid water-induced phase separation, and dried overnight; spectra (128 scans) covered the carbonyl region near 1590–1710 cm⁻¹ and the OH-stretch region near 3380–3540 cm⁻¹. P11

Water sorption. Ten to 20 mg of amorphous powder was dried at 60 °C and stepped from 10% to 90% RH under nitrogen; each measured isotherm was set against Eq. 1. P12·E1 The temperature of these runs and the number of replicates are not stated. X18

Part 3: crystallization without help from the medium

Fifty-percent DL films and an amorphous LPV film were exposed to buffer without HPMC, so that only the polymer inside the ASD could inhibit crystallization; the concentration was followed from 2 to 48 h by filtration and HPLC. P6·C24 The solid was examined by polarized light microscopy at 8, 16, 24, 36 and 48 h. P7

Experimental analysis

The ceiling is well defined and the polymers do not move it

The two pH 3 measurements of the amorphous solubility agree: 17.4 ± 0.3 μg/mL from the HPMC dispersion and 18 ± 2 μg/mL from the onset of liquid–liquid phase separation, close to the 16–17.1 μg/mL reported at pH 6.8 by three methods in earlier work (Table 1). R1.b–f The crystalline solubility, 3.1 ± 0.1 μg/mL alone, stayed between 2.6 ± 1.5 and 3.6 ± 0.2 μg/mL with each polymer present, which the authors describe as similar with slight variations. R1.a·R1.g–l·C17 So the window an amorphous form can open, from about 3 to about 17 μg/mL, is the same in every system, and differences between the ASDs are differences in how much of it they use. The methods describe the polymer as added in "excess" for these tests, the results as a "trace"; no amount is given. X15

Main result: insoluble polymers lower the maximum concentration

Table A. LPV concentration (μg/mL) after 48 h at pH 3 with 20 μg/mL HPMC in the buffer, read from Fig. 2B (approximate), with the Tg pattern of the dry dispersions from Fig. 5. The "ideal" row is the paper's dashed line, 17.4 μg/mL times the drug weight fraction. R3·R8·E7

Table A is the core of the paper. With HPMC the concentration sat at or slightly above the amorphous solubility at every loading, from about 18.9 μg/mL at 15% DL to 17.5 at 85%, with no trend. R3.a·C18 With every insoluble polymer the concentration rose with drug loading and stayed below the ceiling. R3·C19 The spread at low loading is large: at 15% DL, about 13.8 and 12.8 μg/mL for CA and EUD, 4.3 for HPMCAS, 3.1 for CAP, 0.6 for HPMCP and 0.3 for EC. R3.b–g The authors summarize it this way: HPMCP gave the lowest values, EUD and CA had "much lower impacts" and approached the amorphous solubility, HPMCAS suppressed "to a moderate extent", slightly above the ideal line, CAP at 15–50% DL and EC at 50–85% DL were nearly ideal, and above 50% DL CAP "seemed" to suppress more while EC suppressed more below 50%. C20

At the two lowest loadings, EC (about 0.3 and 3.1 μg/mL) and HPMCP (about 0.6 and 2.5 μg/mL) fell below or close to the crystalline solubility measured with those polymers (3.6 ± 0.2 and 3.4 ± 0.5 μg/mL); the authors call this "extreme solubility suppression". R3.e·R3.g·R1.k·R1.l·C63 In their reading, an ASD whose drug sits at or below the crystal's chemical potential has no driving force to crystallize and "should remain physically stable for long periods of time", dry and in suspension, while still releasing once the enteric polymer meets intestinal pH. C64 That last step was not tested: stability was examined only at 50% DL, and release at higher pH not at all. X8

Table B. LPV (μg/mL) at 33% DL after 3 h and 48 h, read from Fig. 2A and 2B (approximate). R2·R3

Is 48 h the ceiling? The authors find the 3 h pattern "similar", with slightly less drug, and attribute the difference to release kinetics ("can be attributed"); HPMC, EC and CA released about the same at both times, HPMCAS, CAP and HPMCP more at 48 h. C21·C22 They later use the similarity of 3 h and 48 h "in most instances" as support for an undissolved solid whose composition does not change. C52 Table B shows that the change is not small for every system: HPMCAS at 33% DL almost doubled, CAP more than doubled, and EUD at 15% DL rose from about 6.9 to about 12.8 μg/mL. R2·R3·X4 The AI writer's reading: for HPMCAS, CAP, HPMCP and EUD the 48 h values may still be on their way up, so they are lower bounds on the maximum achievable concentration rather than proven equilibrium values; the 7-day data that could settle this are not shown, and the dissolved polymer could not be measured because most assay values were below detection. X5·R4·R5

Mixing is what lowers the ceiling

Figure 9 is the cleanest control in the paper. A physical mixture of amorphous LPV and HPMCAS, the same two materials at the same loadings, reached about 17.0 μg/mL at 15% DL and 18.0–18.5 μg/mL from 33 to 85%, at or above the amorphous solubility; the HPMCAS dispersion reached only about 4.3, 9.3, 12.4, 14.2 and 15.7 μg/mL. R14 The authors: suppression was "clearly not observed" for the physical mixture but was composition-dependent for the miscible ASD. C50 Their explanation is that components of a phase-separated system "are not intimately mixed with each other, and thereby behave independently", so the mixing terms of the model do not apply. C49

The Tg data line up with this. CA dispersions showed two Tgs near the two components (about 155–180 °C and 61–79 °C, against 184 and 69 °C) at every loading, and the authors call LPV "largely immiscible" with CA. R8.a·M1·M6·C27 EUD dispersions showed two Tgs at 15–50% DL and one at 70 and 85%, where two close Tgs were "probable" but unresolved. R8.b·C28 CAP showed two Tgs at 15 and 33% DL (about 141/51 and 131/56 °C) and one from 50% DL, "indicating possible increased miscibility" with loading. R8.c·C29 EC, HPMCAS and HPMCP each showed one Tg, falling from about 90–92 °C at 15% DL to about 69–79 °C at 50–85%, read as miscible. R8.d–f·C30 After 48 h at 97% RH the picture was the same with lower Tgs, except that CAP now showed one Tg above 15% DL; the authors conclude that water did not change miscibility "to any considerable extent". R9·C40 Note by the AI writer: the dry CAP dispersion at 33% DL has two Tgs and the humidified one a single Tg, so for CAP at that loading the hydration effect is unresolved; and the discussion cites "Figs. 5 and 9" for the hydrated DSC, whereas Figure 9 is the physical-mixture experiment and the hydrated data are in the supplementary Fig. A. X12·X11 The EUD Tgs come from the first heating and both lie outside the two pure values (about 74–83 °C above both, about 47–51 °C below both at 15–33%), so they show two events but not which phase is which. X13

Read against Table A, the two clearly phase-separated systems, CA and EUD, are the two that suppress least, and the physical mixture does not suppress at all. CAP is the interesting middle case: at 15% DL, where it shows two Tgs, it already lowers LPV to about 3.1 μg/mL, and its departure from the ideal line comes above 50% DL, where it is one phase. R3.f·R8.c·C20

Which polymers interact with lopinavir

Table C. Water uptake (weight gain, %) at 90% RH of the polymer alone, the 50% DL dispersion and the additive prediction of Eq. 1, read from Fig. 7 (approximate); amorphous LPV alone takes up about 4.2%. R12·E1

In Table C, CA and EUD dispersions took up about what the additive model predicts, which together with their two Tgs the authors read as "negligible drug-polymer interactions". R12.a–b·C35 HPMCAS and HPMCP dispersions took up clearly less (about 5.4 against 6.8% and 5.1 against 7.3%), read as molecular interaction. R12.e–f·C35 HPMCP alone took up more water than HPMCAS alone (about 10.4 against 9.4%), yet its dispersion took up less, which "may indicate" a stronger interaction with HPMCP. R12.e–f·C36 EC and LPV take up about the same, so EC cannot be judged this way. C33 Across loadings at 90% RH (Figure 8), HPMCAS and HPMCP stayed below the additive line at every loading (for HPMCP at 33% DL, about 6.4 against 8.3%), while CAP followed it below 50% DL and fell below it above, which the authors read as stronger CAP interaction at high loading. R13·C38·C39 Note by the AI writer: at 50% DL the CAP dispersion is already below its prediction (about 5.9 against 6.6%), though the text calls the 50% isotherm close to additive, and "significantly lower" is stated without replicates or a test. X17

The infrared spectra point the same way. Amorphous LPV shows a carbonyl doublet at 1648 and 1674 cm⁻¹, which "may be assigned" to its amide C=O groups. R10·C31 With CAP the 1648 cm⁻¹ peak weakened as polymer increased and the doublet merged into one broad peak at 1660 cm⁻¹; with HPMCAS and HPMCP the 1674 cm⁻¹ peak became stronger than the 1648 cm⁻¹ peak as loading fell; EC left the carbonyl region unchanged, but its own OH band moved from 3481 to 3470 cm⁻¹ with LPV, "consistent with a change in hydrogen bond interactions"; CA and EUD showed no shifts (spectra not shown). R10·R11·C32 In the discussion this becomes IR "revealed" hydrogen bonding in all four. C53 The AI writer's reading: only EC shows a band shift; for CAP, HPMCAS and HPMCP the evidence is a change in relative intensity within the drug's own doublet, and Figure 6D, labelled "LPV-P", is taken to be LPV–HPMCP. X10

From these data the authors give the "approximate trend" HPMCP > EC > HPMCAS for the strength of suppression among the hydrogen-bonded dispersions, "consistent with" HPMCP forming the stronger or more extensive interactions. C54 They relate it to chemistry: HPMCP has about 31% carboxylic acid-bearing phthalyl groups that can hydrogen-bond strongly with LPV, and LPV's phenyl rings "may also interact" with the phthalates by π–π stacking; the HF grade of HPMCAS has only about 7% succinoyl groups; EC has about 50% unreacted hydroxyl groups. C55·M2·M3·M5 EUD has no hydrogen-bond donor, which is "likely the cause" of its weak interaction, and LPV in it "presumably" keeps the chemical potential of pure amorphous drug. C57 Degree of substitution does not explain why CA differs from EC or CAP from HPMCP; polymer conformation and extra groups such as acetyl "might be" involved. C58 Water content does not explain the ranking either: HPMCAS, HPMCP and CAP dispersions held similar amounts of water and EC slightly less. C59·R12 The AI writer notes that the ranking holds at 33–85% DL but not at 15%, where EC (about 0.3) and HPMCP (about 0.6 μg/mL) cannot be told apart, and that CAP, which the authors count among the hydrogen-bonded systems, is left out of it although at 70 and 85% DL it suppresses more than EC and HPMCAS. X9

Lower supersaturation, slower crystallization

Table D. LPV concentration from 50% DL films in buffer without HPMC, read from Fig. 3 (approximate). R6

Without HPMC in the buffer, amorphous LPV alone peaked at about 10.8 μg/mL at 4 h and lost most of it by 8 h. R6.a The authors sort the dispersions into fast, intermediate and slow crystallization "in spite of differences in the degree of drug supersaturation": EUD and CAP are poor inhibitors whose supersaturation lasted about as long as that of amorphous LPV; HPMCAS, CA and EC "retained supersaturation for 12 h"; HPMCP did not crystallize during the 48 h, "showing a lower, but protracted level of supersaturation". R6·C25 Under polarized light, EUD and CAP particles had marked regions at 8 h and coloured crystals at 16 h, CA and EC had marked regions at 16 h, HPMCAS showed faint birefringence at 16 h, HPMCP showed nothing at 48 h, and amorphous LPV had become a fibrous mass by 8 h (Figure 4). R7·C26 The authors link the two ends: HPMCP, with the most suppression, "correlates" with the longest stability, and its dispersions were "more stable than HPMCAS dispersions", notable given how widely HPMCAS is used. C61 Note by the AI writer: the HPMCP curve runs at about 2.5–3.0 μg/mL and then 1.6–1.8 μg/mL, at or below the crystalline solubility measured with HPMCP (3.4 ± 0.5 μg/mL), so by the paper's own reference it is not supersaturated with respect to the crystal; several other dispersions also end below the Table 1 crystalline values, which the paper does not discuss. X6 The concentrations in Figure 3, taken without HPMC in the buffer, cannot be compared directly with Figure 2; the 50% HPMCP dispersion gives about 4.6 μg/mL at 48 h in one and about 1.8 μg/mL in the other. X7

Discussion

Three ways a polymer can sit with the drug. The authors' discussion sorts the results by where the polymer ends up. With a soluble polymer such as HPMC, the dispersion dissolves quickly, the drug exceeds its amorphous solubility and separates as an amorphous drug-rich phase, and the polymer, being very soluble, is "not substantially mixed with the drug droplets"; the ceiling is the amorphous solubility, whatever the loading. C48·C47·R3.a With a phase-separated dispersion (CA, EUD, the physical mixture), the drug sits in its own domains, behaves as pure amorphous drug and again reaches close to the amorphous solubility. C49·C50·C56 Only when drug and polymer form one phase and the polymer stays undissolved is the drug diluted by polymer and water, so that its chemical potential and its maximum concentration fall, most at low loading. C51·C67

Within the miscible group, Figure 10 pictures three patterns, holding the water contribution equal: strong interaction ($\chi$ below zero, not broken by water) suppresses more than ideal mixing; interaction like that within the pure components ($\chi$ near zero) suppresses by dilution alone; weak interaction ($\chi$ above zero) suppresses less, and at large $\chi$ the system separates and suppression nearly disappears. C51 The authors place HPMCP at the strong end. C54·C55 Their conclusion: the extent of supersaturation suppression "can be rationalized" by ASD composition, mixing state and the extent and strength of drug–polymer interactions. C68 This is a qualitative argument: no $\chi$ was determined, and, as noted above, the ideal line in Figure 2 is not the $\chi =0$ case of Eq. 4, so the placement of each polymer relative to "ideal" is a comparison with dilution by weight. X2·C46

What the authors propose for formulation. The paper ends with two applications. For ASD suspensions in preclinical studies, which must be dosed before they crystallize and ideally stay uncrystallized for days, crystallization is "expected" to depend on the extent of supersaturation, the solubility of the polymer in the medium and the polymer's ability to inhibit crystallization in solution and in the solid. C60 Lowering supersaturation lowers the crystallization tendency, so suppression "can be utilized as an advantage in the case of suspension formulations", and the HPMCP results illustrate it. C62·C61 For controlled release, a poorly soluble polymer lowers the drug's chemical potential in the drug–polymer–water system and with it the release rate and solubility; "in principle", choosing the polymer sets "an effective solubility somewhere between the amorphous and crystalline solubilities", so release rate and supersaturation "can be potentially tuned". C65 Designs that pair poorly soluble drugs with poorly soluble excipients should take these polymer and water effects into account. C66

Note by the AI writer (not discussed in the paper): what this means for an enteric ASD in the stomach. The paper's medium, pH 3, is close to gastric conditions for HPMCAS, HPMCP and CAP, which stay undissolved there. The data say that during that stage such an ASD may supersaturate far less than the drug's amorphous solubility suggests, and less the lower the drug loading and the stronger the interaction: at 15% DL, HPMCAS gave about a quarter of the amorphous solubility and HPMCP less than the crystalline solubility. R3.d·R3.g·R1.l That is good for avoiding crystallization in the stomach and a cost if early release matters. Two simple checks follow from the paper's own design: a physical-mixture control separates suppression by mixing from slow release, and a concentration measured at a fixed time should be shown to have stopped rising before it is read as a solubility. R14·X5

What stays open. The authors name three gaps: the Huang model ignores the change of $\chi$ with water and needs a $\chi$ that is hard to measure, so it is only qualitative; the differences between CA and EC and between CAP and HPMCP are unexplained; and no model exists for dispersions with water-soluble polymers. C46·C58·C47 Beyond these, the 48 h values were not shown to be equilibrium values, the amount of polymer that dissolved was not measured, the long-term stability of the most suppressed ASDs (EC and HPMCP at 15–33% DL) was predicted but not tested, and the work covers one drug at one pH and temperature. X5·X3·X8

Limitations

Limitations stated by the authors. The Huang model "is probably best applied in a qualitative manner": it does not include the effect of water on $\chi$, and $\chi$ is hard to determine. C46 It applies only to one-phase dispersions that stay one phase on hydration and whose polymer is insoluble enough for the solid's composition to stay put. C45 Polymer solubility could not be measured because most assay values were below detection; the authors estimate 3–25 μg/mL and a change of ± 2% in drug loading. R5·C52 The 7-day release data are not shown; the small drifts over that time "may be due to polymer degradation, in particular hydrolysis". R4·C23 For EUD dispersions at 70 and 85% DL, two close Tgs were "probable" but could not be resolved. C28

The AI writer's own judgement (1): the concentration data are solid; the mechanism is indirect. The direct results hold up: suppression by miscible insoluble polymers is seen at 3 h and 48 h in triplicate, it is absent with soluble HPMC, and the physical-mixture control shows that molecular mixing, not the polymer's presence, causes it. R2·R3·R14 What is indirect is the ranking by interaction strength. It rests on Tg, on infrared intensity changes that show a band shift only for EC, and on water sorption below an additive model; $\chi$ was never measured; CAP is left out of the ranking; and the "ideal" line used to sort strong from weak is proportional dilution rather than the model's $\chi =0$ case. X9·X10·X2 The paper's own wording, "approximate trend" and "can be rationalized", is the right strength for that part. C54·C68

The AI writer's own judgement (2): equilibrium and composition. Several curves (HPMCAS, CAP, HPMCP, EUD) rose considerably from 3 h to 48 h, which weakens the paper's use of the 3 h–48 h similarity as evidence that the undissolved solid kept its composition and leaves open whether 48 h is the ceiling. X4·X5 With 2 mg of solids per 10 mL, the ± 2% composition estimate holds only at the low end of the stated polymer solubility; at the high end an 85% DL film would lose most of its polymer. X3 The HPMCP dispersion that "did not crystallize" ran at or below the crystalline solubility, so its stability is not stability of a supersaturated solution. X6 Smaller points: Eq. 3 carries a sign misprint in its enthalpy term; the discussion's "Figs. 5 and 9" for hydrated DSC should point to the supplementary Fig. A; the CAP dispersion at 33% DL shows two Tgs dry and one after 97% RH, which sits uneasily with "did not alter the phase behavior"; the HPMC level behind the Table 1 amorphous solubility (170 μg/mL in the text, about 270 μg/mL by the film's stated composition) is unclear; and the preparation of the physical mixture is not described. X1·X11·X12·X14·X16

Conclusion

For lopinavir at pH 3, the paper shows that the highest concentration an ASD can reach depends on whether the polymer dissolves and whether it is mixed with the drug. Soluble HPMC, phase-separated CA and EUD dispersions and a physical mixture with HPMCAS reached close to the amorphous solubility of about 17 μg/mL; miscible dispersions with insoluble polymers fell below it, most at low loading and most with HPMCP, down to the crystalline solubility at 15–33% DL for EC and HPMCP. R3·R14·C63·C67 The authors rationalize the extent with a Flory–Huggins plus Gibbs–Duhem picture and indirect evidence of drug–polymer interaction, and link lower supersaturation to slower crystallization in suspension. C68·C61 The conclusion section puts it this way: molecular mixing with insoluble polymers "leads to reduced potential for supersaturation", while phase-separated ASDs show a supersaturation close to the pure drug's amorphous solubility. C67